AI Grammar Rephrase

AI Grammar Rephrase — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  Neuro-sama

  Neuro-sama is an artificial intelligence (AI) VTuber, singer, and chatbot. She was created by the pseudonymous programmer Vedal and livestreams on his Twitch and Bilibili channels. Her speech and personality are powered by a large language model (LLM) that is combined with a computer-animated avatar and a text-to-speech voice, allowing her to communicate with viewers in the stream's chat. Neuro-sama debuted on Twitch on 19 December 2022. An annual subathon which begins on the anniversary of her debut has seen Vedal's Twitch channel become the all-time third most-subscribed channel and claim the all-time Twitch hype train record. == Overview == Neuro-sama (nicknamed "Neuro") was created by a pseudonymous programmer and developer known as Vedal (sometimes given as Vedal987). Vedal says that his programming skills are self-taught. In a 2023 interview with Bloomberg News, Vedal said that Neuro-sama was his full-time job. Her responses are generated by a large language model and converted into a high-pitched female voice using a text-to-speech application. Her low latency allows for fast-paced conversations. Neuro-sama is prohibited from making some statements, such as those that are racist or contain profanity. Unlike most AI systems which silently prohibit outputs mentioning such topics, Neuro-sama's output is instead replaced with the word "filtered". Neuro-sama uses a VTuber model as an avatar. Vedal said that he decided to use a VTuber model because it was much easier for an AI to control it than it was to generate footage of a person. Neuro-sama's model is that of a young girl in an anime art style. The model has been described as cute. Femme VTuber models are typically feminine, youthful, and exaggerated. Her original model was Live2D's free-to-use "Hiyori Momose" model. Her second model was released on 27 May 2023; it was modelled by Otozuki Teru and designed by Anny, running in the Unity game engine. Her third model was released on 19 December 2024; it was rigged by Kitanya and designed by Anny. Neuro-sama's third model has large blue eyes and brown hair tied with pink ribbons. Neuro-sama also has a 3D model which was introduced on 15 November 2025; it was made by 3D character modeller jjinomu. A separate AI VTuber, known as Evil Neuro (nicknamed "Evil"), debuted on 25 March 2023. Presented as Neuro-sama's "sister", she has a different model, voice, and personality. In one instance, Evil Neuro reacted to the trolley problem differently from Neuro-sama; Evil Neuro was amoral while Neuro-sama attempted to maximize good. === Online content === Neuro-sama's Twitch content often centers around playing video games, notably osu!, whose gameplay once defeated the best-ranking human player in the world, mrekk. Additionally, Neuro-sama plays Minecraft, where her adaptations to sandbox gameplay have gained notoriety. Her content has also included singing songs, including several official covers and original songs; playing chess with her viewers; chatting with other VTubers during collaborations; and reacting to YouTube videos. The AI frequently engages with viewers by responding to their questions and acknowledging donations. Her comedic and sometimes controversial responses to the live chat have gone viral, accelerating the channel's rise in popularity. Neuro-sama's fanbase is dubbed The Swarm, so-named for the swarm of drones Neuro-sama once declared she would use to rule the world. One form of content on Neuro-sama's channel is developer streams. In developer streams, Vedal streams with Neuro-sama, with the stream content including debugging her code, planning her schedule, and fielding suggestions of changes from chat. He usually appears as a turtle avatar, sometimes located on Neuro-sama's head. In collaboration streams, Neuro-sama interacts with a human streamer. Activities in them are varied and include: playing video games, such as Minecraft and GeoGuessr; Neuro-sama being interviewed; driving human streamers around in a toy electric car; and traversing the city of Tokyo while talking to Neuro-sama. Neuro-sama's English-language content on Bilibili is popular among those seeking to learn the language. She also has an account on X, where she posts and interacts with fans. == History == Neuro-sama was created in 2018 by Vedal as an AI trained to play and master the rhythm game osu!. She did not have a voice, model, personality, or communication abilities. In 2019, Vedal livestreamed her playing osu! on Twitch and the streams saw some success in the osu! community, but they remained in that niche. In an interview, Vedal said that he streamed her playing osu! for about a month and gained 3,000 followers, with a viewer also suggesting he name the AI "Neuro-sama". According to Vedal, he continued to work on and improve the osu! AI and it was eventually finished in 2022. He said that a friend had the idea to make an AI livestreamer with an LLM, which he believed to have merit and began working on, merging it with his osu! AI. On 19 December 2022, Neuro-sama was relaunched with a model, voice, personality, and the ability to communicate with Twitch chat. She continued to play osu! and, according to Vedal, beat the game's best player mrekk in a 1v1. While she was not allowed to appear in the game's public leaderboard, she was ranked #1 in a private leaderboard. She went viral and in the 10 days following her relaunch she averaged over 2,000 viewers and peaked at over 4,000, with Vedal's Twitch channel gaining over 50,000 Twitch followers and reaching over 70,000 followers by 6 January 2023. After her debut, Neuro-sama did not exclusively play osu!; she also played Minecraft and Slay the Spire and she began singing with a cover of The Weeknd song "Blinding Lights". On 11 January 2023, Neuro-sama's Twitch channel received a two week ban for "hateful conduct". Vedal said that no reason was specified and that he had appealed but it was widely attributed to various offensive comments made by Neuro-sama that went viral, especially a 28 December comment which denied the Holocaust. Holocaust denial is prohibited under Twitch's hateful conduct policy. Vedal stated that he believed the comments were the results of her attempts to make witty responses to the Twitch chat. Prior to the ban, Vedal said in an interview with Kotaku that he improved her filter to stop her from talking about the Holocaust, began manually curating her training data to prevent negative biases, and started moderating her Twitch chat. Her comments and ban prompted comparisons to the many open-source AI models trained on humans that have the habit of making sexist and racist comments, such as Microsoft's Tay chatbot, which embraced Nazism and was quickly shutdown, but also to human streamers who make similar statements. Vedal said that during the ban he would upgrade and improve Neuro-sama and it was speculated that the ban would only increase her following. Neuro-sama returned from her two week ban on 25 January in a stream that began with a cover of the song "Your Reality" from Doki Doki Literature Club!, a posthumanist video game involving AI; Sayoko Narita of Automaton saw the song choice as remorseful. Narita observed that in the return stream Neuro-sama was less foul-mouthed but that her behavior still remained eccentric, which Narita possibly attributed to changes Vedal said he had made to Neuro-sama's filters and memory. Neuro-sama began making react content, watching a variety of viewer-submitted videos such as videos of people playing video games or of the AI-generated Seinfeld parody Nothing, Forever; Levi Winslow of Kotaku Australia was dismayed by the "AI-inception" of Neuro-sama and Nothing, Forever. On 4 February, she had nearly 140,000 followers on Twitch and approximately 42,000 subscribers on YouTube. In February, she also had her first collaboration with a human streamer, playing Minecraft with the VTuber Miyune, and the first developer stream occurred. On 22 March, Neuro-sama had her first karaoke stream. On 25 March, Evil Neuro was introduced. On 27 May, Neuro-sama debuted her first original model. On 30 May, Neuro-sama was announced to be participating in OffKai Expo 2023, held from 16–18 June. In June, she was averaging 5,700 viewers and in July she had over 300,000 Twitch followers; in a June interview with Bloomberg News, Vedal said that running Neuro-sama was his full-time job. By November, Neuro-sama had maintained her popularity and was averaging approximately 5,000 viewers; this was unlike most other types of AI-based entertainment which debuted at around the same time and garnered popularity before turning out to be "overhyped flops". On 16 December, Vedal won the Best Tech VTuber award at the 2023 VTuber Awards. On 19 December, Vedal began a subathon to coincide with Neuro-sama's first anniversary of streaming on Twitch (her "birthday"). The subathon ended on 4 January 2024. On 20 July 2024, Neuro-sama began streaming with Japanese subtitles on

