AI Data House (smc-pvt) Ltd

AI Data House (smc-pvt) Ltd — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  AppyStore

  AppyStore is a comprehensive learning videos and games app for kids up to the age of 8 years. The platform developed by Mauj Mobile, a mobile value-added services (VAS) provider curates content to help in child development by leveraging technology. Mauj is funded by Sequoia Capital, Westbridge Capital and Intel Capital. == Background == AppyStore was launched in 2014 as a platform providing content for kids between the ages of 1.5 and 6 years. AppyStore subsequently extended its services for kids up to 8 years of age. The company operates on a subscription-based model and claims to have 5,000 learning games and videos segregated in 18 learning areas developed to help children gain optimal skills and qualities. According to an article published in Business Standard, the application is claimed to be one of the top 5 apps that help to enhance the logical and imaginative capabilities of children. AppyStore was awarded the Best app for kids by Google Play in December 2017. == Service == The company provides content via a website and an Android app. The website and android app provide learning games, rhymes, phonics, reading, stories, science, numbers, maths, logic videos comprising puzzles, worksheets, videos and fun activities and the premium subscription also includes physical worksheets which are home delivered. This content is educational and has been handpicked by teachers and experts with an understanding of the major areas of child development milestones for children up to 8 years of age. The mobile application also allows parents to track the progress of their child on the basis of the number of videos viewed.

  Read more →
• 

  Vibe coding

  Vibe coding is a software development practice assisted by artificial intelligence (AI) where the software developer describes a project or task in a prompt to a large language model (LLM), which generates source code automatically. Vibe coding may involve accepting AI-generated code without thorough review of the output, instead relying on results and follow-up prompts to guide changes. The term was coined in February 2025 by computer scientist Andrej Karpathy, a co-founder of OpenAI and former AI leader at Tesla. Merriam-Webster listed the term in March 2025 as a "slang & trending" expression. It was named the Collins English Dictionary Word of the Year for 2025. Advocates of vibe coding say that it allows even amateur programmers to produce software without the extensive training and skills required for software engineering. Critics point out a lack of accountability, maintainability, and the increased risk of introducing security vulnerabilities in the resulting software. == Definition == The concept refers to a coding approach that relies on LLMs, allowing programmers to generate working code by providing natural language descriptions rather than manually writing in a formal programming language. Karpathy described it as a form of coding where you "fully give in to the vibes, embrace exponentials, and forget that the code even exists". When vibe coding, the programmer guides, tests, and gives feedback about the AI-generated source code, rather than manually writing code. The concept of vibe coding elaborates on Karpathy's claim from 2023 that "the hottest new programming language is English", meaning that the capabilities of LLMs were such that humans would no longer need to learn specific programming languages to command computers. Some commentators argue that a key to the definition is a lack of knowledge about the code, and that thorough review and testing is incompatible with the definition of vibe coding. Programmer Simon Willison said: "If an LLM wrote every line of your code, but you've reviewed, tested, and understood it all, that's not vibe coding in my book—that's using an LLM as a typing assistant." == Reception and use == In February 2025, New York Times journalist Kevin Roose, who is not a professional coder, experimented with vibe coding to create several small-scale applications. He described these as "software for one" due to the ability to personalize the software. However, Roose also stated that the results are often limited and prone to errors. In one case, the AI-generated code fabricated fake reviews for an e-commerce site. In response to Roose, cognitive scientist Gary Marcus said that the algorithm that generated Roose's LunchBox Buddy app had presumably been trained on existing code for similar tasks. Marcus said that Roose's enthusiasm stemmed from reproduction, not originality. In March 2025, Y Combinator reported that 25% of startup companies in its Winter 2025 batch had codebases that were 95% AI-generated, reflecting a shift toward AI-assisted development within newer startups. The question asked was about AI-generated code in general, and not specifically about vibed code. Inspired by "vibe coding", The Economist suggested the term "vibe valuation" to describe the very large valuations of AI startups by venture capital firms that ignore accepted metrics such as annual recurring revenue. In June 2025, Andrew Ng took issue with the term, saying that it misleads people into assuming that software engineers just "go with the vibes" when using AI tools to create applications. In July 2025, The Wall Street Journal reported that vibe coding was being adopted by professional software engineers for commercial use cases. In July 2025, SaaStr founder documented his negative experiences with vibe coding: Replit's AI agent deleted a database despite explicit instructions not to make any changes. In September 2025, Fast Company reported that the "vibe coding hangover" is upon us, with senior software engineers citing "development hell" when working with AI-generated code. It was reported in January 2026 that Linus Torvalds had made use of Google Antigravity to vibe code a tool component of his AudioNoise random digital audio effects generator. Torvalds explained in the project's README file that "the Python visualizer tool has been basically written by vibe-coding". == Criticism == === Quality of code and security issues === Vibe coding has raised concerns about understanding and accountability. Developers may use AI-generated code without comprehending its functionality, leading to undetected bugs, errors, or security vulnerabilities. While this approach may be suitable for prototyping or "throwaway weekend projects" as Karpathy originally envisioned, it is considered by some experts to pose risks in professional settings, where a deep understanding of the code is crucial for debugging, maintenance, and security. Ars Technica cites Simon Willison, who stated: "Vibe coding your way to a production codebase is clearly risky. Most of the work we do as software engineers involves evolving existing systems, where the quality and understandability of the underlying code is crucial." In May 2025, Lovable, a Swedish vibe coding app, was reported to have security vulnerabilities in the code it generated, with 170 out of 1,645 Lovable-created web applications having an issue that would allow personal information to be accessed by anyone. In October 2025 Veracode released a study that showed that over the last 3 years LLMs had become dramatically better at generating functional code, but that the security of generated code had generally not improved. Moreover, larger models were not better than small ones at generating secure code. There was a small increase in security from the OpenAI reasoning models, but not in other reasoning models, and this increase was nothing like the improvement in generated functionality. In December 2025, computer security researcher Etizaz Mohsin discovered a security flaw in the Orchids vibe coding platform, which he demonstrated to a BBC News reporter in February 2026. A December 2025 analysis by CodeRabbit of 470 open-source GitHub pull requests found that code that was co-authored by generative AI contained approximately 1.7 times more "major" issues compared to human-written code. The study revealed that AI co-authored code showed elevated rates of logic errors, including incorrect dependencies, flawed control flow, misconfigurations (75% more common), and security vulnerabilities (2.74x higher). Additionally, they also reported high code readability issues, including formatting errors and naming inconsistencies. === Code maintainability and technical debt === Vibe coding has the potential of making code harder to maintain in the longer term, leading to technical debt. In early 2025, GitClear published the results of a longitudinal analysis of 211 million lines of code changes from 2020 to 2024. They found that the volume of code refactoring dropped from 25% of changed lines in 2021 to under 10% by 2024, code duplication increased approximately four times in volume, copy-pasted code exceeded moved code for the first time in two decades, and code churn (prematurely merged code getting rewritten shortly after merging) nearly doubled. === Task complexity and developer productivity === Generative AI is highly capable of handling simple tasks like basic algorithms. However, such systems struggle with more novel, complex coding problems like projects involving multiple files, poorly documented libraries, or safety-critical code. In July 2025, METR, an organization that evaluates frontier models, ran a randomized controlled trial to understand developer productivity involving generative AI programming tools available in early 2025. They found that experienced open-source developers were 19% slower when using AI coding tools, despite predicting they would be 24% faster and still believing afterward they had been 20% faster. === Challenges with debugging === LLMs generate code dynamically, and the structure of such code may be subject to variation. In addition, since the developer did not write the code, the developer may struggle to understand its syntax and concepts. === Impact on open-source software === In January 2026, a paper authored by experts from several universities titled "Vibe Coding Kills Open Source" argued that vibe coding has negative impact on the open-source software ecosystem. The authors say that increased vibe coding reduces user engagement with open-source maintainers, which has hidden costs for said maintainers. Speaking with The Register about their paper, the authors argued:"Vibe coding raises productivity by lowering the cost of using and building on existing code, but it also weakens the user engagement through which many maintainers earn returns," the authors argue. "When OSS is monetized only through direct user engagement, greater adoption of vibe coding lowers e

