AI Art Legality

AI Art Legality — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  Kernel-phase

  Kernel-phases are observable quantities used in high resolution astronomical imaging used for superresolution image creation. It can be seen as a generalization of closure phases for redundant arrays. For this reason, when the wavefront quality requirement are met, it is an alternative to aperture masking interferometry that can be executed without a mask while retaining phase error rejection properties. The observables are computed through linear algebra from the Fourier transform of direct images. They can then be used for statistical testing, model fitting, or image reconstruction. == Prerequisites == In order to extract kernel-phases from an image, some requirements must be met: Images are nyquist-sampled (at least 2 pixels per resolution element ( λ D {\displaystyle {\frac {\lambda }{D}}} )) Images are taken in near monochromatic light Exposure time is shorter than the timescale of aberrations Strehl ratio is high (good adaptive optics) Linearity of the pixel response (i.e. no saturation) Deviations from these requirements are known to be acceptable, but lead to observational bias that should be corrected by the observation of calibrators. == Definition == The method relies on a discrete model of the instrument's pupil plane and the corresponding list of baselines to provide corresponding vectors φ {\displaystyle \varphi } of pupil plane errors and Φ {\displaystyle \Phi } of image plane Fourier Phases. When the wavefront error in the pupil plane is small enough (i.e. when the Strehl ratio of the imaging system is sufficiently high), the complex amplitude associated to the instrumental phase in one point of the pupil φ k {\displaystyle \varphi _{k}} , can be approximated by e i φ k ≈ 1 + i φ k {\displaystyle e^{i\varphi _{k}}\approx 1+{\mathit {i}}\varphi _{k}} . This permits the expression of the pupil-plane phase aberrations φ {\displaystyle \varphi } to the image plane Fourier phase as a linear transformation described by the matrix A {\displaystyle A} : Φ = Φ 0 + A ⋅ φ {\displaystyle \Phi =\Phi _{0}+A\cdot \varphi } Where Φ 0 {\displaystyle \Phi _{0}} is the theoretical Fourier phase vector of the object. In this formalism, singular value decomposition can be used to find a matrix K {\displaystyle K} satisfying K ⋅ A = 0 {\displaystyle K\cdot A=0} . The rows of K {\displaystyle K} constitute a basis of the kernel of A T {\displaystyle A^{T}} . K ⋅ Φ = K ⋅ Φ 0 + K ⋅ A ⋅ φ {\displaystyle K\cdot \Phi =K\cdot \Phi _{0}+{\cancel {K\cdot A\cdot \varphi }}} The vector K . Φ {\displaystyle K.\Phi } is called the kernel-phase vector of observables. This equation can be used for model-fitting as it represents the interpretation of a sub-space of the Fourier phase that is immune to the instrumental phase errors to the first order. == Applications == The technique was first used in the re-analysis of archival images from the Hubble Space Telescope where it enabled the discovery of a number of brown dwarf in close binary systems. The technique is used as an alternative to aperture masking interferometry, especially for fainter stars because it does not require the use of masks that typically block 90% of the light, and therefore allows higher throughput. It is also considered to be an alternative to coronagraphy for direct detection of exoplanets at very small separations (below 2 λ D {\displaystyle 2{\frac {\lambda }{D}}} ) where coronagraphs are limited by the wavefront errors of adaptive optics. The same framework can be used for wavefront sensing. In the case of an asymmetric aperture, a pseudo-inverse of A {\displaystyle A} can be used to reconstruct the wavefront errors directly from the image. A Python library called xara is available on GitHub and maintained by Frantz Martinache to facilitate the extraction and interpretation of kernel-phases. The KERNEL project, has received funding from the European Research Council to explore the potential of these observables for a number of use-cases, including direct detection of exoplanets, image reconstruction, and image plane wavefront sensing for adaptive optics.

  Read more →
• 

  Model collapse

  Model collapse, also known by other names such as "AI inbreeding", "AI cannibalism", "Habsburg AI", and "model autophagy disorder" or "MAD" is a phenomenon noted in artificial intelligence studies, where machine learning models gradually degrade due to errors coming from uncurated synthetic data, or due to training on the outputs of another model such as prior versions of itself. It is unclear to what extent the phenomenon threatens the long-term development of such models, and some techniques have been proposed to mitigate the effect. == Characteristics == Shumailov et al. coined the term to describe two specific stages to the degradation of machine learning models: early model collapse and late model collapse: In early model collapse, the model begins losing information about the tails of the distribution – mostly affecting minority data. Later work highlighted that early model collapse is hard to notice, since overall performance may appear to improve, while the model loses performance on minority data. In late model collapse, the model loses a significant proportion of its performance, confusing concepts and losing most of its variance. == Mechanism == Using synthetic data as training data can lead to issues with the quality and reliability of the trained model. Model collapse occurs for three main reasons: functional approximation errors sampling errors learning errors Importantly, it happens in even the simplest of models, where not all of the error sources are present. In more complex models the errors often compound, leading to faster collapse. == Disagreement over real-world impact == Some researchers and commentators on model collapse warn that the phenomenon could fundamentally threaten future generative AI development: As AI-generated data is shared on the Internet, it will inevitably end up in future training datasets, which are often crawled from the Internet. If training on "slop" (large quantities of unlabeled synthetic data) inevitably leads to model collapse, this could therefore pose a difficult problem. However, recently, other researchers have disagreed with this argument, showing that if synthetic data accumulates alongside human-generated data, model collapse is avoided. The researchers argue that data accumulating over time is a more realistic description of reality than deleting all existing data every year, and that the real-world impact of model collapse may not be as catastrophic as feared. An alternative branch of the literature investigates the use of machine learning detectors and watermarking to identify model generated data and filter it out. == Mathematical models of the phenomenon == === 1D Gaussian model === In 2024, a first attempt has been made at illustrating collapse for the simplest possible model — a single dimensional normal distribution fit using unbiased estimators of mean and variance, computed on samples from the previous generation. To make this more precise, we say that original data follows a normal distribution X 0 ∼ N ( μ , σ 2 ) {\displaystyle X^{0}\sim {\mathcal {N}}(\mu ,\sigma ^{2})} , and we possess M 0 {\displaystyle M_{0}} samples X j 0 {\displaystyle X_{j}^{0}} for j ∈ { 1 , … , M 0 } {\displaystyle j\in {\{\,1,\dots ,M_{0}\,{}\}}} . Denoting a general sample X j i {\displaystyle X_{j}^{i}} as sample j ∈ { 1 , … , M i } {\displaystyle j\in {\{\,1,\dots ,M_{i}\,{}\}}} at generation i {\displaystyle i} , then the next generation model is estimated using the sample mean and variance: μ i + 1 = 1 M i ∑ j X j i ; σ i + 1 2 = 1 M i − 1 ∑ j ( X j i − μ i + 1 ) 2 . {\displaystyle \mu _{i+1}={\frac {1}{M_{i}}}\sum _{j}X_{j}^{i};\quad \sigma _{i+1}^{2}={\frac {1}{M_{i}-1}}\sum _{j}(X_{j}^{i}-\mu _{i+1})^{2}.} Leading to a conditionally normal next generation model X j i + 1 | μ i + 1 , σ i + 1 ∼ N ( μ i + 1 , σ i + 1 2 ) {\displaystyle X_{j}^{i+1}|\mu _{i+1},\;\sigma _{i+1}\sim {\mathcal {N}}(\mu _{i+1},\sigma _{i+1}^{2})} . In theory, this is enough to calculate the full distribution of X j i {\displaystyle X_{j}^{i}} . However, even after the first generation, the full distribution is no longer normal: It follows a variance-gamma distribution. To continue the analysis, instead of writing the probability density function at each generation, it is possible to explicitly construct them in terms of independent random variables using Cochran's theorem. To be precise, μ 1 {\displaystyle \mu _{1}} and σ 1 {\displaystyle \sigma _{1}} are independent, with μ 1 ∼ N ( μ , σ 2 M 0 ) {\displaystyle \mu _{1}\sim {\mathcal {N}}\left(\mu ,{\frac {\sigma ^{2}}{M_{0}}}\right)} and ( M 0 − 1 ) σ 1 2 ∼ σ 2 Γ ( M 0 − 1 2 , 1 2 ) {\displaystyle (M_{0}-1)\,\sigma _{1}^{2}\sim \sigma ^{2}\,\Gamma \left({\frac {M_{0}-1}{2}},{\frac {1}{2}}\right)} , following a Gamma distribution. Denoting with Z {\displaystyle Z} Gaussian random variables distributed according to N ( 0 , 1 ) {\displaystyle {\mathcal {N}}(0,1)} and with S i {\displaystyle S^{i}} random variables distributed with 1 M i − 1 − 1 Γ ( M i − 1 − 1 2 , 1 2 ) {\displaystyle {\frac {1}{M_{i-1}-1}}\Gamma \left({\frac {M_{i-1}-1}{2}},{\frac {1}{2}}\right)} , it turns out to be possible to write samples at each generation as X j 0 = μ + σ Z j 0 , {\textstyle X_{j}^{0}=\mu +\sigma Z_{j}^{0},} X j 1 = μ + σ M 0 Z 1 + σ S 1 Z j 1 , {\textstyle X_{j}^{1}=\mu +{\frac {\sigma }{\sqrt {M_{0}}}}Z^{1}+\sigma {\sqrt {S^{1}}}Z_{j}^{1},} and more generally X j n = μ + σ M 0 Z 1 + σ M 1 S 1 Z 2 + ⋯ + σ M n − 1 S 1 × ⋯ × S n − 1 Z n + σ S 1 × ⋯ × S n Z j n . {\displaystyle X_{j}^{n}=\mu +{\frac {\sigma }{\sqrt {M_{0}}}}Z^{1}+{\frac {\sigma }{\sqrt {M_{1}}}}{\sqrt {S^{1}}}Z^{2}+\dots +{\frac {\sigma }{\sqrt {M_{n-1}}}}{\sqrt {S^{1}\times \dots \times S^{n-1}}}Z^{n}+\sigma {\sqrt {S^{1}\times \dots \times S^{n}}}Z_{j}^{n}.} Note, that these are not joint distributions, as Z n {\displaystyle Z^{n}} and S n {\displaystyle S^{n}} depend directly on Z j n − 1 {\displaystyle Z_{j}^{n-1}} , but when considering X j n {\displaystyle X_{j}^{n}} on its own the formula above provides all the information about the full distribution. To analyse the model collapse, we can first calculate variance and mean of samples at generation n {\displaystyle n} . This would tell us what kind of distributions we expect to arrive at after n {\displaystyle n} generations. It is possible to find its exact value in closed form, but the mean and variance of the square root of gamma distribution are expressed in terms of gamma functions, making the result quite clunky. Following, it is possible to expand all results to second order in each of 1 / M i {\displaystyle 1/M_{i}} , assuming each sample size to be large. It is then possible to show that 1 σ 2 Var ⁡ ( X j n ) = 1 M 0 + 1 M 1 + ⋯ + 1 M n − 1 + 1 + O ( M i − 2 ) . {\displaystyle {\frac {1}{\sigma ^{2}}}\operatorname {Var} (X_{j}^{n})={\frac {1}{M_{0}}}+{\frac {1}{M_{1}}}+\dots +{\frac {1}{M_{n-1}}}+1+{\mathcal {O}}\left(M_{i}^{-2}\right).} And if all sample sizes M i = M {\displaystyle M_{i}=M} are constant, this diverges linearly as n → ∞ {\displaystyle n\to \infty } : Var ⁡ ( X j n ) = σ 2 ( 1 + n M ) ; E ( X j n ) = μ . {\displaystyle \operatorname {Var} (X_{j}^{n})=\sigma ^{2}\left(1+{\frac {n}{M}}\right);\quad \mathbb {E} (X_{j}^{n})=\mu .} This is the same scaling as for a single dimensional Gaussian random walk. However, divergence of the variance of X j n {\displaystyle X_{j}^{n}} does not directly provide any information about the corresponding estimates of μ n + 1 {\displaystyle \mu _{n+1}} and σ n + 1 {\displaystyle \sigma _{n+1}} , particularly how different they are from the original μ {\displaystyle \mu } and σ {\displaystyle \sigma } . It turns out to be possible to calculate the distance between the true distribution and the approximated distribution at step n + 1 {\displaystyle n+1} , using the Wasserstein-2 distance (which is also sometimes referred to as risk): E [ W 2 2 ( N ( μ , σ 2 ) , N ( μ n + 1 , σ n + 1 2 ) ) ] = 3 2 σ 2 ( 1 M 0 + 1 M 1 + ⋯ + 1 M n ) + O ( M i − 2 ) , {\displaystyle \mathbb {E} \left[\mathbb {W} _{2}^{2}\left({\mathcal {N}}(\mu ,\sigma ^{2}),{\mathcal {N}}(\mu _{n+1},\sigma _{n+1}^{2})\right)\right]={\frac {3}{2}}\sigma ^{2}\left({\frac {1}{M_{0}}}+{\frac {1}{M_{1}}}+\dots +{\frac {1}{M_{n}}}\right)+{\mathcal {O}}\left(M_{i}^{-2}\right),} Var ⁡ [ W 2 2 ( N ( μ , σ 2 ) , N ( μ n + 1 , σ n + 1 2 ) ) ] = 1 2 σ 4 ( 3 M 0 2 + 3 M 1 2 + ⋯ + 3 M n 2 + ∑ i ≠ j 4 M i M j ) + O ( M i − 3 ) . {\displaystyle \operatorname {Var} \left[\mathbb {W} _{2}^{2}\left({\mathcal {N}}(\mu ,\sigma ^{2}),{\mathcal {N}}(\mu _{n+1},\sigma _{n+1}^{2})\right)\right]={\frac {1}{2}}\sigma ^{4}\left({\frac {3}{M_{0}^{2}}}+{\frac {3}{M_{1}^{2}}}+\dots +{\frac {3}{M_{n}^{2}}}+\sum _{i\neq j}{\frac {4}{M_{i}M_{j}}}\right)+{\mathcal {O}}\left(M_{i}^{-3}\right).} This directly shows why model collapse occurs in this simple model. Due to errors from re-sampling the approximated distribution, each generation ends up corresponding to a