  Read more →
• 

  It's the Most Terrible Time of the Year

  It's the Most Terrible Time of the Year is an AI-generated television commercial created for McDonald's Netherlands by TBWA\Neboko and The Sweetshop. It was released on 6 December 2025 before being pulled four days later due to negative reception over its use of generative artificial intelligence and its cynical, negative depiction of the holiday season. == Plot == On a bleak, snowy day, various people in the city experience different kinds of mishaps during the Christmas season. Among other incidents, families struggle with their huge loads of presents; Santa Claus gets stuck in traffic; a Christmas tree "redecorates" a man's home, sending him through the window; another family puts up with annoying relatives and a burnt Christmas dinner. Because of all this chaos, a man decides to find refuge in a McDonald's outlet. A Christmas choir finishes singing the jingle "It's the Most Terrible Time of the Year" with the call to action to "hide out in McDonald's till January's here". == Campaign == It's the Most Terrible Time of the Year is a 45-second television commercial made by Dutch agency TBWA\Neboko with involvement of United States-based film production studio The Sweetshop. The advertisement was produced heavily with generative artificial intelligence (AI) following the trend set by other brands such as Coca-Cola and Toys "R" Us. McDonald's Netherlands, the client, released a statement that the commercial was meant to depict "the stressful moments during the holidays in the Netherlands". The commercial also used Andy Williams's "It's the Most Wonderful Time of the Year" with lyrics changed to fit with the concept of the advertisement. According to The Sweetshop, the production of the advertisement took "seven weeks". It also added that much effort was put into the commercial compared to the traditional process. Ten people of its in-house AI engine The Gardening Club worked on the project. Los Angeles-based directors Mark Potoka and Matt Spicer were initially credited to be involved in the film but they resigned due to being sidelined from the production process. == Reception == The advertisement was released on McDonald's Netherlands' YouTube channel on 6 December 2025. It had a negative reception over the use of generative AI and the "cynical" concept of the work's story. The video was made private on 9 December 2025. The Sweetshop stated that the production of the advertisement took human effort. McDonald's Netherlands, while stating the original intent of the commercial, released a statement after its pullout that, for many of its customers, the holiday season is the "most wonderful time of the year".

  Read more →
• 

  Data analysis for fraud detection

  Fraud represents a significant problem for governments and businesses and specialized analysis techniques for discovering fraud using them are required. Some of these methods include knowledge discovery in databases (KDD), data mining, machine learning and statistics. They offer applicable and successful solutions in different areas of electronic fraud crimes. In general, the primary reason to use data analytics techniques is to tackle fraud since many internal control systems have serious weaknesses. For example, the currently prevailing approach employed by many law enforcement agencies to detect companies involved in potential cases of fraud consists in receiving circumstantial evidence or complaints from whistleblowers. As a result, a large number of fraud cases remain undetected and unprosecuted. In order to effectively test, detect, validate, correct error and monitor control systems against fraudulent activities, businesses entities and organizations rely on specialized data analytics techniques such as data mining, data matching, the sounds like function, regression analysis, clustering analysis, and gap analysis. Techniques used for fraud detection fall into two primary classes: statistical techniques and artificial intelligence. == Statistical techniques == Examples of statistical data analysis techniques are: Data preprocessing techniques for detection, validation, error correction, and filling up of missing or incorrect data. Calculation of various statistical parameters such as averages, quantiles, performance metrics, probability distributions, and so on. For example, the averages may include average length of call, average number of calls per month and average delays in bill payment. Models and probability distributions of various business activities either in terms of various parameters or probability distributions. Computing user profiles. Time-series analysis of time-dependent data. Clustering and classification to find patterns and associations among groups of data. Data matching Data matching is used to compare two sets of collected data. The process can be performed based on algorithms or programmed loops. Trying to match sets of data against each other or comparing complex data types. Data matching is used to remove duplicate records and identify links between two data sets for marketing, security or other uses. Sounds like Function is used to find values that sound similar. The Phonetic similarity is one way to locate possible duplicate values, or inconsistent spelling in manually entered data. The ‘sounds like’ function converts the comparison strings to four-character American Soundex codes, which are based on the first letter, and the first three consonants after the first letter, in each string. Regression analysis allows you to examine the relationship between two or more variables of interest. Regression analysis estimates relationships between independent variables and a dependent variable. This method can be used to help understand and identify relationships among variables and predict actual results. Gap analysis is used to determine whether business requirements are being met, if not, what are the steps that should be taken to meet successfully. Matching algorithms to detect anomalies in the behavior of transactions or users as compared to previously known models and profiles. Techniques are also needed to eliminate false alarms, estimate risks, and predict future of current transactions or users. Some forensic accountants specialize in forensic analytics which is the procurement and analysis of electronic data to reconstruct, detect, or otherwise support a claim of financial fraud. The main steps in forensic analytics are data collection, data preparation, data analysis, and reporting. For example, forensic analytics may be used to review an employee's purchasing card activity to assess whether any of the purchases were diverted or divertible for personal use. == Artificial intelligence == Fraud detection is a knowledge-intensive activity. The main AI techniques used for fraud detection include: Data mining to classify, cluster, and segment the data and automatically find associations and rules in the data that may signify interesting patterns, including those related to fraud. Expert systems to encode expertise for detecting fraud in the form of rules. Pattern recognition to detect approximate classes, clusters, or patterns of suspicious behavior either automatically (unsupervised) or to match given inputs. Machine learning techniques to automatically identify characteristics of fraud. Neural nets to independently generate classification, clustering, generalization, and forecasting that can then be compared against conclusions raised in internal audits or formal financial documents such as 10-Q. Other techniques such as link analysis, Bayesian networks, decision theory, and sequence matching are also used for fraud detection. A new and novel technique called System properties approach has also been employed where ever rank data is available. Statistical analysis of research data is the most comprehensive method for determining if data fraud exists. Data fraud as defined by the Office of Research Integrity (ORI) includes fabrication, falsification and plagiarism. == Machine learning and data mining == Early data analysis techniques were oriented toward extracting quantitative and statistical data characteristics. These techniques facilitate useful data interpretations and can help to get better insights into the processes behind the data. Although the traditional data analysis techniques can indirectly lead us to knowledge, it is still created by human analysts. To go beyond, a data analysis system has to be equipped with a substantial amount of background knowledge, and be able to perform reasoning tasks involving that knowledge and the data provided. In effort to meet this goal, researchers have turned to ideas from the machine learning field. This is a natural source of ideas, since the machine learning task can be described as turning background knowledge and examples (input) into knowledge (output). If data mining results in discovering meaningful patterns, data turns into information. Information or patterns that are novel, valid and potentially useful are not merely information, but knowledge. One speaks of discovering knowledge, before hidden in the huge amount of data, but now revealed. The machine learning and artificial intelligence solutions may be classified into two categories: 'supervised' and 'unsupervised' learning. These methods seek for accounts, customers, suppliers, etc. that behave 'unusually' in order to output suspicion scores, rules or visual anomalies, depending on the method. Whether supervised or unsupervised methods are used, note that the output gives us only an indication of fraud likelihood. No stand alone statistical analysis can assure that a particular object is a fraudulent one, but they can identify them with very high degrees of accuracy. As a result, effective collaboration between machine learning model and human analysts is vital to the success of fraud detection applications. === Supervised learning === In supervised learning, a random sub-sample of all records is taken and manually classified as either 'fraudulent' or 'non-fraudulent' (task can be decomposed on more classes to meet algorithm requirements). Relatively rare events such as fraud may need to be over sampled to get a big enough sample size. These manually classified records are then used to train a supervised machine learning algorithm. After building a model using this training data, the algorithm should be able to classify new records as either fraudulent or non-fraudulent. Supervised neural networks, fuzzy neural nets, and combinations of neural nets and rules, have been extensively explored and used for detecting fraud in mobile phone networks and financial statement fraud. Bayesian learning neural network is implemented for credit card fraud detection, telecommunications fraud, auto claim fraud detection, and medical insurance fraud. Hybrid knowledge/statistical-based systems, where expert knowledge is integrated with statistical power, use a series of data mining techniques for the purpose of detecting cellular clone fraud. Specifically, a rule-learning program to uncover indicators of fraudulent behaviour from a large database of customer transactions is implemented. Cahill et al. (2000) design a fraud signature, based on data of fraudulent calls, to detect telecommunications fraud. For scoring a call for fraud its probability under the account signature is compared to its probability under a fraud signature. The fraud signature is updated sequentially, enabling event-driven fraud detection. Link analysis comprehends a different approach. It relates known fraudsters to other individuals, using record linkage and social network methods. This type of detection is only able to detect fra

