AI Chat Image

AI Chat Image — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  PCPaint

  PCPaint was one of the first IBM PC-based mouse-driven GUI paint programs, released in 1984. It followed after Microsoft Doodle, released in 1983 with the Microsoft Mouse version 1 drivers for DOS, and around the same time as Digital Research’s Draw program. It was developed and created by John Bridges and Doug Wolfgram. It was later developed into Pictor Paint. The hardware manufacturer Mouse Systems bundled PCPaint with millions of computer mice that they sold, making PCPaint one of the best-selling DOS-based paint programs of the mid 1980s. == History == In 1983, Doug Wolfgram bought a Microsoft Mouse and decided to write a drawing program for it. They named it “Mouse Draw”. The interface was primitive but the program functioned well. Wolfgram traveled to SoftCon in New Orleans where he demonstrated the program to Mouse Systems. Mouse Systems was developing an optical mouse and they wanted to bundle a painting program so they agreed to publish Mouse Draw. The original program was written entirely in assembly language with primitive graphics routines developed by Wolfgram. John Bridges worked for an educational software company, Classroom Consortia Media, Inc., developing and writing Apple and IBM graphics libraries for CCM's software. Bridges and Wolfgram were friends who had been connected through a bulletin board system developed and run by Wolfgram. The two collaborated cross country via the BBS, Wolfram in California and Bridges in New York. Mouse Systems wanted the paint program to capture the look and feel of MacPaint. John Bridges and Doug Wolfgram started reworking Mouse Draw into what became PCPaint. The program was completely re-written using Bridge's graphics library and the top-level elements were written in C rather than assembly language. Bridges developed the core graphics code for the first version of PCPaint while Wolfgram worked on the user interface and top-level code. Mouse Systems signed an exclusive agreement with Wolfgram's company, Microtex Industries, Inc., to bundle PCPaint with every mouse they sold. They began publishing PCPaint with their mice in 1984. Microsoft responded in 1985 by bundling a competing product, PC Paintbrush, with version 4 of its DOS drivers for the Microsoft Mouse, replacing its in-house Microsoft Doodle program which it published with version 1 of the DOS drivers in mid-1983. Microsoft’s mouse began to outsell Mouse Systems mouse. In November 1985 Microsoft bundled a cut-down version of PC Paintbrush with Windows 1.0 (called Microsoft Paint), later bundling an updated version of PC Paintbrush with Windows 3.0 (as Paintbrush), impacting PCPaint’s marketshare. In early 1987, Mouse Systems decided that PCPaint wasn't helping to sell mice any longer so they discontinued the bundle deal and returned rights to the code to MicroTex Industries, but retained rights to the name, PCPaint. Wolfgram then combined the paint program with a new animation system he was developing (called GRASP) and Paul Mace Software bought publishing rights to the animation system and PCPaint, which was to be renamed Pictor. Bridges again got involved and took over programming responsibilities for GRASP as well as PCPaint while Wolfgram focused on more