  Read more →
• 

  YouNoodle

  YouNoodle, Inc. is a San Francisco-based company, with offices in Barcelona and Santiago, founded in 2010, building a platform for entrepreneurship competitions all over the world. YouNoodle matches entrepreneurs with competitions, accelerators, and startup programs, and provides a judging and voting SaaS platform to university, non-profit, government and enterprise clients organizing innovation challenges and competitions. Stanford's BASES, UC Berkeley LAUNCH, Start-Up Chile, Amazon Startup Challenge, and NASA are all running one or more competitions on YouNoodle's platform. == History and structure == YouNoodle was founded by Rebeca Hwang and Torsten Kolind in 2010. The company was spun off a project started by Bob Goodson (Quid) and Kirill Makharinsky (Enki) in 2007 with support from Peter Thiel (Founders Fund), Max Levchin (PayPal) and Charles Lho (Amicus Group), founding investor and Chairman of YouNoodle today. This project also spawned Quid (Goodson) and indirectly Ostrovok (Makharinsky). Although also named YouNoodle, this project/company was discontinued in 2010, when the three new entities started operations. The founders of the 2007-2010 entity were Goodson and Makharinsky, both former students of the University of Oxford. Goodson had studied medieval English literature before moving from Oxford to California when Levchin, the co-founder of PayPal, invited him to join a start-up there. Makharinsky's degree was in applied mathematics, and he was also encouraged to pursue opportunities in the United States by Levchin. Other significant employees included Hwang (co-founder of today's YouNoodle), a Stanford University doctoral student whose research is into social network theory. == Startup predictor == YouNoodle's now discontinued "Startup predictor", part of the 2007-2010 entity and developed by Makharinsky and Hwang, used mathematical models to predict the success of new businesses. The user fills in a questionnaire, which takes about half an hour to complete and concentrates on the business concept, finances, founders and advisers. Because the procedure was designed for new companies, questions on revenue and traffic are not included. The site then provided an estimate of what the company's value will be after three years and a score from 1 to 1000 representing its value as an investment. The service was free for the startups themselves, but YouNoodle intended to charge third parties for access to the results. The level of detail required by the questionnaire makes it difficult for people without inside knowledge of a company to provide the data for a prediction on their own. The company's founders have declined to explain the algorithm in detail, but state that it takes into account the entrepreneurs' experience, networks and mutual relations. Information provided by companies which use the site's networking features is used to improve the algorithm. As of August 2008, the algorithm was based on data from 3,000 startups. In the same month the company had four patents pending on the technology.