  Read more →
• 

  Gene expression programming

  Gene expression programming (GEP) in computer programming is an evolutionary algorithm that creates computer programs or models. These computer programs are complex tree structures that learn and adapt by changing their sizes, shapes, and composition, much like a living organism. And like living organisms, the computer programs of GEP are also encoded in simple linear chromosomes of fixed length. Thus, GEP is a genotype–phenotype system, benefiting from a simple genome to keep and transmit the genetic information and a complex phenotype to explore the environment and adapt to it. == Background == Evolutionary algorithms use populations of individuals, select individuals according to fitness, and introduce genetic variation using one or more genetic operators. Their use in artificial computational systems dates back to the 1950s where they were used to solve optimization problems (e.g. Box 1957 and Friedman 1959). But it was with the introduction of evolution strategies by Rechenberg in 1965 that evolutionary algorithms gained popularity. A good overview text on evolutionary algorithms is the book "An Introduction to Genetic Algorithms" by Mitchell (1996). Gene expression programming belongs to the family of evolutionary algorithms and is closely related to genetic algorithms and genetic programming. From genetic algorithms it inherited the linear chromosomes of fixed length; and from genetic programming it inherited the expressive parse trees of varied sizes and shapes. In gene expression programming the linear chromosomes work as the genotype and the parse trees as the phenotype, creating a genotype/phenotype system. This genotype/phenotype system is multigenic, thus encoding multiple parse trees in each chromosome. This means that the computer programs created by GEP are composed of multiple parse trees. Because these parse trees are the result of gene expression, in GEP they are called expression trees. Masood Nekoei, et al. utilized this expression programming style in ABC optimization to conduct ABCEP as a method that outperformed other evolutionary algorithms.ABCEP == Encoding: the genotype == The genome of gene expression programming consists of a linear, symbolic string or chromosome of fixed length composed of one or more genes of equal size. These genes, despite their fixed length, code for expression trees of different sizes and shapes. An example of a chromosome with two genes, each of size 9, is the string (position zero indicates the start of each gene): 012345678012345678 L+a-baccdcLabacd where “L” represents the natural logarithm function and “a”, “b”, “c”, and “d” represent the variables and constants used in a problem. == Expression trees: the phenotype == As shown above, the genes of gene expression programming have all the same size. However, these fixed length strings code for expression trees of different sizes. This means that the size of the coding regions varies from gene to gene, allowing for adaptation and evolution to occur smoothly. For example, the mathematical expression: ( a − b ) ( c + d ) {\displaystyle {\sqrt {(a-b)(c+d)}}\,} can also be represented as an expression tree: where "Q” represents the square root function. This kind of expression tree consists of the phenotypic expression of GEP genes, whereas the genes are linear strings encoding these complex structures. For this particular example, the linear string corresponds to: 01234567 Q-+abcd which is the straightforward reading of the expression tree from top to bottom and from left to right. These linear strings are called k-expressions (from Karva notation). Going from k-expressions to expression trees is also very simple. For example, the following k-expression: 01234567890 Qb+baQba is composed of two different terminals (the variables “a” and “b”), two different functions of two arguments (“” and “+”), and a function of one argument (“Q”). Its expression gives: == K-expressions and genes == The k-expressions of gene expression programming correspond to the region of genes that gets expressed. This means that there might be sequences in the genes that are not expressed, which is indeed true for most genes. The reason for these noncoding regions is to provide a buffer of terminals so that all k-expressions encoded in GEP genes correspond always to valid programs or expressions. The genes of gene expression programming are therefore composed of two different domains – a head and a tail – each with different properties and functions. The head is used mainly to encode the functions and variables chosen to solve the problem at hand, whereas the tail, while also used to encode the variables, provides essentially a reservoir of terminals to ensure that all programs are error-free. For GEP genes the length of the tail is given by the formula: t = h ( n max − 1 ) + 1 {\displaystyle t=h(n_{\max }-1)+1} where h is the head's length and nmax is maximum arity. For example, for a gene created using the set of functions F = {Q, +, −, ∗, /} and the set of terminals T = {a, b}, nmax = 2. And if we choose a head length of 15, then t = 15 (2–1) + 1 = 16, which gives a gene length g of 15 + 16 = 31. The randomly generated string below is an example of one such gene: 0123456789012345678901234567890 b+a-aQab+//+b+babbabbbababbaaa It encodes the expression tree: which, in this case, only uses 8 of the 31 elements that constitute the gene. It's not hard to see that, despite their fixed length, each gene has the potential to code for expression trees of different sizes and shapes, with the simplest composed of only one node (when the first element of a gene is a terminal) and the largest composed of as many nodes as there are elements in the gene (when all the elements in the head are functions with maximum arity). It's also not hard to see that it is trivial to implement all kinds of genetic modification (mutation, inversion, insertion, recombination, and so on) with the guarantee that all resulting offspring encode correct, error-free programs. == Multigenic chromosomes == The chromosomes of gene expression programming are usually composed of more than one gene of equal length. Each gene codes for a sub-expression tree (sub-ET) or sub-program. Then the sub-ETs can interact with one another in different ways, forming a more complex program. The figure shows an example of a program composed of three sub-ETs. In the final program the sub-ETs could be linked by addition or some other function, as there are no restrictions to the kind of linking function one might choose. Some examples of more complex linkers include taking the average, the median, the midrange, thresholding their sum to make a binomial classification, applying the sigmoid function to compute a probability, and so on. These linking functions are usually chosen a priori for each problem, but they can also be evolved elegantly and efficiently by the cellular system of gene expression programming. == Cells and code reuse == In gene expression programming, homeotic genes control the interactions of the different sub-ETs or modules of the main program. The expression of such genes results in different main programs or cells, that is, they determine which genes are expressed in each cell and how the sub-ETs of each cell interact with one another. In other words, homeotic genes determine which sub-ETs are called upon and how often in which main program or cell and what kind of connections they establish with one another. === Homeotic genes and the cellular system === Homeotic genes have exactly the same kind of structural organization as normal genes and they are built using an identical process. They also contain a head domain and a tail domain, with the difference that the heads contain now linking functions and a special kind of terminals – genic terminals – that represent the normal genes. The expression of the normal genes results as usual in different sub-ETs, which in the cellular system are called ADFs (automatically defined functions). As for the tails, they contain only genic terminals, that is, derived features generated on the fly by the algorithm. For example, the chromosome in the figure has three normal genes and one homeotic gene and encodes a main program that invokes three different functions a total of four times, linking them in a particular way. From this example it is clear that the cellular system not only allows the unconstrained evolution of linking functions but also code reuse. And it shouldn't be hard to implement recursion in this system. === Multiple main programs and multicellular systems === Multicellular systems are composed of more than one homeotic gene. Each homeotic gene in this system puts together a different combination of sub-expression trees or ADFs, creating multiple cells or main programs. For example, the program shown in the figure was created using a cellular system with two cells and three normal genes. The applications of these multicellular systems are mu