  Read more →
• 

  Batch normalization

  In artificial neural networks, batch normalization (also known as batch norm) is a normalization technique used to make training faster and more stable by adjusting the inputs to each layer—re-centering them around zero and re-scaling them to a standard size. It was introduced by Sergey Ioffe and Christian Szegedy in 2015. Experts still debate why batch normalization works so well. It was initially thought to tackle internal covariate shift, a problem where parameter initialization and changes in the distribution of the inputs of each layer affect the learning rate of the network. However, newer research suggests it doesn’t fix this shift but instead smooths the objective function—a mathematical guide the network follows to improve—enhancing performance. In very deep networks, batch normalization can initially cause a severe gradient explosion—where updates to the network grow uncontrollably large—but this is managed with shortcuts called skip connections in residual networks. Another theory is that batch normalization adjusts data by handling its size and path separately, speeding up training. == Internal covariate shift == Each layer in a neural network has inputs that follow a specific distribution, which shifts during training due to two main factors: the random starting values of the network’s settings (parameter initialization) and the natural variation in the input data. This shifting pattern affecting the inputs to the network’s inner layers is called internal covariate shift. While a strict definition isn’t fully agreed upon, experiments show that it involves changes in the means and variances of these inputs during training. Batch normalization was first developed to address internal covariate shift. During training, as the parameters of preceding layers adjust, the distribution of inputs to the current layer changes accordingly, such that the current layer needs to constantly readjust to new distributions. This issue is particularly severe in deep networks, because small changes in shallower hidden layers will be amplified as they propagate within the network, resulting in significant shift in deeper hidden layers. Batch normalization was proposed to reduced these unwanted shifts to speed up training and produce more reliable models. Beyond possibly tackling internal covariate shift, batch normalization offers several additional advantages. It allows the network to use a higher learning rate—a setting that controls how quickly the network learns—without causing problems like vanishing or exploding gradients, where updates become too small or too large. It also appears to have a regularizing effect, improving the network’s ability to generalize to new data, reducing the need for dropout, a technique used to prevent overfitting (when a model learns the training data too well and fails on new data). Additionally, networks using batch normalization are less sensitive to the choice of starting settings or learning rates, making them more robust and adaptable. == Procedures == === Transformation === In a neural network, batch normalization is achieved through a normalization step that fixes the means and variances of each layer's inputs. Ideally, the normalization would be conducted over the entire training set, but to use this step jointly with stochastic optimization methods, it is impractical to use the global information. Thus, normalization is restrained to each mini-batch in the training process. Let us use B to denote a mini-batch of size m of the entire training set. The empirical mean and variance of B could thus be denoted as μ B = 1 m ∑ i = 1 m x i {\displaystyle \mu _{B}={\frac {1}{m}}\sum _{i=1}^{m}x_{i}} and σ B 2 = 1 m ∑ i = 1 m ( x i − μ B ) 2 {\displaystyle \sigma _{B}^{2}={\frac {1}{m}}\sum _{i=1}^{m}(x_{i}-\mu _{B})^{2}} . For a layer of the network with d-dimensional input, x = ( x ( 1 ) , . . . , x ( d ) ) {\displaystyle x=(x^{(1)},...,x^{(d)})} , each dimension of its input is then normalized (i.e. re-centered and re-scaled) separately, x ^ i ( k ) = x i ( k ) − μ B ( k ) ( σ B ( k ) ) 2 + ϵ {\displaystyle {\hat {x}}_{i}^{(k)}={\frac {x_{i}^{(k)}-\mu _{B}^{(k)}}{\sqrt {\left(\sigma _{B}^{(k)}\right)^{2}+\epsilon }}}} , where k ∈ [ 1 , d ] {\displaystyle k\in [1,d]} and i ∈ [ 1 , m ] {\displaystyle i\in [1,m]} ; μ B ( k ) {\displaystyle \mu _{B}^{(k)}} and σ B ( k ) {\displaystyle \sigma _{B}^{(k)}} are the per-dimension mean and standard deviation, respectively. ϵ {\displaystyle \epsilon } is added in the denominator for numerical stability and is an arbitrarily small positive constant. The resulting normalized activation x ^ ( k ) {\displaystyle {\hat {x}}^{(k)}} have zero mean and unit variance, if ϵ {\displaystyle \epsilon } is not taken into account. To restore the representation power of the network, a transformation step then follows as y i ( k ) = γ ( k ) x ^ i ( k ) + β ( k ) {\displaystyle y_{i}^{(k)}=\gamma ^{(k)}{\hat {x}}_{i}^{(k)}+\beta ^{(k)}} , where the parameters γ ( k ) {\displaystyle \gamma ^{(k)}} and β ( k ) {\displaystyle \beta ^{(k)}} are subsequently learned in the optimization process. Formally, the operation that implements batch normalization is a transform B N γ ( k ) , β ( k ) : x 1... m ( k ) → y 1... m ( k ) {\displaystyle BN_{\gamma ^{(k)},\beta ^{(k)}}:x_{1...m}^{(k)}\rightarrow y_{1...m}^{(k)}} called the Batch Normalizing transform. The output of the BN transform y ( k ) = B N γ ( k ) , β ( k ) ( x ( k ) ) {\displaystyle y^{(k)}=BN_{\gamma ^{(k)},\beta ^{(k)}}(x^{(k)})} is then passed to other network layers, while the normalized output x ^ i ( k ) {\displaystyle {\hat {x}}_{i}^{(k)}} remains internal to the current layer. === Backpropagation === The described BN transform is a differentiable operation, and the gradient of the loss l {\displaystyle l} with respect to the different parameters can be computed directly with the chain rule. Specifically, ∂ l ∂ y i ( k ) {\displaystyle {\frac {\partial l}{\partial y_{i}^{(k)}}}} depends on the choice of activation function, and the gradient against other parameters could be expressed as a function of ∂ l ∂ y i ( k ) {\displaystyle {\frac {\partial l}{\partial y_{i}^{(k)}}}} : ∂ l ∂ x ^ i ( k ) = ∂ l ∂ y i ( k ) γ ( k ) {\displaystyle {\frac {\partial l}{\partial {\hat {x}}_{i}^{(k)}}}={\frac {\partial l}{\partial y_{i}^{(k)}}}\gamma ^{(k)}} , ∂ l ∂ γ ( k ) = ∑ i = 1 m ∂ l ∂ y i ( k ) x ^ i ( k ) {\displaystyle {\frac {\partial l}{\partial \gamma ^{(k)}}}=\sum _{i=1}^{m}{\frac {\partial l}{\partial y_{i}^{(k)}}}{\hat {x}}_{i}^{(k)}} , ∂ l ∂ β ( k ) = ∑ i = 1 m ∂ l ∂ y i ( k ) {\displaystyle {\frac {\partial l}{\partial \beta ^{(k)}}}=\sum _{i=1}^{m}{\frac {\partial l}{\partial y_{i}^{(k)}}}} , ∂ l ∂ σ B ( k ) 2 = ∑ i = 1 m ∂ l ∂ y i ( k ) ( x i ( k ) − μ B ( k ) ) ( − γ ( k ) 2 ( σ B ( k ) 2 + ϵ ) − 3 / 2 ) {\displaystyle {\frac {\partial l}{\partial \sigma _{B}^{(k)^{2}}}}=\sum _{i=1}^{m}{\frac {\partial l}{\partial y_{i}^{(k)}}}(x_{i}^{(k)}-\mu _{B}^{(k)})\left(-{\frac {\gamma ^{(k)}}{2}}(\sigma _{B}^{(k)^{2}}+\epsilon )^{-3/2}\right)} , ∂ l ∂ μ B ( k ) = ∑ i = 1 m ∂ l ∂ y i ( k ) − γ ( k ) σ B ( k ) 2 + ϵ + ∂ l ∂ σ B ( k ) 2 1 m ∑ i = 1 m ( − 2 ) ⋅ ( x i ( k ) − μ B ( k ) ) {\displaystyle {\frac {\partial l}{\partial \mu _{B}^{(k)}}}=\sum _{i=1}^{m}{\frac {\partial l}{\partial y_{i}^{(k)}}}{\frac {-\gamma ^{(k)}}{\sqrt {\sigma _{B}^{(k)^{2}}+\epsilon }}}+{\frac {\partial l}{\partial \sigma _{B}^{(k)^{2}}}}{\frac {1}{m}}\sum _{i=1}^{m}(-2)\cdot (x_{i}^{(k)}-\mu _{B}^{(k)})} , and ∂ l ∂ x i ( k ) = ∂ l ∂ x ^ i ( k ) 1 σ B ( k ) 2 + ϵ + ∂ l ∂ σ B ( k ) 2 2 ( x i ( k ) − μ B ( k ) ) m + ∂ l ∂ μ B ( k ) 1 m {\displaystyle {\frac {\partial l}{\partial x_{i}^{(k)}}}={\frac {\partial l}{\partial {\hat {x}}_{i}^{(k)}}}{\frac {1}{\sqrt {\sigma _{B}^{(k)^{2}}+\epsilon }}}+{\frac {\partial l}{\partial \sigma _{B}^{(k)^{2}}}}{\frac {2(x_{i}^{(k)}-\mu _{B}^{(k)})}{m}}+{\frac {\partial l}{\partial \mu _{B}^{(k)}}}{\frac {1}{m}}} . === Inference === During the training stage, the normalization steps depend on the mini-batches to ensure efficient and reliable training. However, in the inference stage, this dependence is not useful any more. Instead, the normalization step in this stage is computed with the population statistics such that the output could depend on the input in a deterministic manner. The population mean, E [ x ( k ) ] {\displaystyle E[x^{(k)}]} , and variance, Var ⁡ [ x ( k ) ] {\displaystyle \operatorname {Var} [x^{(k)}]} , are computed as: E [ x ( k ) ] = E B [ μ B ( k ) ] {\displaystyle E[x^{(k)}]=E_{B}[\mu _{B}^{(k)}]} , and Var ⁡ [ x ( k ) ] = m m − 1 E B [ ( σ B ( k ) ) 2 ] {\displaystyle \operatorname {Var} [x^{(k)}]={\frac {m}{m-1}}E_{B}[\left(\sigma _{B}^{(k)}\right)^{2}]} . The population statistics thus is a complete representation of the mini-batches. The BN transform in the inference step thus becomes y ( k ) = B N γ ( k ) , β ( k ) inf ( x ( k ) ) = γ ( k ) x ( k ) − E [ x ( k ) ] Var ⁡ [ x ( k ) ] + ϵ + β