of the business details. In creating the first version of PCPaint, Doug had a dual-floppy machine with a Computer Innovations compiler on one disk and source code on the other. John had the "luxury" of a 10MB hard disk in his XT. Data was exchanged daily via 1200, then 2400 baud modems. === Authorship and Ownership === John Bridges and Wolfgram continued to work on PCPaint and GRASP on behalf of Paul Mace Software until 1990. Also in that year, Doug Wolfgram sold his remaining rights to PCPaint (and its animation system, GRASP) to John Bridges. In 1994, GRASP development stopped and so did development of Pictor Paint. John Bridges terminated his GRASP publishing contract with Paul Mace Software, and went off to create GLPro (the next generation of GRASP) with GMEDIA. Along with GLPro, came GLPaint, the successor to PCPaint and Pictor Paint. == Versions == In June 1984, Mouse Systems shipped PCPaint 1.0, the first GUI based Paint program for the IBM PC family of computers. John Bridges and Doug Wolfgram, were the co-authors of PCPaint 1.0. PCPaint 1.0 saved its graphics in a modified BSaved image format with the extension of ".PIC". The release of PCPaint Version 1.5 followed in late 1984, with the additions of graphics image compression for the .PIC format and support for "larger-than-screen" images. PCjr support was also added in this version after overcoming severe memory shortage problems getting PCPaint to run on the 128k PCjr. October 1985 saw the release of PCPaint 2.0. EGA support and publishing features were added to this version. The .PIC format was further refined, offering support for the rapidly expanding graphics capabilities of the PC and efficient image compression. PCPaint 3.1 was released in 1989. Unlike previous versions, it was not bundled with mice but was sold as a stand-alone software product. PCPaint 3.1 offered improved text and image handling, provided 36 types of flood and fill, worked with VGA adapters in hi-res 16-color and 256-color modes, allowed the user to save and retrieve files in a variety of intercompatible formats (.PIC, .GIF, .PCX, .IMG), and printed selected portions of images on color or black-and-white dot matrix, ink jet, and laser printers such as PostScript and HP Laser Jet. PCPaint 3.1 is still in use today by some users of DOS emulation programs like DOSBox and available for free download. Pictor Paint was an improved version, written by John Bridges, and bundled with GRASP GRaphical System for Presentation also written by John Bridges. It was also called "The Painter's Easel". GLPaint, released in 1995, was the last in this series of paint programs written by John Bridges. By 1998 version 7.0 provided support for TrueColor images and the Pictor PIC format was expanded to handle these. == Pictor PIC Image Format == PCPaint 1.0 saved its graphics in a modified BSAVE image format (which was popular at the time) with the file type (extension) of ".PIC". By PCPaint 1.5 this format was extended further to accommodate image compression. With the release of version 2.0 the PICtor PIC image format was developed almost to its present state, with no similarity to the BSAVE format used by earlier versions. Pictor Paint saved its files in a compressed format with the file extension PIC, which was the same format used by PCPaint.