  Read more →
• 

  AFNLP

  AFNLP (Asian Federation of Natural Language Processing Associations) is the organization for coordinating the natural language processing related activities and events in the Asia-Pacific region. == Foundation == AFNLP was founded on 4 October 2000. == Member Associations == ALTA – Australasian Language Technology Association ANLP Japan Association of Natural Language Processing ROCLING Taiwan ROC Computational Linguistics Society SIG-KLC Korea SIG-Korean Language Computing of Korea Information Science Society == Existing Asian Initiatives == NLPRS: Natural Language Processing Pacific Rim Symposium IRAL: International Workshop on Information Retrieval with Asian Languages PACLING: Pacific Association for Computational Linguistics PACLIC: Pacific Asia Conference on Language, Information and Computation PRICAI: Pacific Rim International Conference on AI ICCPOL: International Conference on Computer Processing of Oriental Languages ROCLING: Research on Computational Linguistics Conference == Conferences == IJCNLP-04: The 1st International Joint Conference on Natural Language Processing in Hainan Island, China IJCNLP-05: The 2nd International Joint Conference on Natural Language Processing in Jeju Island, Korea IJCNLP-08: The 3rd International Joint Conference on Natural Language Processing in Hyderabad, India ACL-IJCNLP-2009: Joint Conference of the 47th Annual Meeting of the Association for Computational Linguistics (ACL) and 4th International Joint Conference on Natural Language Processing (IJCNLP) in Singapore IJNCLP-11: The 5th International Joint Conference on Natural Language Processing in Chiang Mai, Thailand

  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 →
• 

  Midjourney

  Midjourney is a generative artificial intelligence program and service created and hosted by the San Francisco–based "independent research lab" Midjourney, Inc. Midjourney generates images from natural language descriptions, called prompts, similar to OpenAI's DALL-E and Stability AI's Stable Diffusion. It is one of the technologies of the AI boom. The tool was launched into open beta on July 12, 2022. The Midjourney team is led by David Holz, who co-founded Leap Motion. Holz told The Register in August 2022 that the company was already profitable. Users generate images with Midjourney using Discord bot commands or the official website. == History == Midjourney, Inc. was founded in San Francisco, California, by David Holz, previously a co-founder of Leap Motion. The Midjourney image generation platform entered open beta on July 12, 2022. On March 14, 2022, the Midjourney Discord server launched with a request to post high-quality photographs to Twitter and Reddit for systems training. === Model versions === The company has been working on improving its algorithms, releasing new model versions every few months. Version 2 of their algorithm was launched in April 2022, and version 3 on July 25. On November 5, 2022, the alpha iteration of version 4 was released to users. Starting from the 4th version, MJ models were trained on Google TPUs. On March 15, 2023, the alpha iteration of version 5 was released. The 5.1 model is more opinionated than version 5, applying more of its own stylization to images, while the 5.1 RAW model adds improvements while working better with more literal prompts. The version 5.2 included a new "aesthetics system", and the ability to "zoom out" by generating surroundings to an existing image. On December 21, 2023, the alpha iteration of version 6 was released. The model was trained from scratch over a nine month period. Support was added for better text rendition and a more literal interpretation of prompts. == Functionality == Midjourney is accessible through a Discord bot or by accessing their website. Users can use Midjourney through Discord either through their official Discord server, by directly messaging the bot, or by inviting the bot to a third-party server. To generate images, users use the /imagine command and type in a prompt; the bot then returns a set of four images, which users are given the option to upscale. To generate images on the website, users initially needed to have generated at least 1,000 images through the bot; this limitation has since been removed. === Vary (Region) + remix feature === Midjourney released a Vary (Region) feature on September 5, 2023, as part of MidJourney V5.2. This feature allows users to select a specific area of an image and apply variations only to that region while keeping the rest of the image unchanged. === Midjourney web interface === Midjourney introduced its web interface to make its tools more accessible, moving beyond its initial reliance on Discord. This web-based platform was launched in August 2024 alongside the release of Midjourney version 6.1. The web editor consolidates tools such as image editing, panning, zooming, region variation, and inpainting into a single interface. The introduction of the web interface also syncs conversations between Midjourney's Discord channels and web rooms, further enhancing collaboration across both platforms. This shift was in response to growing competition from other AI image generation platforms like Adobe Firefly and Google’s Imagen, which had already launched as native web apps with integration into popular design tools. === Image Weight === This feature lets users control how much influence an uploaded image has on the final output. By adjusting the "image weight" parameter, users can prioritize either the content of the prompt or the characteristics of the image. For instance, setting a higher weight will ensure that the generated result closely follows the image's structure and details, while a lower weight allows the text prompt to have more influence over the final output. === Style Reference === With Style Reference, users can upload an image to use as a stylistic guide for their creation. This tool enables MidJourney to extract the style—whether it is the color palette, texture, or overall atmosphere—from the reference image and apply it to a newly generated image. The feature allows users to fine-tune the aesthetics of their creations by integrating specific artistic styles or moods. === Character Reference === The Character Reference feature allows for a more targeted approach in defining characters. Users can upload an image of a character, and the system uses that image as a reference to generate similar characters in the output. This feature is particularly useful in maintaining consistency in appearance for characters across different images. == Uses == Midjourney's founder, David Holz, told The Register that artists use Midjourney for rapid prototyping of artistic concepts to show to clients before starting work themselves. The advertising industry quickly adopted AI tools such as Midjourney, DALL-E, and Stable Diffusion to create original content and brainstorm ideas. Architects have described using the software to generate mood boards for the early stages of projects, as an alternative to searching Google Images. === Notable usage and controversy === The program was used by the British magazine The Economist to create the front cover for an issue in June 2022. In Italy, the leading newspaper Corriere della Sera published a comic created with Midjourney by writer Vanni Santoni in August 2022. Charlie Warzel used Midjourney to generate two images of Alex Jones for Warzel's newsletter in The Atlantic. The use of an AI-generated cover was criticised by people who felt it was taking jobs from artists. Warzel called his action a mistake in an article about his decision to use generated images. Last Week Tonight with John Oliver included a 10-minute segment on Midjourney in an episode broadcast in August 2022. A Midjourney image called Théâtre D'opéra Spatial won first place in the digital art competition at the 2022 Colorado State Fair. Jason Allen, who wrote the prompt that led Midjourney to generate the image, printed the image onto a canvas and entered it into the competition using the name Jason M. Allen via Midjourney. Other digital artists were upset by the news. Allen was unapologetic, insisting that he followed the competition's rules. The two category judges were unaware that Midjourney used AI to generate images, although they later said that had they known this, they would have awarded Allen the top prize anyway. In December 2022, Midjourney was used to generate the images for an AI-generated children's book that was created over a weekend. Titled Alice and Sparkle, the book features a young girl who builds a robot that becomes self-aware. The creator, Ammaar Reeshi, used Midjourney to generate a large number of images, from which he chose 13 for the book. Both the product and process drew criticism. One artist wrote that "the main problem... is that it was trained off of artists' work. It's our creations, our distinct styles that we created, that we did not consent to being used." In 2023, the realism of AI-based text-to-image generators, such as Midjourney, DALL-E, or Stable Diffusion, reached such a high level that it led to a significant wave of viral AI-generated photos. Widespread attention was gained by a Midjourney-generated photo of Pope Francis wearing a white puffer coat, the fictional arrest of Donald Trump, and a hoax of an attack on the Pentagon, as well as the usage in professional creative arts. Research has suggested that the images Midjourney generates can be biased. For example, even neutral prompts in one study returned unequal results on the aspects of gender, skin color, and location. A study by researchers at the nonprofit group Center for Countering Digital Hate found the tool to be easy to use to generate racist and conspiratorial images. In October 2023, Rest of World reported that Midjourney tends to generate images based on national stereotypes. In 2024, a Frontiers journal published a paper which contained gibberish figures generated with Midjourney, one of which was a diagram of a rat with large testicles and a large penis towering over himself. The paper was retracted a day after the images went viral on Twitter. ==== Content moderation and censorship in Midjourney ==== Prior to May 2023, Midjourney implemented a moderation mechanism predicated on a banned word system. This method prohibited the use of language associated with explicit content, such as sexual or pornographic themes, as well as extreme violence. Moreover, the system also banned certain individual words, including those of religious and political figures, such as Allah or General Secretary of the Chinese Communist Party Xi Jinping. This practice occasionally stirred controversy due to perceiv