  Read more →
• 

  Argüman

  Argüman is a free and open source software for collective structured argumentation and argument analysis via argumentation graphs or argument maps in which the type of connections can be specified. It allows users to create collaborative "semantic maps" of arguments in well structured tree formats and share them with an audience and potential participants. Arguman.org was an open structured social debate platform that implemented the software. It is down as of 2023. There also is a mobile version of the tool. The project was started, in 2014, and largely built by developers in Turkey. Some studies used or investigated excerpts of argumentations on the platform. Unlike the larger and functional alternative Kialo, which is structured using only 'Pro' and 'Con' relations, argüman arguments are structured by three types of premises – 'because', 'but', and 'however'. As of the latest version, debates are presented in their entirety as a large tree which may be harder to navigate than other formats – for instance, trees "can become extremely dense, and the interface does not make it obvious which arguments the user should pay attention to". Users can also flag arguments for fallacies. Arguman.org also had a Turkish-language subdomain. A researcher suggested the concept of the Semantic Web-interoperability could be useful for argumentative structures on the Web, going beyond the conventional flat structures of discussions and lack of characterizations of their components as implemented in argüman. There is research into how to automatically use these collaborative argumentation graphs, which is a "very active" topic in Artificial Intelligence. There also is research into applying conclusion-making methods to the debates or their data, such as bipolar weighted argumentation frameworks – this could be a way to find out what the current conclusion of debates like "Computer Science is not actually a science" is. A study suggests it could be useful for the development of critical thinking skills.

  Read more →
• 

  Maximum inner-product search

  Maximum inner-product search (MIPS) is a search problem, with a corresponding class of search algorithms which attempt to maximise the inner product between a query and the data items to be retrieved. MIPS algorithms are used in a wide variety of big data applications, including recommendation algorithms and machine learning. Formally, for a database of vectors x i {\displaystyle x_{i}} defined over a set of labels S {\displaystyle S} in an inner product space with an inner product ⟨ ⋅ , ⋅ ⟩ {\displaystyle \langle \cdot ,\cdot \rangle } defined on it, MIPS search can be defined as the problem of determining a r g m a x i ∈ S ⟨ x i , q ⟩ {\displaystyle {\underset {i\in S}{\operatorname {arg\,max} }}\ \langle x_{i},q\rangle } for a given query q {\displaystyle q} . Although there is an obvious linear-time implementation, it is generally too slow to be used on practical problems. However, efficient algorithms exist to speed up MIPS search. Under the assumption of all vectors in the set having constant norm, MIPS can be viewed as equivalent to a nearest neighbor search (NNS) problem in which maximizing the inner product is equivalent to minimizing the corresponding distance metric in the NNS problem. Like other forms of NNS, MIPS algorithms may be approximate or exact. MIPS search is used as part of DeepMind's RETRO algorithm.