  Read more →
• 

  Deepfake

  Deepfakes (a portmanteau of 'deep learning' and 'fake') are images, videos, or audio that have been edited or generated using artificial intelligence, AI-based tools or audio-video editing software. They may depict real or fictional people and are considered a form of synthetic media, that is media that is usually created by artificial intelligence systems by combining various media elements into a new media artifact. While the act of creating fake content is not new, deepfakes uniquely leverage machine learning and artificial intelligence techniques, including facial recognition algorithms and artificial neural networks such as variational autoencoders and generative adversarial networks (GANs). In turn, the field of image forensics has worked to develop techniques to detect manipulated images. Deepfakes have garnered widespread attention for their potential use in creating child sexual abuse material, celebrity pornographic videos, revenge porn, fake news, hoaxes, bullying, and financial fraud. Academics have raised concerns about the potential for deepfakes to promote disinformation and hate speech, as well as interfere with elections. In response, the information technology industry and governments have proposed recommendations and methods to detect and mitigate their use. Academic research has also delved deeper into the factors driving deepfake engagement online as well as potential countermeasures to malicious application of deepfakes. From traditional entertainment to gaming, deepfake technology has evolved to be increasingly convincing and available to the public, allowing for the disruption of the entertainment and media industries. == History == Photo manipulation was developed in the 19th century and soon applied to motion pictures. Technology steadily improved during the 20th century, and more quickly with the advent of digital video. Deepfake technology has been developed by researchers at academic institutions beginning in the 1990s, and later by amateurs in online communities. More recently, the methods have been adopted by industry. The development of generative adversarial networks (GANs) in the mid-2010s represented a key technical turning point in the evolution of deepfakes. GANs allowed for the creation of highly realistic fake images and videos by training competing neural networks, achieving a much improved visual fidelity over previous methods of creating the content using rules or by using autoencoders, and formed the basis for modern deepfake methods. === Academic research === Academic research related to deepfakes is split between the field of computer vision, a sub-field of computer science, which develops techniques for creating and identifying deepfakes, and humanities and social science approaches that study the social, ethical, aesthetic implications as well as journalistic and informational implications of deepfakes. As deepfakes have risen in prominence in popularity with innovations provided by AI tools, significant research has gone into detection methods and defining the factors driving engagement with deepfakes on the internet. Deepfakes have been shown to appear on social media platforms and other parts of the internet for purposes ranging from entertainment and education related to deepfakes to misinformation to elicit strong reactions. There are gaps in research related to the propagation of deepfakes on social media. Negativity and emotional response are the primary driving factors for users sharing deepfakes. === Social science and humanities approaches to deepfakes === In cinema studies, deepfakes illustrate how "the human face is emerging as a central object of ambivalence in the digital age". Video artists have used deepfakes to "playfully rewrite film history by retrofitting canonical cinema with new star performers". Film scholar Christopher Holliday analyses how altering the gender and race of performers in familiar movie scenes destabilizes gender classifications and categories. The concept of "queering" deepfakes is also discussed in Oliver M. Gingrich's discussion of media artworks that use deepfakes to reframe gender, including British artist Jake Elwes' Zizi: Queering the Dataset, an artwork that uses deepfakes of drag queens to intentionally play with gender. The aesthetic potentials of deepfakes are also beginning to be explored. Theatre historian John Fletcher notes that early demonstrations of deepfakes are presented as performances, and situates these in the context of theater, discussing "some of the more troubling paradigm shifts" that deepfakes represent as a performance genre. While most English-language academic studies of deepfakes focus on the Western anxieties about disinformation and pornography, digital anthropologist Gabriele de Seta has analyzed the Chinese reception of deepfakes, which are known as huanlian, which translates to "changing faces". The Chinese term does not contain the "fake" of the English deepfake, and de Seta argues that this cultural context may explain why the Chinese response has centered on practical regulatory measures to "fraud risks, image rights, economic profit, and ethical imbalances". === Computer science research on deepfakes === A landmark early project was the "Video Rewrite" program, published in 1997. The program modified existing video footage of a person speaking to depict that person mouthing the words from a different audio track. It was the first system to fully automate this kind of facial reanimation, and it did so using machine learning techniques to make connections between the sounds produced by a video's subject and the shape of the subject's face. Contemporary academic projects have focused on creating more realistic videos and improving deepfake techniques. The "Synthesizing Obama" program, published in 2017, modifies video footage of former president Barack Obama to depict him mouthing the words contained in a separate audio track. The project lists as a main research contribution to its photorealistic technique for synthesizing mouth shapes from audio. The "Face2Face" program, published in 2016, modifies video footage of a person's face to depict them mimicking another person's facial expressions. The project highlights its primary research contribution as the development of the first method for re-enacting facial expressions in real time using a camera that does not capture depth, enabling the technique to work with common consumer cameras. Researchers have also shown that deepfakes are expanding into other domains such as medical imagery. In this work, it was shown how an attacker can automatically inject or remove lung cancer in a patient's 3D CT scan. The result was so convincing that it fooled three radiologists and a state-of-the-art lung cancer detection AI. To demonstrate the threat, the authors successfully performed the attack on a hospital in a White hat penetration test. A survey of deepfakes, published in May 2020, provides a timeline of how the creation and detection of deepfakes have advanced over the last few years. The survey identifies that researchers have been focusing on resolving the following challenges of deepfake creation: Generalization. High-quality deepfakes are often achieved by training on hours of footage of the target. This challenge is to minimize the amount of training data and the time to train the model required to produce quality images and to enable the execution of trained models on new identities (unseen during training). Paired Training. Training a supervised model can produce high-quality results, but requires data pairing. This is the process of finding examples of inputs and their desired outputs for the model to learn from. Data pairing is laborious and impractical when training on multiple identities and facial behaviors. Some solutions include self-supervised training (using frames from the same video), the use of unpaired networks such as Cycle-GAN, or the manipulation of network embeddings. Identity leakage. This is where the identity of the driver (i.e., the actor controlling the face in a reenactment) is partially transferred to the generated face. Some solutions proposed include attention mechanisms, few-shot learning, disentanglement, boundary conversions, and skip connections. Occlusions. When part of the face is obstructed with a hand, hair, glasses, or any other item then artifacts can occur. A common occlusion is a closed mouth which hides the inside of the mouth and the teeth. Some solutions include image segmentation during training and in-painting. Temporal coherence. In videos containing deepfakes, artifacts such as flickering and jitter can occur because the network has no context of the preceding frames. Some researchers provide this context or use novel temporal coherence losses to help improve realism. As the technology improves, the interference is diminishing. Overall, deepfakes are expected to have several implications in media and society, med

  Read more →
• 

  Class activation mapping

  Class activation mapping methods are explainable AI (XAI) techniques used to visualize the regions of an input image that are the most relevant for a particular task, especially image classification, in convolutional neural networks (CNNs). These methods generate heatmaps by weighting the feature maps from a convolutional layer according to their relevance to the target class. In the field of artificial intelligence, generically defined as "the effort to automate intellectual tasks normally performed by humans", machine learning and deep learning were created. They both use statistical and computational methods to learn patterns from data, reducing the need for manually coded rules. Machine learning models are trained on input data and the known respective answers, learning the underlying patterns or structures present in the data. Traditional Machine learning algorithms employ manually designed feature sets, posing a direct link between machine learning designers and employed features. Deep learning is a subfield of machine learning, based on the concept of successive layers of representation, in which the data is progressively unfolded in different ways, to extract relevant and informative patterns in data analysis. Deep learning algorithms are defined as feature learning algorithms automatically learning hierarchical feature representations from raw data, extracting increasingly abstract features through multiple layers. CNNs are a specific architecture of deep learning models, designed to process spatially structured data, such as images, exploiting a series of convolution, non-linear activation and pooling operations to extract relevant features, contained in the so-called feature maps from input data. CNNs have demonstrated to be highly effective in a variety of computer vision and image processing tasks. CNNs (and deep learning models more broadly) are described as black boxes due to their complex and non-transparent internal layers of representation. The need for clearer indications on its internal working and decision-making process gave birth to XAI techniques. Among the proposed XAI techniques for computer vision tasks, Class activation mapping methods can show which pixels in an input image are important to the predicted logit for a class of interest, in a classification task. Class activation mapping methods were originally developed for class-discriminative scenarios to visualize which parts of the input image influenced the classification decision, namely to visually highlight the regions of those feature maps that contribute most strongly to the prediction of a given class. More advanced versions of these methods are not limited to image classification tasks, but have been extended also to several vision-related tasks, such as object detection, image captioning, visual question answering and image segmentation. == Background == The following methods laid the groundwork for the class activation maps approaches, forming the conceptual basis of using gradients to highlight class-discriminative regions. === Class model visualization and saliency maps for convolutional neural networks === The class model visualization and image-specific saliency maps approaches have been presented in the foundational work "Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps" by Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman and it generalizes the deconvnet method by Zeiler and Fergus. Class model visualization synthesizes an artificial input image that strongly activates the output neurons associated with a target class. Given a trained, fixed model, this method starts with a zero-initialized image, backpropagates the gradients from the class score to the image pixels, updates the image pixels increasing the specific class scores and it repeats the pixel updating process, showing an encoded (idealized version) prototype of the class of interest. Image-specific class saliency visualization method provides a visual explanation by highlighting the most relevant pixels in an image for predicting a certain class C of interest. This is done by computing the gradient of the class score with respect to the input image, I 0 , {\displaystyle I_{0},} w = ∂ S C ∂ I | I 0 {\displaystyle w=\left.{\frac {\partial S_{C}}{\partial I}}\right|_{I_{0}}} approximating the model locally (around I 0 {\displaystyle I_{0}} ) as linear, using a first-order Taylor expansion: S C ( I ) ≈ w C T I + b {\displaystyle S_{C}(I)\approx w_{C}^{T}I+b} . The magnitude of w C {\displaystyle w_{C}} , the gradient, indicates the importancy of the pixels: larger gradients suggest greater influence on the prediction. Once the gradient is known, the saliency map is defined as the maximum absolute gradient across the color channels: M i j = m a x C | ∂ S C ∂ I i j C | {\displaystyle M_{ij}=max_{C}\left|{\frac {\partial S_{C}}{\partial I_{ij}^{C}}}\right|} resulting in an saliency map (i.e. heatmap). === Guided backpropagation === The concept of guided backpropagation can be traced for the first time in the paper by Springenberg et al. "Striving For Simplicity: The All Convolutional Net" and also this method builds upon the work by Zeiler and Fergus "Visualizing and Understanding Convolutional Networks". Guided backpropagation core is to understand what a CNN is learning, by visualizing the patterns that activate more strongly individual neurons (or filters), in architectures which do not rely on max-pooling layer. When propagating gradients back through a rectified linear unit (ReLU), guided backpropagation passes the gradient if and only if the input to the ReLU was positive (forward pass) and the output gradient is positive (backward signal), tackling both inactive neurons, negative gradients and suppressing the noise. The result displays sharper, high-resolution visualizations of what each neuron is responding to. Guided backpropagation represents a simple and practical method for model interpretability, helping understand how and where neural networks detect semantic concepts across layers. Moreover, it can be applied to any network architecture, due to its working principle. == Base versions == Class activation mapping and gradient-weighted class activation mapping are the original and most widely used methods for visual explanations in convolutional neural networks. These methods serve as the foundation for many later developments in explainable AI. Notation: In this article, the symbols i and j represent integer indices that disappear inside sums or averages, while x and y are the continuous (or up-sampled integer) coordinates of the final heat-map that is plotted. === Class activation mapping (CAM) === Class activation mapping (CAM) was the first, and the original, version of CAM methods, and it gave the name to the whole category. The approach was firstly introduced by Zhou et al. in their seminal work "Learning Deep Features for Discriminative Localization". This approach achieves class-specific heatmaps by modifying image classification CNN architectures, replacing fully-connected layers with convolutional layers and a final global average pooling layer. Its main scope is to localize and highlight discriminative regions of an input image that a CNN uses to identify a particular class, without needing explicit bounding box annotations. ==== Global average pooling (GAP) ==== Global average pooling (GAP) represents the key element in the original CAM approach. It is a dimensionality reduction technique and, similarly to other pooling layers, it allows the downsampling of the feature maps, calculating representative values for a specific region of the feature map. The particularity of GAP is that it calculates a single value for an entire feature map, significantly reducing the model dimensions. ==== Mathematical description ==== The mathematical description considers as its key the combination of convolutional and GAP layers. In CAM, it is mandatory to have the GAP layer after the last convolutional layer and before the final linear classifier layer. This last element of the architecture connects the output logits (the network predictions) y C {\displaystyle y^{C}} , to the GAP values, with its respective fine-tuned weights, w k C {\displaystyle w_{k}^{C}} . Considering A k {\displaystyle A^{k}} as the last feature maps of the last convolutional layer, GAP produces one value for each feature map, by averaging all the matrix elements (i, j) of the feature map: F k = 1 m n ∑ i = 1 m ∑ j = 1 n A i j k {\displaystyle F^{k}={\frac {1}{mn}}\sum _{i=1}^{m}\sum _{j=1}^{n}A_{ij}^{k}} with A k = [ A 11 k A 12 k ⋯ A 1 n k A 21 k A 22 k ⋯ A 2 n k ⋮ ⋮ ⋱ ⋮ A m 1 k A m 2 k ⋯ A m n k ] = { A i j k ∣ 1 ≤ i ≤ m , 1 ≤ j ≤ n } {\displaystyle A^{k}={\begin{bmatrix}A_{11}^{k}&A_{12}^{k}&\cdots &A_{1n}^{k}\\A_{21}^{k}&A_{22}^{k}&\cdots &A_{2n}^{k}\\\vdots &\vdots &\ddots &\vdots \\A_{m1}^{k}&A_{m2}^{k}&\cdots &A_{mn}^{k}\end{bmatrix}}=\left\{A_{