  Read more →
• 

  VACUUM

  VACUUM is a set of normative guidance principles for achieving training and test dataset quality for structured datasets in data science and machine learning. The garbage-in, garbage out principle motivates a solution to the problem of data quality but does not offer a specific solution. Unlike the majority of the ad-hoc data quality assessment metrics often used by practitioners VACUUM specifies qualitative principles for data quality management and serves as a basis for defining more detailed quantitative metrics of data quality. VACUUM is an acronym that stands for: valid accurate consistent uniform unified model

  Read more →
• 

  Ultra Hal

  Ultra Hal is a chatbot intended to function as a virtual assistant. It was developed by Zabaware, Inc. Ultra Hal uses a natural language interface with animated characters using speech synthesis. Users can communicate with the chatterbot via typing or via a speech recognition engine. It utilizes the WordNet lexical dictionary. Its name is an allusion to HAL 9000, the artificial intelligence from the movie 2001: A Space Odyssey. Ultra Hal won the 2007 Loebner Prize for "most human" chatterbot.

  Read more →
• 

  Speculative decoding

  Speculative decoding is an inference-time optimization for autoregressive large language models (LLMs) that generates multiple tokens per decoding step instead of one. A smaller draft model proposes a sequence of candidate tokens, and the larger target model verifies them in a single forward pass through a modified rejection sampling scheme. The verification preserves the target model's original output distribution, so the technique produces the same results as standard decoding while cutting latency by roughly two to three times. The name is an analogy to speculative execution in CPU design, where a processor runs instructions along a predicted branch before the outcome is known. == Background == Standard autoregressive decoding in large language models generates one token at a time. The model computes a probability distribution over its vocabulary, samples the next token, and feeds that token back as input. For large models, this process is bottlenecked by memory bandwidth rather than arithmetic throughput: loading the model's parameters from high-bandwidth memory (HBM) to the processor takes up most of the wall-clock time at each step. Because of this, a forward pass over one token and a forward pass over several tokens in a batch take roughly the same time. Speculative decoding relies on this property. == Mechanism == The technique alternates between two phases: drafting and verification. During drafting, a fast approximation model generates a short run of K candidate tokens, typically between 3 and 12. The draft model is usually a much smaller version of the target model or a lightweight auxiliary network. During verification, the target model scores the entire draft sequence in one batched forward pass. A modified rejection sampling algorithm compares the draft and target probabilities at each position. If the target model would have been at least as likely to produce a given token, that token is accepted; the first token that fails is resampled from a corrected distribution, and everything after it is thrown out. The result is that the output distribution is the same as if each token had been generated one at a time. How many tokens get accepted per cycle depends on how well the draft model matches the target. For common words and predictable continuations the match tends to be good, so the target model can confirm several tokens at once. == History == An early precursor was blockwise parallel decoding, proposed in 2018 by Stern, Shazeer, and Uszkoreit. Their method predicted multiple future tokens through auxiliary prediction heads and validated them against the autoregressive model, but it only worked with greedy decoding and did not preserve the full sampling distribution. The modern form of the technique came from Yaniv Leviathan, Matan Kalman, and Yossi Matias at Google Research, who posted "Fast Inference from Transformers via Speculative Decoding" on arXiv in November 2022. Separately and at about the same time, Charlie Chen and colleagues at DeepMind arrived at a closely related method they called speculative sampling, published in February 2023. Both papers introduced the use of rejection sampling to guarantee that the output distribution is unchanged. Leviathan et al. showed roughly 2–3x speedup on T5-XXL (11 billion parameters); Chen et al. reported 2–2.5x on the Chinchilla model (70 billion parameters). The Leviathan et al. paper was presented as an oral at the International Conference on Machine Learning in July 2023. == Variants == SpecInfer (Miao et al., 2024) uses multiple small language models to jointly build a tree of candidate continuations rather than a single chain. The target model verifies the whole tree in parallel and keeps the longest valid path, with reported speedups of 1.5–3.5x. Medusa (Cai et al., 2024) takes a different approach by not using a separate draft model at all. Extra lightweight decoding heads are attached to the target model itself, and each one predicts a token at a different future position. The candidates are evaluated through a tree-structured attention mechanism. The authors measured 2.2–3.6x speedup. EAGLE (Li et al., 2024) performs autoregression on the target model's internal feature representations (specifically the second-to-top layer) rather than on tokens directly. On LLaMA 2 Chat 70B, this gave a 2.7–3.5x latency reduction. Later versions added dynamic draft trees (EAGLE-2) and further optimizations (EAGLE-3), reaching 3–6.5x speedup. == Adoption == By 2024, speculative decoding had become a standard part of production LLM serving. Google uses it in the AI Overviews feature of Google Search. Open-source inference frameworks such as vLLM, NVIDIA's TensorRT-LLM, and SGLang all include built-in support for speculative decoding and its variants. Apple, AWS, and Meta have also published research extending the method or deploying it at scale.