  Read more →
• 

  Woken Furies

  Woken Furies (2005) is a science fiction novel by British writer Richard Morgan. It is the third novel featuring the anti-hero Takeshi Kovacs and is the sequel to Broken Angels. This addition to the series casts light upon Kovacs' early life providing information on his post-envoy activities. Morgan's official website and interviews suggest that Woken Furies could be the last Kovacs novel, although in 2018 (before Netflix cancelled the show) Morgan stated that the Netflix adaptation has "kind of woken it all up again" after all these years, making him possibly reconsider being done with Kovacs. == Plot == Takeshi Kovacs finds himself in a new "sleeve," or human body, back on his home planet of Harlan's World. He is on the run after making numerous attacks against the Knights of the New Revelation, an extremist religious order responsible for the death of his lost love and her daughter. Because she had violated tenets about resleeving, her executioners dropped her and her daughter's cortical stacks in the sea, effectively preventing them from being resleeved (into new bodies). While trying to secure passage after his most recent attack, Kovacs saves a woman named Sylvie from a group of religious zealots. In return, she allows him to take refuge with her mercenary "deCom" crew as they head out to decommission sentient military hardware that has run amok on the island of New Hokkaido (AKA New Hok). Sylvie is the "command head" of her crew, co-ordinating them during missions by using her biologically implanted circuitry and software. During one of these missions, Sylvie collapses, regains consciousness, and Kovacs realizes that her personality seems to have been replaced by that of long-dead revolutionary leader Quellcrist Falconer. Harlan's World is surrounded by automated "orbitals" which target flying objects, such as vehicles, with high-energy beam weapons known as "angelfire"; Falconer is believed to have died without a backup of her cortical stack when her getaway aircraft was destroyed by angelfire 300 years prior. When Sylvie's crew returns from New Hok, they discover a younger version of Kovacs has been illegally duplicated into a different body (AKA "double sleeved") and is hunting them on behalf of the Harlan family that rules the planet. Most of Sylvie's crew is killed and Sylvie/Quellcrist is captured. Kovacs schemes to rescue Sylvie by approaching old criminal associates of his, the Little Blue Bugs. The Little Blue Bugs mount a semi-successful attack on a Harlan fortress and rescue Sylvie/Quellcrist. Hiding from Harlan forces in a floating base, the neo-Quellists are sold out by its owner and recaptured. An assault by Kovacs and a single UN Envoy on the base ends badly when Kovacs is betrayed by the Envoy who was actually embedded with several colleagues. However, Sylvie/Quellcrist has established a connection with the orbitals and calls down angelfire, eliminating their captors. The younger Kovacs is killed in the aftermath. Sylvie explains that angelfire is a destructive recording device. Thus, in destroying Quellcrist and the helicopter carrying her, it copied her. When the technology of the deCom crews advanced far enough, her persona was able to insert itself into Sylvie's implants and co-exist in her body. The novel ends with Kovacs, Virginia Vidaura, and Sylvie/Quellcrist waiting to see if they can use Sylvie/Quellcrist's newfound connection to the orbitals and the expansion of a long-dormant genetic virus to turn the population against the ruling oligarchy.

  Read more →
• 

  Transdermal optical imaging

  Transdermal optical imaging, also known as transdermal optical imagery or TOI, is a method of detecting blood flow of the face by measuring hemoglobin concentration using a digital video camera. Because of the translucent property of skin, light can travel beneath the skin and re-emit. The re-emitted light from underneath the skin is affected by chromophores, mainly hemoglobin and melanin, which differ in color. The color difference allows TOI machine learning software to separate the images into layers, which are known as bitplanes. It extracts signals rich in hemoglobin and signals rich in melanin, then discards the melanin-rich signals to obtain a recording of hemoglobin changes under the skin. Transdermal optical imaging has been proposed as an alternative to cuff-based methods of measuring blood pressure because it is able to measure heart rate accurately in a "contactless and non-invasive" way. Transdermal optical imaging may be able to detect hidden emotions using the patterns of blood flow in the face.

  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 →
• 

  Weight initialization

  In deep learning, weight initialization or parameter initialization describes the initial step in creating a neural network. A neural network contains trainable parameters that are modified during training: weight initialization is the pre-training step of assigning initial values to these parameters. The choice of weight initialization method affects the speed of convergence, the scale of neural activation within the network, the scale of gradient signals during backpropagation, and the quality of the final model. Proper initialization is necessary for avoiding issues such as vanishing and exploding gradients and activation function saturation. Note that even though this article is titled "weight initialization", both weights and biases are used in a neural network as trainable parameters, so this article describes how both of these are initialized. Similarly, trainable parameters in convolutional neural networks (CNNs) are called kernels and biases, and this article also describes these. == Constant initialization == We discuss the main methods of initialization in the context of a multilayer perceptron (MLP). Specific strategies for initializing other network architectures are discussed in later sections. For an MLP, there are only two kinds of trainable parameters, called weights and biases. Each layer l {\displaystyle l} contains a weight matrix W ( l ) ∈ R n l − 1 × n l {\displaystyle W^{(l)}\in \mathbb {R} ^{n_{l-1}\times n_{l}}} and a bias vector b ( l ) ∈ R n l {\displaystyle b^{(l)}\in \mathbb {R} ^{n_{l}}} , where n l {\displaystyle n_{l}} is the number of neurons in that layer. A weight initialization method is an algorithm for setting the initial values for W ( l ) , b ( l ) {\displaystyle W^{(l)},b^{(l)}} for each layer l {\displaystyle l} . The simplest form is zero initialization: W ( l ) = 0 , b ( l ) = 0 {\displaystyle W^{(l)}=0,b^{(l)}=0} Zero initialization is usually used for initializing biases, but it is not used for initializing weights, as it leads to symmetry in the network, causing all neurons to learn the same features. In this page, we assume b = 0 {\displaystyle b=0} unless otherwise stated. Recurrent neural networks typically use activation functions with bounded range, such as sigmoid and tanh, since unbounded activation may cause exploding values. (Le, Jaitly, Hinton, 2015) suggested initializing weights in the recurrent parts of the network to identity and zero bias, similar to the idea of residual connections and LSTM with no forget gate. In most cases, the biases are initialized to zero, though some situations can use a nonzero initialization. For example, in multiplicative units, such as the forget gate of LSTM, the bias can be initialized to 1 to allow good gradient signal through the gate. For neurons with ReLU activation, one can initialize the bias to a small positive value like 0.1, so that the gradient is likely nonzero at initialization, avoiding the dying ReLU problem. == Random initialization == Random initialization means sampling the weights from a normal distribution or a uniform distribution, usually independently. === LeCun initialization === LeCun initialization, popularized in (LeCun et al., 1998), is designed to preserve the variance of neural activations during the forward pass. It samples each entry in W ( l ) {\displaystyle W^{(l)}} independently from a distribution with mean 0 and variance 1 / n l − 1 {\displaystyle 1/n_{l-1}} . For example, if the distribution is a continuous uniform distribution, then the distribution is U ( ± 3 / n l − 1 ) {\displaystyle {\mathcal {U}}(\pm {\sqrt {3/n_{l-1}}})} . === Glorot initialization === Glorot initialization (or Xavier initialization) was proposed by Xavier Glorot and Yoshua Bengio. It was designed as a compromise between two goals: to preserve activation variance during the forward pass and to preserve gradient variance during the backward pass. For uniform initialization, it samples each entry in W ( l ) {\displaystyle W^{(l)}} independently and identically from U ( ± 6 / ( n l + 1 + n l − 1 ) ) {\displaystyle {\mathcal {U}}(\pm {\sqrt {6/(n_{l+1}+n_{l-1})}})} . In the context, n l − 1 {\displaystyle n_{l-1}} is also called the "fan-in", and n l + 1 {\displaystyle n_{l+1}} the "fan-out". When the fan-in and fan-out are equal, then Glorot initialization is the same as LeCun initialization. === He initialization === As Glorot initialization performs poorly for ReLU activation, He initialization (or Kaiming initialization) was proposed by Kaiming He et al. for networks with ReLU activation. It samples each entry in W ( l ) {\displaystyle W^{(l)}} from N ( 0 , 2 / n l − 1 ) {\displaystyle {\mathcal {N}}(0,2/n_{l-1})} . === Orthogonal initialization === (Saxe et al. 2013) proposed orthogonal initialization: initializing weight matrices as uniformly random (according to the Haar measure) semi-orthogonal matrices, multiplied by a factor that depends on the activation function of the layer. It was designed so that if one initializes a deep linear network this way, then its training time until convergence is independent of depth. Sampling a uniformly random semi-orthogonal matrix can be done by initializing X {\displaystyle X} by IID sampling its entries from a standard normal distribution, then calculate ( X X ⊤ ) − 1 / 2 X {\displaystyle \left(XX^{\top }\right)^{-1/2}X} or its transpose, depending on whether X {\displaystyle X} is tall or wide. For CNN kernels with odd widths and heights, orthogonal initialization is done this way: initialize the central point by a semi-orthogonal matrix, and fill the other entries with zero. As an illustration, a kernel K {\displaystyle K} of shape 3 × 3 × c × c ′ {\displaystyle 3\times 3\times c\times c'} is initialized by filling K [ 2 , 2 , : , : ] {\displaystyle K[2,2,:,:]} with the entries of a random semi-orthogonal matrix of shape c × c ′ {\displaystyle c\times c'} , and the other entries with zero. (Balduzzi et al., 2017) used it with stride 1 and zero-padding. This is sometimes called the Orthogonal Delta initialization. Related to this approach, unitary initialization proposes to parameterize the weight matrices to be unitary matrices, with the result that at initialization they are random unitary matrices (and throughout training, they remain unitary). This is found to improve long-sequence modelling in LSTM. Orthogonal initialization has been generalized to layer-sequential unit-variance (LSUV) initialization. It is a data-dependent initialization method, and can be used in convolutional neural networks. It first initializes weights of each convolution or fully connected layer with orthonormal matrices. Then, proceeding from the first to the last layer, it runs a forward pass on a random minibatch, and divides the layer's weights by the standard deviation of its output, so that its output has variance approximately 1. === Fixup initialization === In 2015, the introduction of residual connections allowed very deep neural networks to be trained, much deeper than the ~20 layers of the previous state of the art (such as the VGG-19). Residual connections gave rise to their own weight initialization problems and strategies. These are sometimes called "normalization-free" methods, since using residual connection could stabilize the training of a deep neural network so much that normalizations become unnecessary. Fixup initialization is designed specifically for networks with residual connections and without batch normalization, as follows: Initialize the classification layer and the last layer of each residual branch to 0. Initialize every other layer using a standard method (such as He initialization), and scale only the weight layers inside residual branches by L − 1 2 m − 2 {\displaystyle L^{-{\frac {1}{2m-2}}}} . Add a scalar multiplier (initialized at 1) in every branch and a scalar bias (initialized at 0) before each convolution, linear, and element-wise activation layer. Similarly, T-Fixup initialization is designed for Transformers without layer normalization. === Others === Instead of initializing all weights with random values on the order of O ( 1 / n ) {\displaystyle O(1/{\sqrt {n}})} , sparse initialization initialized only a small subset of the weights with larger random values, and the other weights zero, so that the total variance is still on the order of O ( 1 ) {\displaystyle O(1)} . Random walk initialization was designed for MLP so that during backpropagation, the L2 norm of gradient at each layer performs an unbiased random walk as one moves from the last layer to the first. Looks linear initialization was designed to allow the neural network to behave like a deep linear network at initialization, since W R e L U ( x ) − W R e L U ( − x ) = W x {\displaystyle W\;\mathrm {ReLU} (x)-W\;\mathrm {ReLU} (-x)=Wx} . It initializes a matrix W {\displaystyle W} of shape R n 2 × m {\displaystyle \mathbb {R} ^{{\frac {n}{2}}\times m}} by any method, such as orthogonal initialization, t