  Read more →
• 

  Vehicle infrastructure integration

  The Vehicle Infrastructure Integration (VII), also known as "Connected Roadways" or "vehicle-to-everything" (V2X) technology, is a United States Department of Transportation initiative that aims to improve road safety by developing technology that connects road vehicles with their environment. This development draws on several disciplines, including transport engineering, electrical engineering, automotive engineering, telematics, and computer science. Although VII specifically covers road transport, similar technologies are under development for other modes of transport. For example, airplanes may use ground-based beacons for automated guidance, allowing the autopilot to fly the plane without human intervention. == Goals == The goal of VII is to establish a communication link between vehicles (via On-Board Equipment, or OBE) and roadside infrastructure (via Roadside Equipment, or RSE) to enhance the safety, efficiency, and convenience of transportation systems. Two potential approaches are the widespread deployment of a dedicated short-range communications (DSRC) link on the 5.9GHz band, and cellular communication (C-V2X). Either of these methods would allow vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. The initiative has three priorities: Stakeholder evaluation and acceptance of the business model and its deployment schedule, Validation of the technology, with a focus on communications systems, in relation to deployment costs, and Creation of legal structures and policies, especially concerning digital privacy, to improve the system's long-term potential for success. === Safety === Current automotive safety technology relies primarily on vehicle-based radar, lidar, and sonar systems. This technology allows, for instance, a potential reduction in rear-end collisions by monitoring obstacles in front of or behind the vehicle and automatically applying the brakes when necessary. This technology, however, is limited by the sensing range of vehicle-based radar, particularly in angled and left-turn collisions, such as a motorist losing control of the vehicle during an impending head-on collision. The rear-end collisions addressed by current technology are generally less severe than angled, left-turn, or head-on collisions. VII promotes the development of a direct communication link between road vehicles and all other vehicles nearby, allowing for the exchange of information on vehicle speed and orientation or driver awareness and intent. This real-time exchange of information may enable more effective automated emergency maneuvers, such as steering, decelerating, or braking. In addition to nearby vehicle awareness, VII promotes a communication link between vehicles and roadway infrastructure. Such a link may allow for improved real-time traffic information, better queue management, and feedback to vehicles. Existing implementations of VII use vehicle-based sensors that can recognize and respond to roadway markings or signs, automatically adjusting vehicle parameters to follow the recognized instructions. However, this information may also be acquired via roadside beacons or stored in a centralized database accessible to all vehicles. === Efficiency === With a VII system in place, vehicles will be linked together. The headway between vehicles may therefore be reduced so that there is less empty space on the road, increasing the available capacity per lane. More capacity per lane will in turn imply fewer lanes in general, possibly satisfying the community's concerns about the impact of roadway widening. VII will enable precise traffic-signal coordination by tracking vehicle platoons and will benefit from accurate timing by drawing on real-time traffic data covering volume, density, and turning movements. Real-time traffic data can also be used in the design of new roadways or modification of existing systems as the data could be used to provide accurate origin-destination studies and turning-movement counts for uses in transportation forecasting and traffic operations. Such technology would also lead to improvements for transport engineers to address problems whilst reducing the cost of obtaining and compiling data. Tolling is another prospect for VII technology as it could enable roadways to be automatically tolled. Data could be collectively transmitted to road users for in-vehicle display, outlining the lowest cost, shortest distance, and/or fastest route to a destination on the basis of real-time conditions. === Existing applications === To some extent, results along these lines have been achieved in trials performed around the globe, making use of GPS, mobile phone signals, and vehicle registration plates. GPS is becoming standard in many new high-end vehicles and is an option on most new low- and mid-range vehicles. In addition, many users also have mobile phones that transmit trackable signals (and may also be GPS-enabled). Mobile phones can already be traced for purposes of emergency response. GPS and mobile phone tracking, however, do not provide fully reliable data. Furthermore, integrating mobile phones in vehicles may be prohibitively difficult. Data from mobile phones, though useful, might even increase risks to motorists as they tend to look at their phones rather than concentrate on their driving. Automatic registration plate recognition can provide large quantities of data, but continuously tracking a vehicle through a corridor is a difficult task with existing technology. Today's equipment is designed for data acquisition and functions such as enforcement and tolling, not for returning data to vehicles or motorists for response. GPS will nevertheless be one of the key components in VII systems. == Limitations == === Privacy === VII architecture is designed to prevent identification of individual vehicles, with all data exchange between the vehicle and the system occurring anonymously. Exchanges between the vehicles and third parties such as OEMs and toll collectors will occur, but the network traffic will be sent via encrypted tunnels and will therefore not be decipherable by the VII system. Data sharing with law enforcement or Homeland Security was not included in system design as of 2006. === Technical issues === ==== Coordination ==== A major issue facing the deployment of VII is the problem of how to set up the system initially. The costs associated with installing the technology in vehicles and providing communications and power at every intersection are significant. ==== Maintenance ==== Another factor for consideration in regard to the technology's distribution is how to update and maintain the units. Traffic systems are highly dynamic, with new traffic controls implemented every day and roadways constructed or repaired every year. The vehicle-based option could be updated via the internet (preferably wireless) but may subsequently require all users to have access to internet technology. Alternatively, if receivers were placed in all vehicles and the VII system was primarily located along the roadside, information could be stored in a centralized database. This would allow the agency responsible to issue updates at any time. These would then be disseminated to the roadside units for passing motorists. Operationally, this method is currently considered to provide the greatest effectiveness but at a high cost to the authorities. ==== Security ==== Security of the units is another concern, especially in light of the public acceptance issue. Criminals could tamper, remove, or destroy VII units regardless of whether they are installed inside vehicles or along the roadside. Magnets, electric shocks, and malicious software (viruses, hacking, or jamming) could be used to damage VII systems – regardless of whether units are located inside vehicle or along the roadside. == Recent developments == Much of the current research and experimentation is conducted in the United States where coordination is ensured through the Vehicle Infrastructure Integration Consortium; consisting of automobile manufacturers (Ford, General Motors, Daimler Chrysler, Toyota, Nissan, Honda, Volkswagen, BMW), IT suppliers, U.S. Federal and state transportation departments, and professional associations. Trialing is taking place in Michigan and California. The specific applications now being developed under the U.S. initiative are: Warning drivers of unsafe conditions or imminent collisions. Warning drivers if they are about to run off the road or speed around a curve too fast. Informing system operators of real-time congestion, weather conditions and incidents. Providing operators with information on corridor capacity for real-time management, planning and provision of corridor-wide advisories to drivers. In mid-2007, a VII environment covering some 20 square miles (52 km2) near Detroit was used to test 20 prototype VII applications. Several automobile manufacturers are also conducting their own VII research and triali