  Read more →
• 

  International Clinical Trials Registry Platform

  The International Clinical Trials Registry Platform (ICTRP) is a platform for the registration of clinical trials operated by the World Health Organization. The ICTRP combines data from multiple cooperating clinical trials registries to generate a global view of clinical trials worldwide, with a search portal that allows access to the entire dataset. It requires a minimum standard set of database fields, the WHO Trial Registration Data Set, to be present for a trial to be registered. All entries are given a Universal Trial Number (UTN) that identifies them uniquely. The organization has sought to assist various national governments in establishing their own clinical trials databases. It combines data from the following primary registries and data providers: Australian New Zealand Clinical Trials Registry (ANZCTR) Brazilian Clinical Trials Registry (ReBec) Chinese Clinical Trial Registry (ChiCTR) Clinical Research Information Service (CRiS), Republic of Korea ClinicalTrials.gov Clinical Trials Information System (CTIS), European Medicines Agency Clinical Trials Registry - India (CTRI) Cuban Public Registry of Clinical Trials (RPCEC) EU Clinical Trials Register (EU-CTR) German Clinical Trials Register (DRKS) Iranian Registry of Clinical Trials (IRCT) ISRCTN (UK) International Traditional Medicine Clinical Trial Registry (ITMCTR) Japan Registry of Clinical Trials (jRCT) Japan Primary Registries Network (JPRN) Lebanese Clinical Trials Registry (LBCTR) Overview of Medical Research in the Netherlands (OMON) Thai Clinical Trials Registry (TCTR) Pan African Clinical Trial Registry (PACTR) Peruvian Clinical Trial Registry (REPEC) Sri Lanka Clinical Trials Registry (SLCTR)

  Read more →
• 

  Scene text

  Scene text is text that appears in an image captured by a camera in an outdoor environment. The detection and recognition of scene text from camera captured images are computer vision tasks which became important after smart phones with good cameras became ubiquitous. The text in scene images varies in shape, font, colour and position. The recognition of scene text is further complicated sometimes by non-uniform illumination and focus. To improve scene text recognition, the International Conference on Document Analysis and Recognition (ICDAR) conducts a robust reading competition once in two years. The competition was held in 2003, 2005 and during every ICDAR conference. International association for pattern recognition (IAPR) has created a list of datasets as Reading systems. == Text detection == Text detection is the process of detecting the text present in the image, followed by surrounding it with a rectangular bounding box. Text detection can be carried out using image based techniques or frequency based techniques. In image based techniques, an image is segmented into multiple segments. Each segment is a connected component of pixels with similar characteristics. The statistical features of connected components are utilised to group them and form the text. Machine learning approaches such as support vector machine and convolutional neural networks are used to classify the components into text and non-text. In frequency based techniques, discrete Fourier transform (DFT) or discrete wavelet transform (DWT) are used to extract the high frequency coefficients. It is assumed that the text present in an image has high frequency components and selecting only the high frequency coefficients filters the text from the non-text regions in an image. == Word recognition == In word recognition, the text is assumed to be already detected and located and the rectangular bounding box containing the text is available. The word present in the bounding box needs to be recognized. The methods available to perform word recognition can be broadly classified into top-down and bottom-up approaches. In the top-down approaches, a set of words from a dictionary is used to identify which word suits the given image. Images are not segmented in most of these methods. Hence, the top-down approach is sometimes referred as segmentation free recognition. In the bottom-up approaches, the image is segmented into multiple components and the segmented image is passed through a recognition engine. Either an off the shelf Optical character recognition (OCR) engine or a custom-trained one is used to recognise the text.