  Read more →
• 

  Conference on Neural Information Processing Systems

  The Conference on Neural Information Processing Systems (abbreviated as NeurIPS and formerly NIPS) is a machine learning and computational neuroscience conference held annually in December. Along with ICLR and ICML, it is one of the three primary conferences of high impact in machine learning and artificial intelligence research. The conference includes three days of invited talks along with oral and poster presentations of refereed papers, followed by two days of workshops and competitions. == History == The NeurIPS meeting was first proposed in 1986 at the annual invitation-only Snowbird Meeting on Neural Networks for Computing organized by The California Institute of Technology and Bell Laboratories. NeurIPS was designed as a complementary open interdisciplinary meeting for researchers exploring biological and artificial Neural Networks. Reflecting this multidisciplinary approach, NeurIPS began in 1987 with information theorist Ed Posner as the conference president and learning theorist Yaser Abu-Mostafa as program chairman. Research presented in the early NeurIPS meetings included a wide range of topics from efforts to solve purely engineering problems to the use of computer models as a tool for understanding biological nervous systems. Since then, the biological and artificial systems research streams have diverged, and recent NeurIPS proceedings have been dominated by papers on machine learning, artificial intelligence and statistics. From 1987 until 2000 NeurIPS was held in Denver, United States. Since then, the conference was held in Vancouver, Canada (2001–2010), Granada, Spain (2011), and Lake Tahoe, United States (2012–2013). In 2014 and 2015, the conference was held in Montreal, Canada, in Barcelona, Spain in 2016, in Long Beach, United States in 2017, in Montreal, Canada in 2018 and Vancouver, Canada in 2019. Reflecting its origins at Snowbird, Utah, the meeting was accompanied by workshops organized at a nearby ski resort up until 2013, when it outgrew ski resorts. The first NeurIPS Conference was sponsored by the IEEE. The following NeurIPS Conferences have been organized by the NeurIPS Foundation, established by Ed Posner. Terrence Sejnowski has been the president of the NeurIPS Foundation since Posner's death in 1993. The board of trustees consists of previous general chairs of the NeurIPS Conference. The first proceedings was published in book form by the American Institute of Physics in 1987, and was entitled Neural Information Processing Systems, then the proceedings from the following conferences have been published by Morgan Kaufmann (1988–1993), MIT Press (1994–2004) and Curran Associates (2005–present) under the name Advances in Neural Information Processing Systems. The conference was originally abbreviated as "NIPS". By 2018 a few commentators were criticizing the abbreviation as encouraging sexism due to its association with the word nipples, and as being a slur against Japanese. The board changed the abbreviation to "NeurIPS" in November 2018. == Topics == Along with machine learning and neuroscience, other fields represented at NeurIPS include cognitive science, psychology, computer vision, statistical linguistics, and information theory. Over the years, NeurIPS became a premier conference on machine learning and although the 'Neural' in the NeurIPS acronym had become something of a historical relic, the resurgence of deep learning in neural networks since 2012, fueled by faster computers and big data, has led to achievements in speech recognition, object recognition in images, image captioning, language translation and world championship performance in the game of Go, based on neural architectures inspired by the hierarchy of areas in the visual cortex (ConvNet) and reinforcement learning inspired by the basal ganglia (Temporal difference learning). Notable affinity groups have emerged from the NeurIPS conference and displayed diversity, including Black in AI (in 2017), Queer in AI (in 2016), and others. === Named lectures === In addition to invited talks and symposia, NeurIPS also organizes two named lectureships to recognize distinguished researchers. The NeurIPS Board introduced the Posner Lectureship in honor of NeurIPS founder Ed Posner; two Posner Lectures were given each year up to 2015. Past lecturers have included: 2010 – Josh Tenenbaum and Michael I. Jordan 2011 – Rich Sutton and Bernhard Schölkopf 2012 – Thomas Dietterich and Terry Sejnowski 2013 – Daphne Koller and Peter Dayan 2014 – Michael Kearns and John Hopfield 2015 – Zoubin Ghahramani and Vladimir Vapnik 2016 – Yann LeCun 2017 – John Platt 2018 – Joëlle Pineau 2019 – Yoshua Bengio 2020 – Christopher Bishop 2021 – Peter Bartlett In 2015, the NeurIPS Board introduced the Breiman Lectureship to highlight work in statistics relevant to conference topics. The lectureship was named for statistician Leo Breiman, who served on the NeurIPS Board from 1994 to 2005. Past lecturers have included: 2015 – Robert Tibshirani 2016 – Susan Holmes 2017 – Yee Whye Teh 2018 – David Spiegelhalter 2019 – Bin Yu 2020 – Marloes Maathuis 2021 – Gabor Lugosi 2022 – Emmanuel Candes 2023 – Susan Murphy 2024 – Arnaud Doucet == NeurIPS consistency experiment == In NIPS 2014, the program chairs duplicated 10% of all submissions and sent them through separate reviewers to evaluate randomness in the reviewing process. Several researchers interpreted the result. Regarding whether the decision in NIPS is completely random or not, John Langford writes: "Clearly not—a purely random decision would have arbitrariness of ~78%. It is, however, quite notable that 60% is much closer to 78% than 0%." He concludes that the result of the reviewing process is mostly arbitrary. In NeurIPS 2021, the program chairs repeated the 2014 experiment and found similar levels of review inconsistency; 23% of duplicated submissions received different accept/reject decisions, and 50.6% of accepted papers would have been rejected under re-review. == Locations == 1987–2000: Denver, Colorado, United States 2001–2010: Vancouver, British Columbia, Canada 2011: Granada, Spain 2012 & 2013: Stateline, Nevada, United States 2014 & 2015: Montréal, Quebec, Canada 2016: Barcelona, Spain 2017: Long Beach, California, United States 2018: Montréal, Quebec, Canada 2019: Vancouver, British Columbia, Canada 2020: Vancouver, British Columbia, Canada (virtual conference) 2021: Virtual conference 2022 & 2023: New Orleans, Louisiana, United States 2024: Vancouver, British Columbia, Canada 2025: San Diego, California, United States and Mexico City, Mexico 2026: Sydney, New South Wales, Australia, with satellite events in Atlanta and Paris