  Read more →
• 

  Tilly Norwood

  Tilly Norwood is a character created using generative artificial intelligence in 2025 by Xicoia, the AI division of Particle6 Group, a production company founded by Eline Van der Velden. "AI Commissioner", the first project to feature the Norwood character, was criticised by reviewers for The Guardian, PC Gamer, and The A.V. Club. A press release that talent agencies expressed interest in representing the character attracted strong criticism from Hollywood actors and firms, prompting allegations of personality rights violations and arguments over the impact of the character on production costs in the media industry. == History == Norwood was created by Xicoia, which was founded in February 2025 as the artificial intelligence (AI) division of Particle6, a production company founded by Dutch actress and producer Eline Van der Velden in 2015. Van der Velden had previously starred in a satirical comedy series for BBC Three based around her character Miss Holland, whom she created in 2012 as a parody of beauty standards. She stated that the process of creating Norwood took "a long time" and compared the process to that of writers creating characters. An Instagram account under Norwood's name, with posts dating back to 6 May 2025, had gained 50,000 followers by October 3, and featured AI-generated modelling shots, selfies, and epic film scenes. Van der Velden stated in July 2025 that she intended Norwood to be the next Scarlett Johansson or Natalie Portman and later said that audiences were more interested in a film's story than whether its actors were real. Particle6 has claimed that using Norwood could cut production costs by 90%. On 30 July 2025, a comedy sketch named "AI Commissioner" was released, featuring Norwood as an "actress" along with other AI-generated characters. It was created with ten AI software tools, with a script generated by ChatGPT. Stuart Heritage of The Guardian described it as technically competent but "relentlessly unfunny to watch", with "sloppily written, woodenly delivered dialogue", and that Norwood's teeth kept "blurring into a single white block." Joshua Wolens of PC Gamer wrote that Norwood's exaggerated mouth movements gave the impression "that her skeleton was about to leave her body", while William Hughes of The A.V. Club wrote that the sketch's attempt at mimicking human body and mouth movements produced "such a hideous uncanny valley effect" that it gave them "a full-on case of the screaming fantods". By October 2, the sketch had been viewed more than 700,000 times on YouTube. Xicoia was officially announced on 27 September 2025, at the Zurich Summit, part of the Zurich Film Festival; there, van der Velden unveiled Norwood and later joined a panel with Verena Puhm, head of Luma AI's Studio Dream Lab LA. They suggested that media companies were quietly embracing AI and that public announcements of AI-generated works were imminent. Van der Velden claimed that studios had dropped their objections by May after being opposed in February, and that multiple talent agencies were considering representing Norwood. The latter claim drew heightened attention to the character and was printed as fact by Deadline under the headline "Talent Agents Circle AI Actress Tilly Norwood." The report caused controversy, with Vulture describing the reaction to it as "Hollywood [lurching] into a fresh wave of existential panic" while being critical of Deadline's reporting, writing that "when Deadline called it a 'revelation' and published the supposed interest as fact without verification, [it] metastasized into a full-fledged cyberpunk news cycle", and that "by Tuesday, it had grown like wildfire." By September 2025, AI-generated videos had been released depicting Norwood on a red carpet, crying on the sofa of The Graham Norton Show, and starring in mock trailers for sci-fi, fantasy, horror, and action films. Later that month, actresses Melissa Barrera, Kiersey Clemons, and Natasha Lyonne suggested boycotting any agency who signed Norwood, while Mara Wilson asked why none of the "hundreds of living young women whose faces were composited together" to create Norwood could be hired instead. Also around this time, Emily Blunt described Norwood as "really, really scary", and Sophie Turner, Toni Collette, Ralph Ineson, and Ariel Winter also expressed disapproval, while Lukas Gage, Odessa A'zion, and Trace Lysette joked about having supposedly worked with Norwood and finding her incompetent and unpleasant to work with, with Gage claiming that "She was a nightmare to work with!" and "She couldn't hit her mark and she was late!" and Lysette adding "She cut me in line at lunch one day and didn't even say excuse me. She won't get far." Jenelle Riley, Nicholas Alexander Chavez, and the American union SAG-AFTRA stated that they do not consider Norwood an actress. The Gersh Agency and WME both announced that they would not sign Norwood. Whoopi Goldberg and Charlie Fink expressed scepticism that AI could replace jobs. Esquire UK reported that a post on Deadline's Instagram account about Norwood also sparked "varying levels of disgust and outrage" in its comments section from Adelaide Kane, Eiza González, Katie Cassidy, Jewel Staite, Lucy Hale, Stephen Sean Ford, and others, singling out González's comment, saying "Shame on whoever is trying to normalize this. Horrific and terrifying." Actor Bronson Pinchot expressed concern that Norwood could take his job. The British union Equity and the Canadian union ACTRA also condemned Norwood. Following this criticism, Van der Velden released a statement claiming Norwood was "not a replacement for a human being, but a creative work." She also denied that a £120,000 grant from the British Film Institute to fund Particle6 had been used to create Norwood, stating that Norwood had been a self-funded project solely for Xicoia. In late October, businessman Kevin O'Leary, while advocating for the use of AI to replace background actors, stated that they could be replaced with "100 Norwell Tillies" without being able to tell the difference. Ryan Reynolds and a real woman named Natalie "Tilly" Norwood also starred in an advertisement for Mint Mobile's internet service provider Minternet that mocked the character of Norwood. In November 2025, Van der Velden stated in an interview with Deadline that she planned to create 40 further "very diverse" characters alongside Norwood in order to expand the character's "whole universe". Also that month, actress Jameela Jamil criticized the idea of Norwood as "deeply disturbing" for being "a teenage-looking girl who can't say no to a type of sex scene" or "advocate for herself". Van der Velden announced later that month that Particle6 would be producing the History Channel's Streets of the Past, a Dutch documentary series which would be hosted by reality television personality Corjan Mol and would use AI to recreate historical scenes. In March 2026, a music video titled "Take The Lead" featuring Norwood was released on YouTube. It addressed the backlash of Norwood's creation by opening with the lyrics: "When they talk about me, they don't see/ The human spark, the creativity," and, "I'm just a tool, but I've got life." It also featured a disclaimer that says: "made by 18 real humans — from production designers to costume designers to prompters, editors and an actor." The vocals were generated by Suno. == Commentary == Charles Pulliam-Moore of The Verge argued that Norwood's introduction was a stunt to normalize "AI actors" despite Norwood essentially being a digital puppet. Straight Arrow News compared Tilly Norwood to Aki Ross, a CGI character from 2001 that was similarly intended to become a "digital star" and appear in multiple films, while Nicholas Schrivens, writing for The Conversation, likened Norwood to the posthumous use of footage of Carrie Fisher as Princess Leia for Star Wars: The Rise of Skywalker in 2019 and the Los Angeles Times likened Norwood to Hatsune Miku. Scrivens also wrote that "no AI creation has achieved the media cut-through that Tilly has". Moises Mendez II of Out dismissed this as "vapid bullshit", writing, "Nobody wants AI actresses." Scottish actress Briony Monroe alleged that Norwood had been modeled after her likeness and mannerisms, and stated that she was consulting Equity regarding the matter. Musician Stella Hennen said in a viral TikTok video, which was uploaded in October 2025 and featured a side-by-side comparison between herself and Norwood, that Norwood was her "doppleganger". On April 14, 2026, Marie Claire published an article titled "Is Tilly Norwood the Most Dangerous 'Actress' in Hollywood?", though it noted that AI-generated characters are "still not very good at, well, acting," "audiences have not been kind to AI-led productions," and "Norwood's 'performances' have already faced negative reviews as well". The University of Southern California's Entertainment Technology Center's AI media director Yves Bergquist dismissed th

  Read more →
• 

  Theaitre

  Theaitre (stylized as THEaiTRE) is an interdisciplinary research project investigating to what extent artificial intelligence is able to generate theatre play scripts. The first theatre play produced within the project, AI: When a Robot Writes a Play, premiered online on February 26, 2021. == Goal == Following similar previous projects such as Sunspring, a short sci-fi movie with an automatically generated script, the THEaiTRE project investigates whether current language generation approaches are mature enough to generate a theatre play script that could be successfully performed in front of an audience. The project falls within the area of generative art, famously represented e.g. by the portrait of Edmond de Belamy which was generated by an artificial neural network. In this field, artists are trying to use automated techniques to create "art", questioning the modern definition of art itself. More broadly, the project aims at promoting cooperation rather than competition of humans and artificial intelligence as the more beneficial approach for both. The first theatre play created within the project, titled AI: When a Robot Writes a Play, was presented in February 2021 at the 100th anniversary of the premiere of the R.U.R. theatre play by the Czech author Karel Čapek to celebrate the invention of the word "robot". While R.U.R. was a play written by a human about robots (and humans), THEaiTRE tried to reverse this idea by presenting a play written by a "robot" (artificial intelligence) about humans (and robots). The script of the play was published online, with marked parts of the text which were written manually or manually post-edited. The analysis shows that 90% of the script is automatically generated, with 10% manually written or manually post-edited. The project also plans to produce a second play in 2022, addressing some of the many shortcomings of the approach used to generate the first play, as well as attempting to further minimize the amount of human influence on the script. == Approach == At the core of the project is the GPT-2 language model by OpenAI with various adjustments motivated by the task of generating theatre play scripts, for which the model is not particularly trained. The GPT-2 model is used in the usual way, providing it with a start of a document and prompting it to generate a continuation of the document. Specifically, the input for GPT-2 in this project is typically a short description of the scene setting, followed by a few lines to introduce the characters and start the dialogue. The model then generates 10 continuation lines, and hands control to the user, who can then either ask the model to continue generating, or make various edits before letting the model to generate further, deleting some parts of the script or adding new lines into the script. The adjustments include restricting the generator to only produce lines pertaining to characters appearing in the input prompt, limiting the repetitiveness of the generated text, and employing automatic summarization of the input prompt and the generated text to overcome the limitation of the GPT-2 model which only attends to the last 1,024 subword tokens. The limitations of the model include, among other, a lack of distinctiveness and self-consistency of the characters, an inability to generate the script for the whole play (scripts for individual scenes are generated independently), and errors due to the employment of automated machine translation, as GPT-2 generates English texts but the final play script is being produced in Czech language. The source codes of the project are available under the MIT licence. The project has also published some sample outputs. == Team == The project is a cooperation of the following experts, all based in Prague, Czech Republic: computational linguists from the Faculty of Mathematics and Physics, Charles University theatre experts from the Švanda Theatre and from the Theatre Faculty of the Academy of Performing Arts in Prague hackers from CEE Hacks The project is financially supported by the Technology Agency of the Czech Republic.