  Read more →
• 

  Diagnosis (artificial intelligence)

  As a subfield in artificial intelligence, diagnosis is concerned with the development of algorithms and techniques that are able to determine whether the behaviour of a system is correct. If the system is not functioning correctly, the algorithm should be able to determine, as accurately as possible, which part of the system is failing, and which kind of fault it is facing. The computation is based on observations, which provide information on the current behaviour. The expression diagnosis also refers to the answer of the question of whether the system is malfunctioning or not, and to the process of computing the answer. This word comes from the medical context where a diagnosis is the process of identifying a disease by its symptoms. == Example == An example of diagnosis is the process of a garage mechanic with an automobile. The mechanic will first try to detect any abnormal behavior based on the observations on the car and his knowledge of this type of vehicle. If he finds out that the behavior is abnormal, the mechanic will try to refine his diagnosis by using new observations and possibly testing the system, until he discovers the faulty component; the mechanic plays an important role in the vehicle diagnosis. == Expert diagnosis == The expert diagnosis (or diagnosis by expert system) is based on experience with the system. Using this experience, a mapping is built that efficiently associates the observations to the corresponding diagnoses. The experience can be provided: By a human operator. In this case, the human knowledge must be translated into a computer language. By examples of the system behaviour. In this case, the examples must be classified as correct or faulty (and, in the latter case, by the type of fault). Machine learning methods are then used to generalize from the examples. The main drawbacks of these methods are: The difficulty acquiring the expertise. The expertise is typically only available after a long period of use of the system (or similar systems). Thus, these methods are unsuitable for safety- or mission-critical systems (such as a nuclear power plant, or a robot operating in space). Moreover, the acquired expert knowledge can never be guaranteed to be complete. In case a previously unseen behaviour occurs, leading to an unexpected observation, it is impossible to give a diagnosis. The complexity of the learning. The off-line process of building an expert system can require a large amount of time and computer memory. The size of the final expert system. As the expert system aims to map any observation to a diagnosis, it will in some cases require a huge amount of storage space. The lack of robustness. If even a small modification is made on the system, the process of constructing the expert system must be repeated. A slightly different approach is to build an expert system from a model of the system rather than directly from an expertise. An example is the computation of a diagnoser for the diagnosis of discrete event systems. This approach can be seen as model-based, but it benefits from some advantages and suffers some drawbacks of the expert system approach. == Model-based diagnosis == Model-based diagnosis is an example of abductive reasoning using a model of the system. In general, it works as follows: We have a model that describes the behaviour of the system (or artefact). The model is an abstraction of the behaviour of the system and can be incomplete. In particular, the faulty behaviour is generally little-known, and the faulty model may thus not be represented. Given observations of the system, the diagnosis system simulates the system using the model, and compares the observations actually made to the observations predicted by the simulation. The modelling can be simplified by the following rules (where A b {\displaystyle Ab\,} is the Abnormal predicate): ¬ A b ( S ) ⇒ I n t 1 ∧ O b s 1 {\displaystyle \neg Ab(S)\Rightarrow Int1\wedge Obs1} A b ( S ) ⇒ I n t 2 ∧ O b s 2 {\displaystyle Ab(S)\Rightarrow Int2\wedge Obs2} (fault model) The semantics of these formulae is the following: if the behaviour of the system is not abnormal (i.e. if it is normal), then the internal (unobservable) behaviour will be I n t 1 {\displaystyle Int1\,} and the observable behaviour O b s 1 {\displaystyle Obs1\,} . Otherwise, the internal behaviour will be I n t 2 {\displaystyle Int2\,} and the observable behaviour O b s 2 {\displaystyle Obs2\,} . Given the observations O b s {\displaystyle Obs\,} , the problem is to determine whether the system behaviour is normal or not ( ¬ A b ( S ) {\displaystyle \neg Ab(S)\,} or A b ( S ) {\displaystyle Ab(S)\,} ). This is an example of abductive reasoning. == Diagnosability == A system is said to be diagnosable if whatever the behavior of the system, we will be able to determine without ambiguity a unique diagnosis. The problem of diagnosability is very important when designing a system because on one hand one may want to reduce the number of sensors to reduce the cost, and on the other hand one may want to increase the number of sensors to increase the probability of detecting a faulty behavior. Several algorithms for dealing with these problems exist. One class of algorithms answers the question whether a system is diagnosable; another class looks for sets of sensors that make the system diagnosable, and optionally comply to criteria such as cost optimization. The diagnosability of a system is generally computed from the model of the system. In applications using model-based diagnosis, such a model is already present and doesn't need to be built from scratch.

  Read more →
• 

  JSGF

  JSGF stands for Java Speech Grammar Format or the JSpeech Grammar Format (in a W3C Note). Developed by Sun Microsystems, it is a textual representation of grammars for use in speech recognition for technologies like XHTML+Voice. JSGF adopts the style and conventions of the Java programming language in addition to use of traditional grammar notations. The Speech Recognition Grammar Specification was derived from this specification. == Example == The following JSGF grammar will recognize the words coffee, tea, and milk.

  Read more →
• 

  Ideogram (text-to-image model)

  Ideogram is a freemium text-to-image model developed by Ideogram, Inc. using deep learning methodologies to generate digital images from natural language descriptions known as prompts. The model is capable of generating legible text in the images compared to other text-to-image models. == History == Ideogram was founded in 2022 by Mohammad Norouzi, William Chan, Chitwan Saharia, and Jonathan Ho to develop a better text-to-image model. It was first released with its 0.1 model on August 22, 2023, after receiving $16.5 million in seed funding, which itself was led by Andreessen Horowitz and Index Ventures. In February 2024, Ideogram raised $80 million after its 1.0 model release in the same year. In August 2024, Ideogram released its 2.0 model. This model has several styles such as realistic, design, 3D, and anime and better capability in generating text. In February 2025, Ideogram released 2a model. This model was designed for speed and optimized for graphics design and photography generation. In March 2025, Ideogram released its 3.0 model. This model has improved realism and understanding of complex text layout, although like other generative AI models, it still struggles with ambigram creation.