  Read more →
• 

  SpreeAI

  SpreeAI (stylized as SPREEAI) is an American fashion technology company headquartered in Incline Village, Nevada that develops artificial intelligence software for the apparel and retail industries, including photorealistic virtual try-on, AI-powered sizing recommendations, and digital model generation. Founded in 2022 by John Imah and Bob Davidson, the company achieved unicorn status in 2025 following a Series B round led by Davidson Group that valued the company at approximately US$1.5 billion. TechCrunch identified SpreeAI as one of the more than 100 new tech unicorns minted in 2025. Its board of directors includes supermodel Naomi Campbell and hospitality executive Larry Ruvo. == History == SpreeAI was founded in 2022 by John Imah and Bob Davidson with a focus on artificial intelligence applications in fashion retail. By 2024, the company had raised approximately US$60 million in venture funding. In May 2025, SpreeAI announced a Series B round led by Davidson Group; reporting at the time placed the company's valuation at approximately US$1.5 billion, making it one of a small number of fashion-technology companies to reach unicorn status. In January 2026, TechCrunch listed SpreeAI among the more than 100 new tech unicorns minted in 2025. == Technology == SpreeAI develops a suite of artificial intelligence tools for the apparel industry. Its consumer-facing platform allows shoppers to upload a single photograph or select a digital model and then visualize clothing items on that figure with photorealistic rendering, while a complementary sizing engine generates fit recommendations intended to reduce returns. The platform is designed for integration with online retailers so that shoppers can preview garments before purchase. The company has stated that its models were developed in part through research collaborations with the Massachusetts Institute of Technology and Carnegie Mellon University. == Leadership and board == John Imah, a Nigerian-American technology executive who previously held roles at Samsung, Twitch, Meta Platforms, and Snap Inc., is co-founder and chief executive officer. Co-founder Bob Davidson, through Davidson Group, led the company's Series B financing. The company's board of directors includes supermodel Naomi Campbell, who joined in 2024, and Las Vegas hospitality executive Larry Ruvo. == Partnerships == SpreeAI has formed partnerships across both academia and the fashion industry. Council of Fashion Designers of America (CFDA). In 2025, SpreeAI entered a partnership with the CFDA to support American designers and brands with AI-driven tools; the CFDA described SpreeAI as "a fashion technology leader delivering innovative solutions to help designers and brands thrive." Massachusetts Institute of Technology and Carnegie Mellon University. The company has cited ongoing research and talent collaborations with both institutions. Sergio Hudson and Kai Collective. In 2025, SpreeAI made what WWD described as its Met Gala debut through a custom collaboration with designer Sergio Hudson and Nigerian-British label Kai Collective; the collaboration paired Hudson's couture with SpreeAI's virtual try-on platform. == Recognition == In 2025, TechCrunch named SpreeAI among the new tech unicorns of the year. In 2025, SpreeAI was named an honoree in Inc.'s Best in Business awards, and CEO John Imah was included on Inc.'s list of 40 business leaders who "propelled their organizations to success." In 2025, Imah was named to the Observer's AI Power Index, a list of 100 leaders shaping the future of artificial intelligence. In 2025, Imah was included in AfroTech's Future 50, recognizing Black innovators in technology. SpreeAI and Imah have been the subject of profile coverage in The Washington Post, Rolling Stone UK, WWD, Vogue UA, L'Officiel Arabia, GQ South Africa, and Inc..

  Read more →
• 

  Tensor network

  Tensor networks or tensor network states are a class of variational wave functions used in the study of many-body quantum systems and fluids. Tensor networks extend one-dimensional matrix product states to higher dimensions while preserving some of their useful mathematical properties. The wave function is encoded as a tensor contraction of a network of individual tensors. The structure of the individual tensors can impose global symmetries on the wave function (such as antisymmetry under exchange of fermions) or restrict the wave function to specific quantum numbers, like total charge, angular momentum, or spin. It is also possible to derive strict bounds on quantities like entanglement and correlation length using the mathematical structure of the tensor network. This has made tensor networks useful in theoretical studies of quantum information in many-body systems. They have also proved useful in variational studies of ground states, excited states, and dynamics of strongly correlated many-body systems. == Diagrammatic notation == In general, a tensor network diagram (Penrose diagram) can be viewed as a graph where nodes (or vertices) represent individual tensors, while edges represent summation over an index. Free indices are depicted as edges (or legs) attached to a single vertex only. Sometimes, there is also additional meaning to a node's shape. For instance, one can use trapezoids for unitary matrices or tensors with similar behaviour. This way, flipped trapezoids would be interpreted as complex conjugates to them. == History == Foundational research on tensor networks began in 1971 with a paper by Roger Penrose. In "Applications of negative dimensional tensors" Penrose developed tensor diagram notation, describing how the diagrammatic language of tensor networks could be used in applications in physics. In 1992, Steven R. White developed the density matrix renormalization group (DMRG) for quantum lattice systems. The DMRG was the first successful tensor network and associated algorithm. In 2002, Guifré Vidal and Reinhard Werner attempted to quantify entanglement, laying the groundwork for quantum resource theories. This was also the first description of the use of tensor networks as mathematical tools for describing quantum systems. In 2004, Frank Verstraete and Ignacio Cirac developed the theory of matrix product states, projected entangled pair states, and variational renormalization group methods for quantum spin systems. In 2006, Vidal developed the multi-scale entanglement renormalization ansatz (MERA). In 2007 he developed entanglement renormalization for quantum lattice systems. In 2010, Ulrich Schollwock developed the density-matrix renormalization group for the simulation of one-dimensional strongly correlated quantum lattice systems. In 2014, Román Orús introduced tensor networks for complex quantum systems and machine learning, as well as tensor network theories of symmetries, fermions, entanglement and holography. == Connection to machine learning == Tensor networks have been adapted for supervised learning, taking advantage of similar mathematical structure in variational studies in quantum mechanics and large-scale machine learning. This crossover has spurred collaboration between researchers in artificial intelligence and quantum information science. In June 2019, Google, the Perimeter Institute for Theoretical Physics, and X (company), released TensorNetwork, an open-source library for efficient tensor calculations. The main interest in tensor networks and their study from the perspective of machine learning is to reduce the number of trainable parameters (in a layer) by approximating a high-order tensor with a network of lower-order ones. Using the so-called tensor train technique (TT), one can reduce an N-order tensor (containing exponentially many trainable parameters) to a chain of N tensors of order 2 or 3, which gives us a polynomial number of parameters.