  Read more →
• 

  Conversica

  Conversica is a US-based cloud software technology company, headquartered in San Mateo, California, that provides two-way AI-driven conversational software and a suite of Intelligent Virtual Assistants for businesses to engage customers via email, chat, and SMS. == History == 2007: The company was founded by Ben Brigham in Bellingham, Washington, originally as AutoFerret.com. The company's initial product was a Customer Relationship Management (CRM) targeted at automotive dealerships. This soon expanded to lead generation, and then lead validation and qualification. The AI Conversica uses currently was made to follow up on and filter out low-quality leads. The focus of the company shifted toward this automated lead engagement technology. 2010: The company started commercially selling AVA, the first Automated Virtual Assistant for sales, and the company name was changed to AVA.ai. Early customers for AVA were automotive dealerships. As the company moved away from generating leads themselves, and providing the CRM themselves, it became necessary to integrate with existing CRM and Marketing Automation platforms, such as DealerSocket, VinSolutions and Salesforce. 2013: The company raised $16m Series A funding, led by Kennet Partners, and named Mark Bradley as CEO. It also moved its headquarters from Bellingham, Washington to Foster City, California. 2014: The company changed its name from AVA.ai to Conversica. 2015: Alex Terry joined Conversica as its CEO. The business expanded to include customers in additional verticals, including technology, education, and financial services. 2016: The company raised $34m Series B funding, led by Providence Strategic Growth. 2017: Conversica expanded its intelligent automation platform and IVAs to support additional communication channels (e-mail and SMS text messaging) and communication languages. Conversica also opened a new technology center in Seattle, Washington to expand its AI and machine learning capabilities. 2018: The company raised $31m Series C funding, led by Providence Strategic Growth. Conversica also acquired Intelligens.ai, providing a regional presence in Latin America with an office in Las Condes, Santiago, Chile. The company launched an AI-powered Admissions Assistant for Higher Education industry. 2019: Conversica was selected by Fast Company magazine as one of the Top 10 Most Innovative AI Companies in the World, and was named Marketo's Technology Partner of the Year. The company officially expanded into the EMEA region with the opening of a London office. As of August 2019, Conversica has over 50 different integrations with third parties. In October Conversica won three awards at the fourth annual Global Annual Achievement Awards for Artificial Intelligence. Also that month, Alex Terry stepped down from his role as CEO and was replaced by Jim Kaskade. 2020: As part of Conversica's response to COVID-19, they optimized the business to become profitable in both 2Q20 and 3Q20, before reinvesting in 4Q20. The company transitioned both international operations in EMEA and LATAM to an indirect model with partners (LeadFabric and Nectia Cloud Solutions respectively), and moved a portion of its US-based employees to near-shore centers in Mexico and Brazil, effectively downsizing the company from 250 to 200. Conversica's reseller partner, Nectia, is a major Latin American affiliate and Chile's number one Salesforce partner, and, as part of the partnership, Nectia devoted capital to a brand new company segment, Predict-IA, dedicated to web-based artificial intelligent solutions. Predict-IA was able to immediately service all LATAM opportunities and clients with Conversica's AI Assistants with end-to-end services (marketing, sales, professional services, customer success, and technical support). Conversica's reseller partner, Leadfabric, has offices in Belgium, Amsterdam, Paris, UK, Taiwan, and Romania. == Technology == Conversica's Revenue Digital Assistants™ are AI assistants who engage with leads, prospects, customers, employees, and other persons of interest (Contacts) in a two-way human-like manner, via email, SMS text, and website chat, in English, French, German, Spanish, Portuguese, and Japanese. The RDAs are built on an Intelligent Automation platform that leverages natural language understanding, natural language processing, natural language generation, deep learning and machine learning. The Assistants are generally deployed alongside sales and marketing, customer success, account management, and higher education admissions teams, as part of an augmented workforce. The Intelligent Automation platform integrates with over 50 external systems, including CRM, Marketing Automation, and other systems of record. A partial list of integration partners includes: Salesforce, Marketo, Oracle, HubSpot, DealerSocket, Reynolds & Reynolds, CDK Global, VinSolutions and many more.