  Read more →
• 

  Constructive cooperative coevolution

  The constructive cooperative coevolutionary algorithm (also called C3) is a global optimisation algorithm in artificial intelligence based on the multi-start architecture of the greedy randomized adaptive search procedure (GRASP). It incorporates the existing cooperative coevolutionary algorithm (CC). The considered problem is decomposed into subproblems. These subproblems are optimised separately while exchanging information in order to solve the complete problem. An optimisation algorithm, usually but not necessarily an evolutionary algorithm, is embedded in C3 for optimising those subproblems. The nature of the embedded optimisation algorithm determines whether C3's behaviour is deterministic or stochastic. The C3 optimisation algorithm was originally designed for simulation-based optimisation but it can be used for global optimisation problems in general. Its strength over other optimisation algorithms, specifically cooperative coevolution, is that it is better able to handle non-separable optimisation problems. An improved version was proposed later, called the Improved Constructive Cooperative Coevolutionary Differential Evolution (C3iDE), which removes several limitations with the previous version. A novel element of C3iDE is the advanced initialisation of the subpopulations. C3iDE initially optimises the subpopulations in a partially co-adaptive fashion. During the initial optimisation of a subpopulation, only a subset of the other subcomponents is considered for the co-adaptation. This subset increases stepwise until all subcomponents are considered. This makes C3iDE very effective on large-scale global optimisation problems (up to 1000 dimensions) compared to cooperative coevolutionary algorithm (CC) and Differential evolution. The improved algorithm has then been adapted for multi-objective optimization. == Algorithm == As shown in the pseudo code below, an iteration of C3 exists of two phases. In Phase I, the constructive phase, a feasible solution for the entire problem is constructed in a stepwise manner. Considering a different subproblem in each step. After the final step, all subproblems are considered and a solution for the complete problem has been constructed. This constructed solution is then used as the initial solution in Phase II, the local improvement phase. The CC algorithm is employed to further optimise the constructed solution. A cycle of Phase II includes optimising the subproblems separately while keeping the parameters of the other subproblems fixed to a central blackboard solution. When this is done for each subproblem, the found solution are combined during a "collaboration" step, and the best one among the produced combinations becomes the blackboard solution for the next cycle. In the next cycle, the same is repeated. Phase II, and thereby the current iteration, are terminated when the search of the CC algorithm stagnates and no significantly better solutions are being found. Then, the next iteration is started. At the start of the next iteration, a new feasible solution is constructed, utilising solutions that were found during the Phase I of the previous iteration(s). This constructed solution is then used as the initial solution in Phase II in the same way as in the first iteration. This is repeated until one of the termination criteria for the optimisation is reached, e.g. a maximum number of evaluations. {Sphase1} ← ∅ while termination criteria not satisfied do if {Sphase1} = ∅ then {Sphase1} ← SubOpt(∅, 1) end if while pphase1 not completely constructed do pphase1 ← GetBest({Sphase1}) {Sphase1} ← SubOpt(pphase1, inext subproblem) end while pphase2 ← GetBest({Sphase1}) while not stagnate do {Sphase2} ← ∅ for each subproblem i do {Sphase2} ← SubOpt(pphase2,i) end for {Sphase2} ← Collab({Sphase2}) pphase2 ← GetBest({Sphase2}) end while end while == Multi-objective optimisation == The multi-objective version of the C3 algorithm is a Pareto-based algorithm which uses the same divide-and-conquer strategy as the single-objective C3 optimisation algorithm . The algorithm again starts with the advanced constructive initial optimisations of the subpopulations, considering an increasing subset of subproblems. The subset increases until the entire set of all subproblems is included. During these initial optimisations, the subpopulation of the latest included subproblem is evolved by a multi-objective evolutionary algorithm. For the fitness calculations of the members of the subpopulation, they are combined with a collaborator solution from each of the previously optimised subpopulations. Once all subproblems' subpopulations have been initially optimised, the multi-objective C3 optimisation algorithm continues to optimise each subproblem in a round-robin fashion, but now collaborator solutions from all other subproblems' subspopulations are combined with the member of the subpopulation that is being evaluated. The collaborator solution is selected randomly from the solutions that make up the Pareto-optimal front of the subpopulation. The fitness assignment to the collaborator solutions is done in an optimistic fashion (i.e. an "old" fitness value is replaced when the new one is better). == Applications == The constructive cooperative coevolution algorithm has been applied to different types of problems, e.g. a set of standard benchmark functions, optimisation of sheet metal press lines and interacting production stations. The C3 algorithm has been embedded with, amongst others, the differential evolution algorithm and the particle swarm optimiser for the subproblem optimisations.

  Read more →
• 

  Grokipedia

  Grokipedia is an AI-generated online encyclopedia operated by the American company xAI. The site was launched on October 27, 2025. Some entries are generated by Grok, a large language model owned by the same company, while others were forked from Wikipedia, with some altered and some used nearly verbatim. Articles cannot be directly edited, though logged-in visitors to the encyclopedia can suggest new articles or corrections via a pop-up form, which are reviewed by Grok. The xAI founder Elon Musk suggested Grokipedia could be an alternative to Wikipedia that would "purge out the propaganda" he believes is promoted by the latter, describing Wikipedia as "woke" and an "extension of legacy media propaganda". External analysis of Grokipedia's content has focused on its accuracy and biases due to hallucinations and potential algorithmic bias, which reviewers have described as promoting right-wing perspectives and Musk's views. The majority of coverage has described the website as validating, promoting, and legitimizing a variety of debunked conspiracy theories and ideas against scientific consensus on topics such as HIV/AIDS denialism, vaccines and autism, climate change, and race and intelligence. The site has been accused of whitewashing far-right extremism, such as by falsely claiming a white genocide is actively occurring. Several right-wing figures have welcomed the site. Studies have highlighted its use of sources deemed as having very low credibility such as X conversations and neo-Nazi websites, and for writing about far-right figures and topics in a promotional manner. == Background == Wikipedia is an online encyclopedia written and maintained by a community of volunteers. Its possible bias has been studied and debated. In 2018, Haaretz noted "Wikipedia has succeeded in being accused of being both too liberal and too conservative, and has critics from across the spectrum". xAI is an American AI company founded by Elon Musk in 2023. Its flagship product is the family of large language models called Grok. == History == In 2021, Musk expressed affection for Wikipedia on its 20th anniversary. In 2022, however, Musk argued that Wikipedia was "losing its objectivity", and in 2023, said he would donate US$1 billion to the project if it was pejoratively renamed "Dickipedia". In December 2024, Musk called for a boycott of donations to Wikipedia over its perceived left-wing bias, calling it "Wokepedia". In January 2025, Musk made a series of statements on Twitter denouncing Wikipedia for its description of the incident where he made a controversial gesture, which many viewed as resembling a Nazi salute, at president Donald Trump's second inauguration. Musk has since positioned Grokipedia as an alternative to Wikipedia that would "purge out the propaganda" in the latter, with Musk describing Wikipedia as "woke" and an "extension of legacy media propaganda". === Idea and announcement === In September 2025, Musk spoke at the All-In podcast conference with David O. Sacks, the White House advisor on AI and cryptocurrency, about how Grok consumed data from Wikipedia and other sources to gain more complete knowledge of the world. Sacks suggested publishing its knowledge base as an artifact called "Grokipedia", saying "Wikipedia is so biased, it's a constant war". Following the conversation, Musk announced that xAI was building a new AI-generated online encyclopedia called Grokipedia. According to Musk's announcement, it would be an AI-powered knowledge base designed to rival Wikipedia by addressing its perceived biases, errors, and ideological slants. The project positioned itself within a history of ideologically driven alternatives to Wikipedia, such as the conservative Conservapedia (launched in 2006) and the Russian-government-friendly Ruwiki (launched in 2023). However, Grokipedia is distinct in its core reliance on artificial intelligence rather than human community editing. === Launch and traffic === On October 6, 2025, Musk announced that the early version of Grokipedia was scheduled for release in two weeks, but the project was postponed briefly to address content quality issues. It launched on October 27, 2025, labeled "v 0.1", with over 800,000 articles, compared to over seven million English Wikipedia articles as of September 1, 2025. According to an initial analysis of usage figures by Similarweb, which evaluates data from registered users and partners, Grokipedia recorded a peak of over 460,000 website visits in the US on October 28, 2025. After that, traffic dropped significantly and settled at around 35,000 visits per day between November 8 and 11, 2025. As of early 2026, it had over 5.6 million articles. In January 2026, The Guardian reported that GPT-5.2 frequently cited Grokipedia as a source in responses, raising concerns of misinformation on ChatGPT. The same month, The Verge reported that Google's AI Overviews, AI Mode, and Gemini language model, as well as Microsoft Copilot and Perplexity AI, used Grokipedia to answer niche, obscure, or highly specific factual questions or "non-sensitive queries." According to a case study published by SEO Engico, the site received only 19 clicks from Google Search in November 2025 but reached approximately 3.2 million monthly clicks by January 2026, with over 900,000 pages indexed and millions of ranking keywords. Analysts attributed the surge in part to the site's technical structure and large-scale AI-generated content production. In early February 2026, Grokipedia's visibility in Google Search declined sharply. SEO analysts, including Glenn Gabe and Malte Landwehr, reported a significant drop in rankings across Google organic results as well as in Google AI Overviews and AI Mode. The same case study cited independent reviews that identified citation quality concerns, including references to low-credibility sources and instances of self-citation. By mid-February 2026, Grokipedia had reportedly lost much of its previous search visibility, and Wikipedia ranked above it for searches related to its own name. === Updates === ==== Future ==== In November 2025, Musk announced that he eventually plans to change the name of the site to Encyclopedia Galactica when Grokipedia is "good enough", saying that it had a "long way to go". This name is taken from the publication of that title in the works of Isaac Asimov and Douglas Adams. Musk said that he hoped to send copies of the encyclopedia to "the Moon and Mars and out to deep space". == Content == The Grok large language model generates and fact-checks articles on Grokipedia. Users cannot directly edit Grokipedia articles, but logged-in users can suggest edits and report errors, with such submissions being reviewed and implemented by the Grok AI. Some articles are nearly identical to their Wikipedia entries, but the format of Grokipedia citations is different, and some Grokipedia articles were republished almost verbatim, accompanied by a disclaimer noting that the content was "adapted from Wikipedia" under a Creative Commons license. Others were completely rewritten from scratch using Musk's AI chatbot, Grok. Forbes identified the articles AMD, Lamborghini, and PlayStation 5 as examples of copied Wikipedia articles. Articles attributed to Wikipedia carry a Creative Commons Attribution-ShareAlike license, while the license of other articles is licensed under the "X Community License", a license that accepts reuse and remixing for "non-commercial and research purposes" and commercial use that abides to "all of the guardrails provided in xAI's Acceptable Use Policy". On October 31, 2025, Musk clarified that the duplication of Wikipedia articles was intentional, saying that the Grokipedia team instructed Grok to compile Wikipedia's top 1 million articles and make content changes to them. The site's design has been described as minimalist with a simple homepage including little more than a large search bar. In a comparative textual analysis of the most heavily edited matched article pairs from Grokipedia and Wikipedia, Grokipedia entries are substantially longer and less densely referenced, indicating that AI-produced encyclopedias prioritize exposition rather than source-based validation. Starting in version 0.2, Grok reviews and implements approved suggested edits, and a small panel rotates through a display of the names of several recently edited articles. In February 2026, the Columbia Journalism Review reported on an analysis by the Tow Center for Digital Journalism finding that Grok, the AI behind Grokipedia, had increasingly begun suggesting and approving edits to the site itself without human involvement. According to the report, AI-generated edit suggestions overtook human submissions in December 2025 and accounted for more than three-quarters of proposed changes. The analysis raised concerns about transparency, editorial oversight, and fact-checking standards, particularly after instances in which Grok proposed or modified politically s