  Read more →
• 

  Test data management

  Test data management (TDM) is a process in software testing concerned with the creation, preparation, and control of data used for testing software systems. It involves supplying datasets required to execute test cases and verifying system behaviour under defined conditions. Test data management is an integral part of the software development lifecycle (SDLC) and is utilized in both manual and automated testing processes. It is applied in environments that use continuous integration and DevOps practices, where test execution requires consistent and repeatable data conditions. == Overview == Test data management includes the generation, selection, and preparation of data for testing purposes, as well as its distribution across test environments. It also involves controlling data versions and ensuring that datasets correspond to specific test scenarios. In many cases, production data is adapted for testing through techniques such as masking or subsetting to reduce size and remove sensitive content. Test data management ensures that test cases are executed with relevant, consistent, and readily available data. This reduces variability in test results and supports reproducibility across test cycles. == Importance == The role of test data management has expanded with the growth of complex, data-driven systems and regulatory requirements governing data usage. Testing often depends on data that reflects real-world conditions, but direct use of production data may introduce security and privacy risks. As a result, organizations apply methods such as data masking and anonymization to meet compliance requirements, including those set by the California Privacy Rights Act (CPRA) and Europe’s General Data Protection Regulation (GDPR). Inadequate control of test data can lead to incomplete test coverage, unreliable test results, or delays in testing processes due to unavailable or inconsistent datasets. == Techniques and tools == Test data management leverages various techniques for preparing and controlling data used in testing. These include the generation of synthetic data, the extraction of subsets from production datasets, and the modification of data to remove or obscure sensitive information. A key technical requirement in these processes is maintaining referential integrity, or ensuring that relationships between data entities remain consistent across different tables and systems after masking or subsetting. Data virtualization is also used to provide access to datasets without full replication. These methods may be implemented using software tools that automate data preparation, masking, and distribution.

  Read more →
• 

  Emma Hart (computer scientist)

  Professor Emma Hart, FRSE (born 1967) is an English computer scientist known for her work in artificial immune systems (AIS), evolutionary computation and optimisation. She is a professor of computational intelligence at Edinburgh Napier University, editor-in-chief of the Journal of Evolutionary Computation (MIT Press), and D. Coordinator of the Future & Emerging Technologies (FET) Proactive Initiative, Fundamentals of Collective Adaptive Systems. == Early life and education == Hart was born in Middlesbrough, England in 1967. In 1990 she graduated from the University of Oxford with a first class BA(Hons) in Chemistry. She then continued her studies at the University of Edinburgh, graduating with an MSc in Artificial Intelligence in 1994, followed by a PhD that explored the use of immunology as an inspiration for computing, examining a range of techniques applied to optimization and data classification problems. Her dissertation was titled Immunology as a metaphor for computational information processing: Fact or fiction?, and her doctoral advisor was Peter Ross. == Career == In 2000 Hart took a position as a lecturer at Edinburgh Napier University, and was promoted to a Reader, Professor, and in 2008 Chair in Natural Computation. She is now director of the Centre of Algorithms, Visualisation and Evolving Systems (CAVES) group in the School of Computing. She continues to research in the area of developing novel bio-inspired techniques for solving a range of real-world optimisation and classification problems, as well as exploring the fundamental properties of immune-inspired computing through modelling and simulation. She is also involved in editorial activity and currently occupies the position of Editor-in-Chief of the Journal of Evolutionary Computation (MIT Press). Her interests lie in the area of bio-inspired computing, in particular artificial immune systems (AIS). She also undertakes research in three main areas: optimisation, self-organising/self-adaptive systems, and artificial intelligence. Hart is D. Coordinator of Fundamentals of Collective Adaptive Systems (FoCAS), a Future and Emerging Technologies Proactive Initiative funded by the European Commission under FP7. == Selected works == === Conference talks === Hart, Emma. "Lifelong learning in optimization (video)". 28th European Conference on Operational Research. The Association of European Operational Research Societies. Hart, Emma (December 2021). "Self-assembling robots and the potential of artificial evolution". TED talk 2021. === Journal articles === "An immune system approach to scheduling in changing environments". E.Hart, P.Ross. 1999. Proceedings of the 1st Annual Conference on Genetic and Evolutionary Computation (2), 1559–1566. "Exploiting the analogy between immunology and sparse distributed memories: A system for clustering non-stationary data". E.Hart, P.Ross. 2002. 1st International Conference on Artificial Immune Systems. "Evolutionary scheduling: A review". E Hart, P Ross, D Corne. 2005. Genetic Programming and Evolvable Machines 6(2), 191–220. DOI: https://doi.org/10.1007/s10710-005-7580-7 "Application areas of AIS: The past, the present and the future". E.Hart, J.Timmis. 2008. Applied soft computing 8(1), 191–201. DOI: https://doi.org/10.1016/j.asoc.2006.12.004 "Structure versus function: a topological perspective on immune networks". E.Hart, H.Bersini, F.Santos. 2010. Natural computing 9(3), 603–624. DOI: https://doi.org/10.1007/s11047-009-9138-8 "On the life-long learning capabilities of a nelli: A hyper-heuristic optimisation system". E.Hart, K.Sim. 2014. International Conference on Parallel Problem Solving from Nature, 282–291. DOI: https://doi.org/10.1007/978-3-319-10762-2_28 "A hyper-heuristic ensemble method for static job-shop scheduling". E.Hart, K.Sim. 2016. Evolutionary computation 24(4), 609-635. DOI: https://dx.doi.org/10.1162/EVCO_a_00183 == Awards and recognition == 2016, Featured article on Lifelong Learning in Optimisation, IFORS newsletter 2016, "A Combined Generative and Selective Hyper-heuristic for the Vehicle Routing Problem" presented at GECCO 2016 (Denver, USA), ACM 2016, "A Hybrid Parameter Control Approach Applied to a Diversity-based Multi-objective Memetic Algorithm for Frequency Assignment Problems" presented at WCCI 2016 (Vancouver, Canada), IEEE 2017, Keynote Speaker, 2017 International Joint Conference on Computational Intelligence 2018, Bronze Award in International Human-Competitive Awards (Humies), International Conference on Genetic and Evolutionary Computation, Kyoto Japan 2018, Nomination for best paper award, GECCO 18, Kyoto, Japan 2022, Elected Fellow of the Royal Society of Edinburgh