  Read more →
• 

  News analytics

  In trading strategy, news analysis refers to the measurement of the various qualitative and quantitative attributes of textual (unstructured data) news stories. Some of these attributes are: sentiment, relevance, and novelty. Expressing news stories as numbers and metadata permits the manipulation of everyday information in a mathematical and statistical way. This data is often used in financial markets as part of a trading strategy or by businesses to judge market sentiment and make better business decisions. News analytics are usually derived through automated text analysis and applied to digital texts using elements from natural language processing and machine learning such as latent semantic analysis, support vector machines, "bag of words" among other techniques. == Applications and strategies == The application of sophisticated linguistic analysis to news and social media has grown from an area of research to mature product solutions since 2007. News analytics and news sentiment calculations are now routinely used by both buy-side and sell-side in alpha generation, trading execution, risk management, and market surveillance and compliance. There is however a good deal of variation in the quality, effectiveness and completeness of currently available solutions. A large number of companies use news analysis to help them make better business decisions. Academic researchers have become interested in news analysis especially with regards to predicting stock price movements, volatility and traded volume. Provided a set of values such as sentiment and relevance as well as the frequency of news arrivals, it is possible to construct news sentiment scores for multiple asset classes such as equities, Forex, fixed income, and commodities. Sentiment scores can be constructed at various horizons to meet the different needs and objectives of high and low frequency trading strategies, whilst characteristics such as direction and volatility of asset returns as well as the traded volume may be addressed more directly via the construction of tailor-made sentiment scores. Scores are generally constructed as a range of values. For instance, values may range between 0 and 100, where values above and below 50 convey positive and negative sentiment, respectively. === Absolute return strategies === The objective of absolute return strategies is absolute (positive) returns regardless of the direction of the financial market. To meet this objective, such strategies typically involve opportunistic long and short positions in selected instruments with zero or limited market exposure. In statistical terms, absolute return strategies should have very low correlation with the market return. Typically, hedge funds tend to employ absolute return strategies. Below, a few examples show how news analysis can be applied in the absolute return strategy space with the purpose to identify alpha opportunities applying a market neutral strategy or based on volatility trading. Example 1 Scenario: The gap between the news sentiment scores for direction, S {\displaystyle S} , of Company X {\displaystyle X} and Market Y {\displaystyle Y} has moved beyond + 20 {\displaystyle +20} . That is, S X − S Y {\displaystyle S_{X}-S_{Y}} ≥ 20 {\displaystyle 20} . Action: Buy the stock on Company X {\displaystyle X} and short the future on Market Y {\displaystyle Y} . Exit Strategy: When the gap in the news sentiment scores for direction of Company X {\displaystyle X} and Market Y {\displaystyle Y} has disappeared, S X − S Y {\displaystyle S_{X}-S_{Y}} = 0 {\displaystyle 0} , sell the stock on Company X {\displaystyle X} and go long the future on Market Y {\displaystyle Y} to close the positions. Example 2 Scenario: The news sentiment score for volatility of Company X {\displaystyle X} goes above 70 {\displaystyle 70} out of 100 {\displaystyle 100} indicating an expected volatility above the option implied volatility. Action: Buy a short-dated straddle (the purchase of both a put and a call) on the stock of Company X {\displaystyle X} . Exit Strategy: Keep the straddle on Company X {\displaystyle X} until expiry or until a certain profit target has been reached. === Relative return strategies === The objective of relative return strategies is to either replicate (passive management) or outperform (active management) a theoretical passive reference portfolio or benchmark. To meet these objectives such strategies typically