  Read more →
• 

  Database application

  A database application is a computer program whose primary purpose is retrieving information from a computerized database. From here, information can be inserted, modified or deleted which is subsequently conveyed back into the database. Early examples of database applications were accounting systems and airline reservations systems, such as SABRE, developed starting in 1957. A characteristic of modern database applications is that they facilitate simultaneous updates and queries from multiple users. Systems in the 1970s might have accomplished this by having each user in front of a 3270 terminal to a mainframe computer. By the mid-1980s it was becoming more common to give each user a personal computer and have a program running on that PC that is connected to a database server. Information would be pulled from the database, transmitted over a network, and then arranged, graphed, or otherwise formatted by the program running on the PC. Starting in the mid-1990s it became more common to build database applications with a Web interface. Rather than develop custom software to run on a user's PC, the user would use the same Web browser program for every application. A database application with a Web interface had the advantage that it could be used on devices of different sizes, with different hardware, and with different operating systems. Examples of early database applications with Web interfaces include amazon.com, which used the Oracle relational database management system, the photo.net online community, whose implementation on top of Oracle was described in the book Database-Backed Web Sites (Ziff-Davis Press; May 1997), and eBay, also running Oracle. Electronic medical records are referred to on emrexperts.com, in December 2010, as "a software database application". A 2005 O'Reilly book uses the term in its title: Database Applications and the Web. Some of the most complex database applications remain accounting systems, such as SAP, which may contain thousands of tables in only a single module. Many of today's most widely used computer systems are database applications, for example, Facebook, which was built on top of MySQL. The etymology of the phrase "database application" comes from the practice of dividing computer software into systems programs, such as the operating system, compilers, the file system, and tools such as the database management system, and application programs, such as a payroll check processor. On a standard PC running Microsoft Windows, for example, the Windows operating system contains all of the systems programs while games, word processors, spreadsheet programs, photo editing programs, etc. would be application programs. As "application" is short for "application program", "database application" is short for "database application program". Not every program that uses a database would typically be considered a "database application". For example, many physics experiments, e.g., the Large Hadron Collider, generate massive data sets that programs subsequently analyze. The data sets constitute a "database", though they are not typically managed with a standard relational database management system. The computer programs that analyze the data are primarily developed to answer hypotheses, not to put information back into the database and therefore the overall program would not be called a "database application". == Examples of database applications == Amazon Student Data CNN eBay Facebook Fandango Filemaker (Mac OS) LibreOffice Base Microsoft Access Oracle relational database SAP (Systems, Applications & Products in Data Processing) Ticketmaster Wikipedia Yelp YouTube Google MySQL

  Read more →
• 

  Computing Machinery and Intelligence

  "Computing Machinery and Intelligence" is a paper written by Alan Turing on the topic of artificial intelligence. The paper, published in 1950 in Mind, was the first to introduce his concept of what is now known as the Turing test to the general public. Turing's paper considers the question "Can machines think?" Turing says that since the words "think" and "machine" cannot clearly be defined, we should "replace the question by another, which is closely related to it and is expressed in relatively unambiguous words." To achieve this objective, Turing proposes a three-step approach. First, he identifies a simple and unambiguous concept to substitute for the term "think." Second, he delineates the specific "machines" under consideration. Third, armed with these tools, he poses a new question related to the first, which he believes he can answer in the affirmative. == Turing's test == Rather than trying to determine if a machine is thinking, Turing suggests we should ask if the machine can win a game, called the "Imitation Game". The original Imitation game, that Turing described, is a simple party game involving three players. Player A is a man, player B is a woman and player C (who plays the role of the interrogator) can be of either sex. In the Imitation Game, player C is unable to see either player A or player B (and knows them only as X and Y), and can communicate with them only through written notes or any other form that does not give away any details about their gender. By asking questions of player A and player B, player C tries to determine which of the two is the man and which is the woman. Player A's role is to trick the interrogator into making the wrong decision, while player B attempts to assist the interrogator in making the right one. Turing proposes a variation of this game that involves the computer: We now ask the question, "What will happen when a machine takes the part of A in this game?" Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? These questions replace our original, "Can machines think?" So the modified game becomes one that involves three participants in isolated rooms: a computer (which is being tested), a human, and a (human) judge. The human judge can converse with both the human and the computer by typing into a terminal. Both the computer and the human try to convince the judge that they are the human. If the judge cannot consistently tell which is which, then the computer wins the game. Researchers in the United Kingdom had been exploring "machine intelligence" for up to ten years prior to the founding of the field of artificial intelligence (AI) research in 1956. It was a common topic among the members of the Ratio Club, an informal group of British cybernetics and electronics researchers that included Alan Turing. Turing, in particular, had been running the notion of machine intelligence since at least 1941 and one of the earliest-known mentions of "computer intelligence" was made by him in 1947. As Stevan Harnad notes, the question has become "Can machines do what we (as thinking entities) can do?" In other words, Turing is no longer asking whether a machine can "think"; he is asking whether a machine can act indistinguishably from the way a thinker acts. This question avoids the difficult philosophical problem of pre-defining the verb "to think" and focuses instead on the performance capacities that being able to think makes possible, and how a causal system can generate them. Since Turing introduced his test, it has been both highly influential and widely criticised, and has become an important concept in the philosophy of artificial intelligence. Some of its criticisms, such as John Searle's Chinese room, are themselves controversial. Some have taken Turing's question to have been "Can a computer, communicating over a teleprinter, fool a person into believing it is human?" but it seems clear that Turing was not talking about fooling people but about generating human cognitive capacity. == Digital machines == Turing also notes that we need to determine which "machines" we wish to consider. He points out that a human clone, while man-made, would not provide a very interesting example. Turing suggested that we should focus on the capabilities of digital machinery—machines which manipulate the binary digits of 1 and 0, rewriting them into memory using simple rules. He gave two reasons. First, there is no reason to speculate whether or not they can exist. They already did in 1950. Second, digital machinery is "universal". Turing's research into the foundations of computation had proved that a digital computer can, in theory, simulate the behaviour of any other digital machine, given enough memory and time. (This is the essential insight of the Church–Turing thesis and the universal Turing machine.) Therefore, if any digital machine can "act like it is thinking", then every sufficiently powerful digital machine can. Turing writes, "all digital computers are in a sense equivalent." This allows the original question to be made even more specific. Turing now restates the original question as "Let us fix our attention on one particular digital computer C. Is it true that by modifying this computer to have an adequate storage, suitably increasing its speed of action, and providing it with an appropriate programme, C can be made to play satisfactorily the part of A in the imitation game, the part of B being taken by a man?" Hence, Turing states that the focus is not on "whether all digital computers would do well in the game nor whether the computers that are presently available would do well, but whether there are imaginable computers which would do well". What is more important is to consider the advancements possible in the state of our machines today regardless of whether we have the available resource to create one or not. == Nine common objections == Having clarified the question, Turing turned to answering it: he considered the following nine common objections, which include all the major arguments against artificial intelligence raised in the years since his paper was first published. Religious Objection: This states that thinking is a function of man's immortal soul; therefore, a machine cannot think. "In attempting to construct such machines," wrote Turing, "we should not be irreverently usurping His power of creating souls, any more than we are in the procreation of children: rather we are, in either case, instruments of His will providing mansions for the souls that He creates." 'Heads in the Sand' Objection: "The consequences of machines thinking would be too dreadful. Let us hope and believe that they cannot do so." This thinking is popular among intellectual people, as they believe superiority derives from higher intelligence and the possibility of being overtaken is a threat (as machines have efficient memory capacities and processing speed, machines exceeding the learning and knowledge capabilities are highly probable). This objection is a fallacious appeal to consequences, confusing what should not be with what can or cannot be (Wardrip-Fruin, 56). The Mathematical Objection: This objection uses mathematical theorems, such as Gödel's incompleteness theorem, to show that there are limits to what questions a computer system based on logic can answer. Turing suggests that humans are too often wrong themselves and pleased at the fallibility of a machine. (This argument would be made again by philosopher John Lucas in 1961 and physicist Roger Penrose in 1989, and later would be called Penrose–Lucas argument.) Argument From Consciousness: This argument, suggested by Professor Geoffrey Jefferson in his 1949 Lister Oration (acceptance speech for his 1948 award of Lister Medal) states that "not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols, could we agree that machine equals brain." Turing replies by saying that we have no way of knowing that any individual other than ourselves experiences emotions, and that therefore we should accept the test. He adds, "I do not wish to give the impression that I think there is no mystery about consciousness ... [b]ut I do not think these mysteries necessarily need to be solved before we can answer the question [of whether machines can think]." (This argument, that a computer can't have conscious experiences or understanding, would be made in 1980 by philosopher John Searle in his Chinese room argument. Turing's reply is now known as the "other minds reply". See also Can a machine have a mind? in the philosophy of AI.) Arguments from various disabilities. These arguments all have the form "a computer will never do X". Turing offers a selection:Be kind, resourceful, beautiful, friendly, have initiative, have a sense of humour, tell right from wrong, make mistakes, fall in love, enjo