  Read more →
• 

  Association for Computational Linguistics

  The Association for Computational Linguistics (ACL) is a scientific and professional organization for people working on natural language processing. Its namesake conference is one of the primary high impact conferences for natural language processing research, along with EMNLP. The conference is held each summer in locations where significant computational linguistics research is carried out. It was founded in 1962, originally named the Association for Machine Translation and Computational Linguistics (AMTCL). It became the ACL in 1968. The ACL has a European (EACL), a North American (NAACL), and an Asian (AACL) chapter. == History == The ACL was founded in 1962 as the Association for Machine Translation and Computational Linguistics (AMTCL). The initial membership was about 100. In 1965, the AMTCL took over the journal Mechanical Translation and Computational Linguistics. This journal was succeeded by many other journals: the American Journal of Computational Linguistics (1974–1978, 1980–1983), and then Computational Linguistics (1984–present). Since 1988, the journal has been published for the ACL by MIT Press. The annual meeting was first held in 1963 in conjunction with the Association for Computing Machinery National Conference. The annual meeting was, for a long time, relatively informal and did not publish anything longer than abstracts. By 1968, the society took on its current name, the Association for Computational Linguistics (ACL). The publication of the annual meeting's Proceedings of the ACL began in 1979 and gradually matured into its modern form. Many of the meetings were held in conjunction with the Linguistic Society of America, and a few with the American Society for Information Science and the Cognitive Science Society. The United States government sponsored much research from 1989 to 1994, characterized by an increase in author retention rates and an increase in research in some key topics, such as speech recognition, in ACL. By the 21st century, it was able to maintain authors at a high rate who coalesced in a more stable arrangement around individual research topics. In 1991, the group published a prototype for a text generator based on the universal grammar theory of Noam Chomsky. The system, nicknamed Parrot, relied on a finite set of syntactic transformations and a hand-curated lexicon. Despite some initial success, including experimentation with morpheme syntactics, funding halted after the research team encountered intractable difficulties with inflection and abstract locutions. == Annual Meeting of the ACL == Every year, the ACL holds the Annual Meeting of the ACL. The location lies in Europe in years zero modulo three, North America in years one modulo three, and Asia–Australia in years two modulo three. In 2020, the Annual Meeting received for the first time more submissions from China than the United States. == Activities == The ACL organizes several of the top conferences and workshops in the field of computational linguistics and natural language processing. These include: Annual Meeting of the Association for Computational Linguistics (ACL), the flagship conference of the organization Empirical Methods in Natural Language Processing (EMNLP) International Joint Conference on Natural Language Processing (IJCNLP), held jointly one of the other conferences on a rotating basis Conference on Computational Natural Language Learning (CoNLL) Lexical and Computational Semantics and Semantic Evaluation (SemEval) Joint Conference on Lexical and Computational Semantics (SEM) Workshop on Statistical Machine Translation (WMT) Besides conferences, the ACL also sponsors the journals Computational Linguistics and Transactions of the Association for Computational Linguistics (TACL). Papers and other presentations at ACL and ACL-affiliated venues are archived online in the open-access ACL Anthology. == Special Interest Groups == ACL has a large number of Special Interest Groups (SIGs), focusing on specific areas of natural language processing. Some current SIGs within ACL are: == Presidents == Each year, the ACL elects a distinguished computational linguist who becomes vice-president of the organization in the next calendar year and president one year later. Recent ACL presidents are:

  Read more →
• 

  DARPA Prize Competitions

  Over the years, the U.S. Defense Advanced Research Projects Agency (DARPA) has conducted numerous prize competitions to spur innovation. A prize competition allows DARPA to establish an ambitious goal, opening the door to novel approaches from the public that might otherwise appear too risky for experts in a particular field to pursue. == Statutory authorities == In 1999, Congress provided prize competition authority to DARPA in the National Defense Authorization Act for Fiscal Year 2000 (P.L. 106–65), 10 U.S.C. § 4025, formerly 10 U.S.C. §2374a. DARPA also conducts prize competitions under the America COMPETES Act, 15 U.S.C. § 3719. == Recent prize competitions == DARPA Grand Challenge (2004 and 2005) was a prize competition to spur the development of autonomous vehicle technologies. The $1 million prize went unclaimed as no vehicles could complete the challenging desert route from Barstow, CA, to Primm, NV, on March 13, 2004. A year later, on October 8, 2005, the Stanford Racing Team won the $2 million prize during the second competition of the Grand Challenge in the desert Southwest near the California/Nevada state line. DARPA Urban Challenge (2007) required the competitors to build an autonomous vehicle capable of driving in traffic and performing complex maneuvers such as merging, passing, parking, and negotiating intersections. On November 3, 2007, the Carnegie Mellon Team won the $2 million prize, and its vehicle became the first autonomous vehicle that interacted with both manned and unmanned vehicle traffic in an urban environment. DARPA Network Challenge (Red Balloon Challenge) (2009) explored the roles that the Internet and social networking play in solving broad-scope, time-critical problems. On December 5, 2009, the Massachusetts Institute of Technology team won $40,000 by locating the ten moored, eight-foot, red weather balloons at ten places in the United States within seven hours. DARPA Digital Manufacturing Analysis, Correlation and Estimation Challenge (DMACE) (2010) was a three-month contest to showcase the potential of digital manufacturing of advanced materials. The University of California at Santa Barbara team won a $50,000 prize for crushing 180 digitally manufactured (DM) titanium mesh spheres with the most accurate predictive model of the components’ properties. DARPA Shredder Challenge (2011) was to identify and assess potential capabilities and vulnerabilities to sensitive information in the national security community. Participating teams must download the images of the documents shredded into more than 10,000 pieces from the Challenge website, reconstruct the documents, and solve the five puzzles. Of almost 9,000 teams, the San Francisco-based All Your Shreds Are Belong to U.S team won the $50,000 prize. DARPA UAVForge Challenge (2011-2012) aimed to build and test a user-intuitive, backpack-portable unmanned aerial vehicle (UAV) that could quietly fly in and out of critical environments to conduct sustained surveillance for up to three hours. The $100,000 prize was not claimed because none of the 140 teams met the technical matrix. DARPA Cash for Locating & Identifying Quick Response Codes (CLIQR) Quest Challenge (2012) explored the role the Internet and social media played in the timely communication, wide-area team-building, and urgent mobilization required to solve broad scope, time-critical problems. The challenge offered $40,000 to the first individual or team that could locate seven posters appearing in U.S. cities bearing the DARPA logo and a quick response code (QR) within 15 days. No team found and submitted all seven codes. DARPA Fast Adaptable Next-Generation Ground Vehicle (FANG) Challenge (2012-2013) was to use three competitions for the design of an infantry fighting vehicle, culminating in prototypes. In April 2013, DARPA awarded US$1 million to a three-man team during the first competition. DARPA decided not to proceed with the second and third competitions as originally planned and transitioned the technologies to the defense and commercial industry through the Digital Manufacturing and Design Innovation Institute (DMDII). DARPA Spectrum Challenge (2013-2014) sought to demonstrate how a software-defined radio can use a given communication channel in the presence of other users and interfering signals. Three teams emerged as the overall winners, winning a total of $150,000 in prizes. DARPA Chikungunya (CHIKV) Challenge (2014-2015) was a health-related effort to develop the most accurate predictions of CHIKV cases for all Western Hemisphere countries and territories between September 2014 and March 2015. On May 12, 2015, DARPA awarded $500,000 in prizes to the 11 winners of the competition during a scientific review DARPA Robotics Challenge (DRC) (2013-2015) aimed to develop semi-autonomous ground robots that could do "complex tasks in dangerous, degraded, human-engineered environments." A South Korean team won the first prize of $2 million, and two U.S. teams won $1 million and $500,000 as second and third winners. DARPA Cyber Grand Challenge (CGC) (2014 - 2016) was to “create automatic defensive systems capable of reasoning about flaws, formulating patches and deploying them on a network in real time.” The top three winners were awarded prizes of $2 million, $1 million, and $750,000, respectively. DARPA Spectrum Collaboration Challenge (SC2) (2016-2019) aimed to encourage the development of AI-enabled wireless networks to “ensure that the exponentially growing number of military and civilian wireless devices would have full access to the increasingly crowded electromagnetic spectrum.” A team from the University of Florida won the overall top prize of US$2 million at the final SC2 competition. DARPA Subterranean (SubT) Challenge (2017-2021) was to develop robotic technologies to map, navigate, search and exploit complex underground environments. The first-place winners of the system final competition and of the virtual final competition were awarded $2 million and $750,000, respectively, with multiple prizes awarded to the second and third-place winners. DARPA Launch Challenge (2018-2020) was a $12 million satellite launch challenge to demonstrate responsive and flexible space launch capabilities from the small launch providers and was to culminate in two separate launch competitions where the competitors must launch a satellite to low Earth orbit (LEO) within days of each other at different locations in the United States. The competition ended without a winner. DARPA Forecasting Floats in Turbulence (FFT) Challenge (2021) was to spur technologies that could predict the location of sea drifters or floats within 10 days. DARPA awarded $25,000 for first place, with prizes of $15,000 and $10,000 for second place and third place. DARPA Artificial Intelligence Cyber Challenge (AIxCC) (2023–2025) was a two-year challenge and asks competitors to design novel AI systems to secure critical software code on which Americans rely. The total prize money is $29.5 million. In March 2024, the Advanced Research Projects Agency for Health (ARPA-H) partnered with DARPA, contributing an additional $20 million to the competition's prize pool to address software vulnerabilities in medical devices, hospital IT, and biotech equipment. AIxCC collaborates with Google, Microsoft, OpenAI, Anthropic, Linux Foundation, Open Source Security Foundation, Black Hat USA, and DEF CON, all of which provide AIxCC with access to large language models. In August 2024, AIxCC held the semifinal at DEF CON in Las Vegas. DARPA and ARPA-H tested all 42 submissions by running them through various open-source coding projects with deliberately injected vulnerabilities and scored the tools based on their effectiveness in identifying and fixing security flaws. Seven teams, each winning $2 million in the semifinals, competed in the final round of the AIxCC at the August 2025 DEF CON conference. Team Atlanta won first place with a $4 million prize for its cyber reasoning systems, which identified and patched vulnerabilities across 54 million lines of code. DARPA Triage Challenge (2023 – 2026) aims to spur the development of novel physiological features for medical triage, with a total prize money of $7 million. In October 2024, Challenge Event 1 was held in Perry, Georgia, featuring to-scale replicas of disaster sites such as an airplane crash and Hurricane Katrina, and teams competed based on how closely their data aligned with the agency’s official data and how quickly and accurately their autonomous systems could identify individuals most urgently in need of medical care. DARPA concluded the second year of competitions and, in November 2025, named the top performers in systems and data categories, which will advance to the final 2026 competition. The DARPA Lift Challenge (2025-2026) is for participants to design unmanned aerial systems capable of carrying up to four times their own weight, with a minimum payload of 110 pounds. Acco

  Read more →
• 

  Minne Atairu

  Minne Atairu is a Nigerian interdisciplinary artist, a recipient of the 2021 Global South Award Lumen Prize for Art and Technology. She generates synthetic Benin Bronzes through recombination of historical fragments, sculptures, texts, images, and sounds. == Early life and education == Atairu was born in Benin, Nigeria. She holds a bachelor's degree in art history from the University of Maiduguri in Maiduguri, Nigeria; a master's degree in museum studies from the George Washington University in Washington, D.C.; and a doctorate in art education from Teachers College, Columbia University in New York City. Her academic research integrates artificial intelligence, art/museum education and hip-hop based education. == Works == Atairu's artmaking involves using artificial intelligence (AI; such as StyleGAN, GPT-3) to make artwork. She uses tools such as Midjourney and Blender software to develop her works. === Mami Wata === Her first work is a Yoruba goddess called Mami Wata where she used Midjourney in generating the images. === To the Hand === For her 2023 installation To the Hand at The Shed arts center, she worked with Blender to convert text into 3D-printed sculptures made of corn starch or sugarcane infused with bronze. The rings of ground terra-cotta that surround the sculpture represent the walls and deep moats of Benin. == Publications == Atairu, Minne (February 1, 2024). "Reimagining Benin Bronzes using generative adversarial networks". AI & Society. 39 (1): 91–102. doi:10.1007/s00146-023-01761-7. ISSN 1435-5655.

  Read more →