involve long positions in selected instruments. In statistical terms, relative return strategies often have high correlation with the market return. Typically, mutual funds tend to employ relative return strategies. Below, a few examples show how news analysis can be applied in the relative return strategy space with the purpose to outperform the market applying a stock picking strategy and by making tactical tilts to ones asset allocation model. Example 1 Scenario: The news sentiment score for direction of Company X {\displaystyle X} goes above 70 {\displaystyle 70} out of 100 {\displaystyle 100} . Action: Buy the stock on Company X {\displaystyle X} . Exit Strategy: When the news sentiment score for direction of Company X {\displaystyle X} falls below 60 {\displaystyle 60} , sell the stock on Company X {\displaystyle X} to close the position. Example 2 Scenario: The news sentiment score for direction of Sector Z {\displaystyle Z} goes above 70 {\displaystyle 70} out of 100 {\displaystyle 100} . Action: Include Sector Z {\displaystyle Z} as a tactical bet in the asset allocation model. Exit Strategy: When the news sentiment score for direction of Sector Z {\displaystyle Z} falls below 60 {\displaystyle 60} , remove the tactical bet for Sector Z {\displaystyle Z} from the asset allocation model. === Financial risk management === The objective of financial risk management is to create economic value in a firm or to maintain a certain risk profile of an investment portfolio by using financial instruments to manage risk exposures, particularly credit risk and market risk. Other types include Foreign exchange, Shape, Volatility, Sector, Liquidity, Inflation risks, etc. Below, a few examples show how news analysis can be applied in the financial risk management space with the purpose to either arrive at better risk estimates in terms of Value at Risk (VaR) or to manage the risk of a portfolio to meet ones portfolio mandate. Example 1 Scenario: The bank operates a VaR model to manage the overall market risk of its portfolio. Action: Estimate the portfolio covariance matrix taking into account the development of the news sentiment score for volume. Implement the relevant hedges to bring the VaR of the bank in line with the desired levels. Example 2 Scenario: A portfolio manager operates his portfolio towards a certain desired risk profile. Action: Estimate the portfolio covariance matrix taking into account the development of the news sentiment score for volume. Scale the portfolio exposure according to the targeted risk profile. === Computer algorithms using news analytics === Within 0.33 seconds, computer algorithms using news analytics can notify subscribers which company the news is about, if the news article sentiment is positive or negative, if the news is ranked as high or low relative importance … relative relevance. the stock price reaction and the increase in trade volume is concentrated in the first 5 seconds after an news article is released. === Algorithmic order execution === The objective of algorithmic order execution, which is part of the concept of algorithmic trading, is to reduce trading costs by optimizing on the timing of a given order. It is widely used by hedge funds, pension funds, mutual funds, and other institutional traders to divide up large trades into several smaller trades to manage market impact, opportunity cost, and risk more effectively. The example below shows how news analysis can be applied in the algorithmic order execution space with the purpose to arrive at more efficient algorithmic trading systems. Example 1 Scenario: A large order needs to be placed in the market for the stock on Company X {\displaystyle X} . Action: Scale the daily volume distribution for Company X {\displaystyle X} applied in the algorithmic trading system, thus taking into account the news sentiment score for volume. This is followed by the creation of the desired trading distribution forcing greater market participation during the periods of the day when volume is expected to be heaviest. == Effects == Being able to express news stories as numbers permits the manipulation of everyday information in a statistical way that allows computers not only to make decisions once made only by humans, but to do so more efficiently. Since market participants are always looking for an edge, the speed of computer connections and the delivery of news analysis, measured in milliseconds, have become essential.