  Read more →
• 

  Fuzzy cognitive map

  A fuzzy cognitive map (FCM) is a cognitive map within which the relations between the elements (e.g. concepts, events, project resources) of a "mental landscape" can be used to compute the "strength of impact" of these elements. Fuzzy cognitive maps were introduced by Bart Kosko. Robert Axelrod introduced cognitive maps as a formal way of representing social scientific knowledge and modeling decision making in social and political systems, then brought in the computation. == Details == Fuzzy cognitive maps are signed fuzzy directed graphs. Spreadsheets or tables are used to map FCMs into matrices for further computation. FCM is a technique used for causal knowledge acquisition and representation, it supports causal knowledge reasoning process and belong to the neuro-fuzzy system that aim at solving decision making problems, modeling and simulate complex systems. Learning algorithms have been proposed for training and updating FCMs weights mostly based on ideas coming from the field of Artificial Neural Networks. Adaptation and learning methodologies used to adapt the FCM model and adjust its weights. Kosko and Dickerson (Dickerson & Kosko, 1994) suggested the Differential Hebbian Learning (DHL) to train FCM. There have been proposed algorithms based on the initial Hebbian algorithm; others algorithms come from the field of genetic algorithms, swarm intelligence and evolutionary computation. Learning algorithms are used to overcome the shortcomings that the traditional FCM present i.e. decreasing the human intervention by suggested automated FCM candidates; or by activating only the most relevant concepts every execution time; or by making models more transparent and dynamic. Fuzzy cognitive maps (FCMs) have gained considerable research interest due to their ability in representing structured knowledge and model complex systems in various fields. This growing interest led to the need for enhancement and making more reliable models that can better represent real situations. A first simple application of FCMs is described in a book of William R. Taylor, where the war in Afghanistan and Iraq is analyzed. In Bart Kosko's book Fuzzy Thinking, several Hasse diagrams illustrate the use of FCMs. As an example, one FCM quoted from Rod Taber describes 11 factors of the American cocaine market and the relations between these factors. For computations, Taylor uses pentavalent logic (scalar values out of {-1,-0.5,0,+0.5,+1}). That particular map of Taber uses trivalent logic (scalar values out of {-1,0,+1}). Taber et al. also illustrate the dynamics of map fusion and give a theorem on the convergence of combination in a related article. While applications in social sciences introduced FCMs to the public, they are used in a much wider range of applications, which all have to deal with creating and using models of uncertainty and complex processes and systems. Examples: In business FCMs can be used for product planning and decision support. In economics, FCMs support the use of game theory in more complex settings. In education for modeling Critical Success Factors of Learning Management Systems. In medical applications to model systems, provide diagnosis, develop decision support systems and medical assessment. In engineering for modeling and control mainly of complex systems and reliability engineering In project planning FCMs help to analyze the mutual dependencies between project resources. In robotics FCMs support machines to develop fuzzy models of their environments and to use these models to make crisp decisions. In computer assisted learning FCMs enable computers to check whether students understand their lessons. In expert systems a few or many FCMs can be aggregated into one FCM in order to process estimates of knowledgeable persons. In IT project management, a FCM-based methodology helps to success modelling, risk analysis and assessment, IT scenarios FCMappers is an international online community for the analysis and the visualization of fuzzy cognitive maps. FCMappers offer support for starting with FCM and also provide a Microsoft Excel-based tool that is able to check and analyse FCMs. The output is saved as Pajek file and can be visualized within third party software like Pajek, Visone, etc. They also offer to adapt the software to specific research needs. Additional FCM software tools, such as Mental Modeler, have recently been developed as a decision-support tool for use in social science research, collaborative decision-making, and natural resource planning.

  Read more →
• 

  Fuzzy architectural spatial analysis

  Fuzzy architectural spatial analysis (FASA) (also fuzzy inference system (FIS) based architectural space analysis or fuzzy spatial analysis) is a spatial analysis method of analysing the spatial formation and architectural space intensity within any architectural organization. Fuzzy architectural spatial analysis is used in architecture, interior design, urban planning and similar spatial design fields. == Overview == Fuzzy architectural spatial analysis was developed by Burcin Cem Arabacioglu (2010) from the architectural theories of space syntax and visibility graph analysis, and is applied with the help of a fuzzy system with a Mamdani inference system based on fuzzy logic within any architectural space. Fuzzy architectural spatial analysis model analyses the space by considering the perceivable architectural element by their boundary and stress characteristics and intensity properties. The method is capable of taking all sensorial factors into account during analyses in conformably with the perception process of architectural space which is a multi-sensorial act.

  Read more →
• 

  PARRY

  PARRY was an early example of a chatbot, implemented in 1972 by psychiatrist Kenneth Colby. == History == PARRY was written in 1972 by psychiatrist Kenneth Colby, then at Stanford University. While ELIZA was a simulation of a Rogerian therapist, PARRY attempted to simulate a person with paranoid schizophrenia. The program implemented a crude model of the behavior of a person with paranoid schizophrenia based on concepts, conceptualizations, and beliefs (judgements about conceptualizations: accept, reject, neutral). It also embodied a conversational strategy, and as such was a much more serious and advanced program than ELIZA. It was described as "ELIZA with attitude". PARRY was tested in the early 1970s using a variation of the Turing Test. A group of experienced psychiatrists analysed a combination of real patients and computers running PARRY through teleprinters. Another group of 33 psychiatrists were shown transcripts of the conversations. The two groups were then asked to identify which of the "patients" were human and which were computer programs. The psychiatrists were able to make the correct identification only 48 percent of the time — a figure consistent with random guessing. PARRY and ELIZA (also known as "the Doctor") interacted several times. The most famous of these exchanges occurred at the ICCC 1972, where PARRY and ELIZA were hooked up over ARPANET and responded to each other.

  Read more →
• 

  Computational law

  Computational law is the branch of legal informatics concerned with the automation of legal reasoning. What distinguishes Computational Law systems from other instances of legal technology is their autonomy, i.e. the ability to answer legal questions without additional input from human legal experts. While there are many possible applications of Computational Law, the primary focus of work in the field today is compliance management, i.e. the development and deployment of computer systems capable of assessing, facilitating, or enforcing compliance with rules and regulations. Some systems of this sort already exist. TurboTax is a good example. And the potential is particularly significant now due to recent technological advances – including the prevalence of the Internet in human interaction and the proliferation of embedded computer systems (such as smart phones, self-driving cars, and robots). There are also applications that do not involve governmental laws. The regulations can just as well be the terms of contracts (e.g. delivery schedules, insurance covenants, real estate transactions, financial agreements). They can be the policies of corporations (e.g. constraints on travel, expenditure reporting, pricing rules). They can even be the rules of games (embodied in computer game playing systems). == History == Speculation about potential benefits to legal practice through applying methods from computational science and AI research to automate parts of the law date back at least to the middle 1940s. Further, AI and law and computational law do not seem easily separable, as perhaps most of AI research focusing on the law and its automation appears to utilize computational methods. The forms that speculation took are multiple and not all related in ways to readily show closeness to one another. This history will sketch them as they were, attempting to show relationships where they can be found to have existed. By 1949, a minor academic field aiming to incorporate electronic and computational methods to legal problems had been founded by American legal scholars, called jurimetrics. Though broadly said to be concerned with the application of the "methods of science" to the law, these methods were actually of a quite specifically defined scope. Jurimetrics was to be "concerned with such matters as the quantitative analysis of judicial behavior, the application of communication and information theory to legal expression, the use of mathematical logic in law, the retrieval of legal data by electronic and mechanical means, and the formulation of a calculus of legal predictability". These interests led in 1959 to the founding a journal, Modern Uses of Logic in Law, as a forum wherein articles would be published about the applications of techniques such as mathematical logic, engineering, statistics, etc. to the legal study and development. In 1966, this Journal was renamed as Jurimetrics. Today, however, the journal and meaning of jurimetrics seems to have broadened far beyond what would fit under the areas of applications of computers and computational methods to law. Today the journal not only publishes articles on such practices as found in computational law, but has broadened jurimetrical concerns to mean also things like the use of social science in law or the "policy implications [of] and legislative and administrative control of science". Independently in 1958, at the Conference for the Mechanization of Thought held at the National Physical Laboratory in Teddington, Middlesex, UK, the French jurist Lucien Mehl presented a paper both on the benefits of using computational methods for law and on the potential means to use such methods to automate law for a discussion that included AI luminaries like Marvin Minsky. Mehl believed that the law could by automated by two basic distinct, though not wholly separable, types of machine. These were the "documentary or information machine", which would provide the legal researcher quick access to relevant case precedents and legal scholarship, and the "consultation machine", which would be "capable of answering any question put to it over a vast field of law". The latter type of machine would be able to basically do much of a lawyer's job by simply giving the "exact answer to a [legal] problem put to it". By 1970, Mehl's first type of machine, one that would be able to retrieve information, had been accomplished but there seems to have been little consideration of further fruitful intersections between AI and legal research. There were, however, still hopes that computers could model the lawyer's thought processes through computational methods and then apply that capacity to solve legal problems, thus automating and improving legal services via increased efficiency as well as shedding light on the nature of legal reasoning. By the late 1970s, computer science and the affordability of computer technology had progressed enough that the retrieval of "legal data by electronic and mechanical means" had been achieved by machines fitting Mehl's first type and were in common use in American law firms. During this time, research focused on improving the goals of the early 1970s occurred, with programs like Taxman being worked on in order to both bring useful computer technology into the law as practical aids and to help specify the exact nature of legal concepts. Nonetheless, progress on the second type of machine, one that would more fully automate the law, remained relatively inert. Research into machines that could answer questions in the way that Mehl's consultation machine would picked up somewhat in the late 1970s and 1980s. A 1979 convention in Swansea, Wales marked the first international effort solely to focus upon applying artificial intelligence research to legal problems in order to "consider how computers can be used to discover and apply the legal norms embedded within the written sources of the law". Considerable progress on the development of the second type of machine was made in the following decade, with the development of a variety of expert systems. According to Thorne McCarty, "these systems all have the following characteristics: They do backward chaining inference from a specified goal; they ask questions to elicit information from the user; and they produce a suggested answer along with a trace of the supporting legal rules." According to Prakken and Sartor the representation of the British Nationality Act as a logic program, which introduced this approach, was "hugely influential for the development of computational representations of legislation, showing how logic programming enables intuitively appealing representations that can be directly deployed to generate automatic inferences". In 2021, this work received the Inaugural CodeX Prize as "one of the first and best-known works in computational law, and one of the most widely cited papers in the field." In a 1988 review of Anne Gardner's book An Artificial Intelligence Approach to Legal Reasoning (1987), the Harvard academic legal scholar and computer scientist Edwina Rissland wrote that "She plays, in part, the role of pioneer; artificial intelligence ("AI") techniques have not yet been widely applied to perform legal tasks. Therefore, Gardner, and this review, first describe and define the field, then demonstrate a working model in the domain of contract offer and acceptance." Eight years after the Swansea conference had passed, and still AI and law researchers merely trying to delineate the field could be described by their own kind as "pioneer[s]". In the 1990s and early 2000s more progress occurred. Computational research generated insights for law. The First International Conference on AI and the Law occurred in 1987, but it is in the 1990s and 2000s that the biannual conference began to build up steam and to delve more deeply into the issues involved with work intersecting computational methods, AI, and law. Classes began to be taught to undergraduates on the uses of computational methods to automating, understanding, and obeying the law. Further, by 2005, a team largely composed of Stanford computer scientists from the Stanford Logic group had devoted themselves to studying the uses of computational techniques to the law. Computational methods in fact advanced enough that members of the legal profession began in the 2000s to both analyze, predict and worry about the potential future of computational law and a new academic field of computational legal studies seems to be now well established. As insight into what such scholars see in the law's future due in part to computational law, here is quote from a recent conference about the "New Normal" for the legal profession: "Over the last 5 years, in the fallout of the Great Recession, the legal profession has entered the era of the New Normal. Notably, a series of forces related to technological change, globalization, and the pressure to do more with less (in both corpo