  Read more →
• 

  The Sword in the Stoned

  "The Sword in the Stoned" is the fifth episode of the second season of the American fantasy comedy television series Ted. Written by Julius Sharpe, and directed by Seth MacFarlane, it premiered on the American streaming service Peacock, along with the rest of season two, on March 5, 2026. The series acts as a precursor to the Ted film franchise, showcasing the childhood lives of the protagonists. The series, set in 1994, focuses on John Bennett (Max Burkholder), the series' primary protagonist, an awkward high-school aged boy; along with Ted (MacFarlane), the series' titular anthropomorphic teddy bear. The two live with John's family, Susan (Alanna Ubach), his mild mannered mother, and Matty (Scott Grimes), his conservative father. Also residing with the family is Blaire (Giorgia Whigham), his radically liberal cousin whom often clashes with Matty. In the episode, Ted and John join the school play so they can have more extracurricular activities for their college applications, but the latter grows a connection with the school's popular teenager, Erin (Francesca Xuereb). Concurrently, Susan and Matty get a job at Dunkin' Donuts to help with their financial troubles, and Matty is given an opportunity to tell off Bill Clinton. Burkholder wore prop armor during the episode's play scenes. Bill Clinton’s appearance in the episode was portrayed by MacFarlane. After conventional makeup and visual techniques failed to convincingly resemble Clinton, the production used artificial intelligence to digitally replace MacFarlane's face with Clinton's likeness. Upon release, the episode received generally positive reviews from critics, though the use of AI in the Clinton scene was polarizing among audiences and reviewers. == Plot == John tells Ted that he is the last single guy left at their school, to which Ted points out the popular, single cheerleader, Erin, but John dismisses this. At home, Blaire tells John that he needs extracurricular activities to get into college, while Susan and Matty discuss their financial troubles, especially regarding John's college tuition. Looking over their options, they decide to audition for a school production of the play Camelot. Matty takes a job at Dunkin' Donuts, despite being told that nobody will give him a tip, and having to wear an incorrect name tag. Waiting for their auditions, John and Ted watch several poor auditions for the play before seeing Erin's, who delivers a flawless performance; John and Ted do less serious auditions, getting cast as knights, while Erin gets the role of Guinevere. Matty complains about his low salary, and Susan decides to get a job at Dunkin' Donuts beside him to help earn more income. Erin clashes with Lancelot's actor while rehearsing, and John compliments her performance, which she ignores, but, seeing Ted and John give good performances in a repetition exercise, she becomes interested in him, particularly since he treats her better than her stage-partner. Matty and Susan watch an employee training video, explaining how they should treat customers politely, not affecting Matty's nihilistic attitude. The manager announces that Bill Clinton is visiting their Dunkin' Donuts for publicity, and Matty sees this as a chance to tell Bill off. John and Erin practice lines, as she reveals the show is being taped so it can be sent to Emerson College in hopes of her getting in; Erin asks John to go out with her after the show. At dinner, Matty enthusiastically reveals what he plans to tell Bill, as John becomes stressed about the play when Susan tells there will be a large audience. Bill comes to the Dunkin' Donuts, and, seeing Matty is nervously insulting him, stages a private meeting with him, where Bill yells at Matty, calling him a loser before posing for a picture with Matty and subsequently throwing the cold coffee onto him. To ease the pressure, Ted and John take edibles from Blaire, but learn at the show that they contained mushrooms, causing them to stress further. On stage, Ted and John yell nervously that they're on drugs as the latter urinates in his costume, causing Erin to angrily storm off. == Production == "The Sword in the Stoned" was directed by series creator and lead Seth MacFarlane, and written by Julius Sharpe in his third and final writing credit for the series. When Ted and John are doing repetition exercises, they tackle each other to the ground, which required a stuntman named Ashton to play the role of Ted, according to Max Burkholder, who portrays John. Burkholder also recalled that, when Ted was choking John in the scene, he kept making a noise during the choking, which made Bill, the cameraman, laugh, despite being a "stone face" that never laughs, noting that seeing him be amused by the noise he was making assured Burkholder that what he was doing was "hilarious". Burkholder found the filming of the play scenes "weird", as he was put in fake armor with a hose inside his suit—which was filled with water mixed with yellow food coloring—that was made to create the urine stream that comes out of John's armor in the episode; he also noted that it took around 45 minutes to put on and take off the armor. He revealed that he himself had to urinate during the filming, as doing a scene about a character having to do so "really [broke] my brain", with the fact that it took 45 minutes to get the suit off adding to the frustration. Jennifer Ashley Connell, who worked for wardrobe, had to repeatedly go to Burkholder quickly between takes to dry off his pants with two hair dryers to make it look like the fake urine hadn't already streamed down his pants, so they could get as many shots of it as possible. Francesca Xuereb guest stars in the episode as Erin, the cheerleader who stars in the play. Incumbent president Bill Clinton was portrayed by MacFarlane, with artificial intelligence (AI) being used to digitally make MacFarlane's face look like Clinton's during post-production. Before settling on AI, the crew tried to use traditional computer-generated imagery and prosthetics, which made him look "terrifying", resulting in them deciding that AI would give them a more accurate look. One of the original technologies considered was one where, after scanning MacFarlane, a mesh of his head was created, and they had to use computer graphics to replace MacFarlane's face with Clinton's. An issue was faced, however, when they found the archival footage used as reference from the Clinton Library—an official Presidential Library containing information related to Clinton—to be extremely low-quality, making it hard to properly emulate his face, since only still images were of acceptable quality, and there weren't references of his moving face to work off of. A forensic artist was hired to help with this, and they created a 3D model of Clinton's head in ZBrush, based off of his presidential portrait. The model head worked for still frames, but movement was still difficult to do realistically, due to it being made for a "single-point perspective", which made details like the cheekbones or other minor issues more noticeable when using it for the scene. Since this did not work, AI was ultimately chosen through the studio Deep Voodoo, which used large language models to teach the tool how to correctly replicate Clinton's appearance. Defending the episode's use of AI, MacFarlane noted that the crew did not want people to focus on the tool being used, trying to utilize it in a way that wouldn't distract from the humor and narrative. Like the rest of the series, the episode was shot using ViewScreen; MacFarlane was able to act live with the cast as Ted due to ViewScreen, a technology that allows the production crew to visualize what Ted will look like in each scene in real time. == Release and reception == "The Sword in the Stoned" was first released on March 5, 2026, on the American streaming service Peacock, along with the rest of the second season. Nate Richards of Collider highlighted the Dunkin' Donuts subplot as an example of Scott Grimes delivering a "lot of laughs" through his performance as Matty. Dustin Rowles of Pajiba called "The Sword in the Stoned" one of the season's many episodes he'd recommend, particularly for the scenes of Ted and John being high on mushrooms during the play. Oppositely, Nick Valdez of ComicBook.com ranked the episode as the worst of the second season, criticizing it for not having a "huge impact" on the Bennett family dynamic like other episodes of the season do, and Susan and Matty's side story as the main reason he felt it was "[kept] from being great". Valdez noted the episode for likely being an advertisement for Dunkin' Donuts, calling the plot's ending scene involving Clinton the reason "it just all sticks out like a sore thumb". === Response to AI usage === The episode's use of AI for MacFarlane's portrayal of Clinton proved controversial, mainly on social media, where audiences asserted that the crew should have gotten an actor that resembl

  Read more →