  Read more →
• 

  Generative AI pornography

  Generative AI pornography or simply AI pornography is a digitally created pornography produced through generative artificial intelligence (AI) technologies. Unlike traditional pornography, which involves real actors and cameras, this content is synthesized entirely by AI algorithms. These algorithms, including generative adversarial networks (GANs) and text-to-image models, generate lifelike images, videos, or animations from textual descriptions or datasets. == Functions and production strategies == AI pornography platforms, beyond account creation and social media linking, primarily enable users to generate sexual images through feature selection or text prompting. Users can customize bodies, clothing, and sociodemographic traits, and browse categorized galleries of user‑generated content. Several sites also support short pornographic videos or GIFs and modification tools such as nudifiers, deepfakes, and facemorphing. Platforms often allow fine‑tuning of parameters such as settings, style, or theme, and provide prompt enhancers or suggestions to improve outputs. Users may edit generated images, refine prior prompts, modify others’ work, or upload personal material as a basis, with iterative and collaborative content creation. Some websites additionally host interactive “erobots,” customizable in real time for appearance, personality, memories, speech, and profession, enabling tailored sexual and non‑sexual interactions. Less common features include VR integration, AI porn games, audio or doodle prompts, and consensual replication of individuals with verification. == History == The use of generative AI in the adult industry began in the late 2010s, initially focusing on AI-generated art, music, and visual content. This trend accelerated in 2022 with Stability AI's release of Stable Diffusion (SD), an open-source text-to-image model that enables users to generate images, including NSFW content, from text prompts using the LAION-Aesthetics subset of the LAION-5B dataset. Despite Stability AI's warnings against sexual imagery, SD's public release led to dedicated communities exploring both artistic and explicit content, sparking ethical debates over open-access AI and its use in adult media. By 2020, AI tools had advanced to generate highly realistic adult content, amplifying calls for regulation. === AI-generated influencers === One application of generative AI technology is the creation of AI-generated influencers on platforms such as OnlyFans and Instagram. These AI personas interact with users in ways that can mimic real human engagement, offering an entirely synthetic but convincing experience. While popular among niche audiences, these virtual influencers have prompted discussions about authenticity, consent, and the blurring line between human and AI-generated content, especially in adult entertainment. === The growth of AI porn sites === By 2023, websites dedicated to AI-generated adult content had gained traction, catering to audiences seeking customizable experiences. These platforms allow users to create or view AI-generated pornography tailored to their preferences. These platforms enable users to create or view AI-generated adult content appealing to different preferences through prompts and tags, customizing body type, facial features, and art styles. Tags further refine the output, creating niche and diverse content. Many sites feature extensive image libraries and continuous content feeds, combining personalization with discovery and enhancing user engagement. AI porn sites, therefore, attract those seeking unique or niche experiences, sparking debates on creativity and the ethical boundaries of AI in adult media. == Ethical concerns and misuse == The growth of generative AI pornography has also attracted some cause for criticism. AI technology can be exploited to create non-consensual pornographic material, posing risks similar to those seen with deepfake revenge porn and AI-generated NCII (Non-Consensual Intimate Image). A 2023 analysis found that 98% of deepfake videos online are pornographic, with 99% of the victims being women. Some famous celebrities victims of deepfake include Scarlett Johansson, Taylor Swift, and Maisie Williams. OpenAI is exploring whether NSFW content, such as erotica, can be responsibly generated in age-appropriate contexts while maintaining its ban on deepfakes. This proposal has attracted criticism from child safety campaigners who argue it undermines OpenAI's mission to develop "safe and beneficial" AI. Additionally, the Internet Watch Foundation has raised concerns about AI being used to generate sexual abuse content involving children. === AI-generated non-consensual intimate imagery (AI Undress) === Generative AI have extensively been used to produce pornography images and videos of non-consenting individuals. 404 Media reported a particular AI generated porn bot on Telegram has more than 100,000 monthly users. Alibaba, the Chinese tech company, released an AI video generation model in 2025 called Wan 2.1, which was modified to produce non-consensual pornography. Several US states are taking actions against using deepfake apps and sharing them on the internet. In 2024, San Francisco filed a landmark lawsuit to shut down "undress" apps that allow users to generate non-consensual AI nude images, citing violations of state laws. The case aligns with California's recent legislation—SB 926, SB 942, and SB 981—championed by Senators Aisha Wahab and Josh Becker and signed by Governor Gavin Newsom. These bills aim to protect individuals from AI-generated explicit images by criminalizing non-consensual distribution, mandating disclosures, and empowering victims to report and remove harmful content from platforms. === Differences from deepfake pornography === While both generative AI pornography and deepfake pornography rely on synthetic media, they differ in their methods and ethical considerations. Deepfake pornography typically involves altering existing footage of real individuals, often without their consent, using AI to superimpose faces, undress said persons, or modify scenes. In contrast, generative AI pornography is created using algorithms, producing hyper-realistic content without the need to upload real pictures of people. Hany Farid, digital image analysis expert, also described the difference between "AI porn" and "deepfake porn." == Legality == The legality of generative AI pornography varies widely by jurisdiction and remains an evolving issue. In some countries, laws addressing digital impersonation, obscenity, or deepfake technologies may indirectly apply, particularly when AI-generated content involves the likeness of real individuals without consent. The absence of a physical performer further complicates traditional regulatory frameworks, which are often grounded in performer protection and distribution laws. In the United States, legal responses have primarily focused on non-consensual deepfakes and impersonation. Some states, such as Virginia, California, and Texas, have enacted legislation criminalising the creation or distribution of non-consensual explicit deepfake content. However, there is no comprehensive federal law addressing AI-generated pornography, leaving a patchwork of legal interpretations and enforcement standards across different jurisdictions. According to a 2023 report, South Korea accounts for approximately 53% of global deepfake pornography production. In September 2024, South Korea's National Assembly amended the Act on Special Cases Concerning the Punishment of Sexual Crimes, introducing two significant reforms related to deepfake content. The first criminalises the possession, viewing, purchase, and storage of non-consensual deepfake material, with penalties of up to three years in prison or fines of up to 30 million won (approximately USD 20,000). The second reform specifically addresses the exploitation of minors, establishing that individuals who use deepfakes to threaten or blackmail minors face a minimum of three years' imprisonment, and at least five years if they coerce minors into unwanted acts. In England and Wales the Data (Use and Access) Act 2025 has legislated against the creation, or the request for creation, of intimate images by nudifying software or websites of another person who has not consented to this. However as of January 2026 this has not yet been brought into force.

  Read more →