  Read more →
• 

  A Logical Calculus of the Ideas Immanent in Nervous Activity

  "A Logical Calculus of the Ideas Immanent in Nervous Activity" is a 1943 paper written by Warren Sturgis McCulloch and Walter Pitts, published in the journal The Bulletin of Mathematical Biophysics. The paper proposed a mathematical model of the nervous system as a network of simple logical elements, later known as artificial neurons, or McCulloch–Pitts neurons. These neurons receive inputs, perform a weighted sum, and fire an output signal based on a threshold function. By connecting these units in various configurations, McCulloch and Pitts demonstrated that their model could perform all logical functions. It is a seminal work in cognitive science, computational neuroscience, computer science, and artificial intelligence. It was a foundational result in automata theory. John von Neumann cited it as a significant result. == Mathematics == The artificial neuron used in the original paper is slightly different from the modern version. They considered neural networks that operate in discrete steps of time t = 0 , 1 , … {\displaystyle t=0,1,\dots } . The neural network contains a number of neurons. Let the state of a neuron i {\displaystyle i} at time t {\displaystyle t} be N i ( t ) {\displaystyle N_{i}(t)} . The state of a neuron can either be 0 or 1, standing for "not firing" and "firing". Each neuron also has a firing threshold θ {\displaystyle \theta } , such that it fires if the total input exceeds the threshold. Each neuron can connect to any other neuron (including itself) with positive synapses (excitatory) or negative synapses (inhibitory). That is, each neuron can connect to another neuron with a weight w {\displaystyle w} taking an integer value. A peripheral afferent is a neuron with no incoming synapses. We can regard each neural network as a directed graph, with the nodes being the neurons, and the directed edges being the synapses. A neural network has a circle or a circuit if there exists a directed circle in the graph. Let w i j ( t ) {\displaystyle w_{ij}(t)} be the connection weight from neuron j {\displaystyle j} to neuron i {\displaystyle i} at time t {\displaystyle t} , then its next state is N i ( t + 1 ) = H ( ∑ j = 1 n w i j ( t ) N j ( t ) − θ i ( t ) ) , {\displaystyle N_{i}(t+1)=H\left(\sum _{j=1}^{n}w_{ij}(t)N_{j}(t)-\theta _{i}(t)\right),} where H {\displaystyle H} is the Heaviside step function (outputting 1 if the input is greater than or equal to 0, and 0 otherwise). === Symbolic logic === The paper used, as a logical language for describing neural networks, "Language II" from The Logical Syntax of Language by Rudolf Carnap with some notations taken from Principia Mathematica by Alfred North Whitehead and Bertrand Russell. Language II covers substantial parts of classical mathematics, including real analysis and portions of set theory. To describe a neural network with peripheral afferents N 1 , N 2 , … , N p {\displaystyle N_{1},N_{2},\dots ,N_{p}} and non-peripheral afferents N p + 1 , N p + 2 , … , N n {\displaystyle N_{p+1},N_{p+2},\dots ,N_{n}} they considered logical predicate of form P r ( N 1 , N 2 , … , N p , t ) {\displaystyle Pr(N_{1},N_{2},\dots ,N_{p},t)} where P r {\displaystyle Pr} is a first-order logic predicate function (a function that outputs a boolean), N 1 , … , N p {\displaystyle N_{1},\dots ,N_{p}} are predicates that take t {\displaystyle t} as an argument, and t {\displaystyle t} is the only free variable in the predicate. Intuitively speaking, N 1 , … , N p {\displaystyle N_{1},\dots ,N_{p}} specifies the binary input patterns going into the neural network over all time, and P r ( N 1 , N 2 , … , N n , t ) {\displaystyle Pr(N_{1},N_{2},\dots ,N_{n},t)} is a function that takes some binary input patterns, and constructs an output binary pattern P r ( N 1 , N 2 , … , N n , 0 ) , P r ( N 1 , N 2 , … , N n , 1 ) , … {\displaystyle Pr(N_{1},N_{2},\dots ,N_{n},0),Pr(N_{1},N_{2},\dots ,N_{n},1),\dots } . A logical sentence P r ( N 1 , N 2 , … , N n , t ) {\displaystyle Pr(N_{1},N_{2},\dots ,N_{n},t)} is realized by a neural network iff there exists a time-delay T ≥ 0 {\displaystyle T\geq 0} , a neuron i {\displaystyle i} in the network, and an initial state for the non-peripheral neurons N p + 1 ( 0 ) , … , N n ( 0 ) {\displaystyle N_{p+1}(0),\dots ,N_{n}(0)} , such that for any time t {\displaystyle t} , the truth-value of the logical sentence is equal to the state of the neuron i {\displaystyle i} at time t + T {\displaystyle t+T} . That is, ∀ t = 0 , 1 , 2 , … , P r ( N 1 , N 2 , … , N p , t ) = N i ( t + T ) {\displaystyle \forall t=0,1,2,\dots ,\quad Pr(N_{1},N_{2},\dots ,N_{p},t)=N_{i}(t+T)} === Equivalence === In the paper, they considered some alternative definitions of artificial neural networks, and have shown them to be equivalent, that is, neural networks under one definition realizes precisely the same logical sentences as neural networks under another definition. They considered three forms of inhibition: relative inhibition, absolute inhibition, and extinction. The definition above is relative inhibition. By "absolute inhibition" they meant that if any negative synapse fires, then the neuron will not fire. By "extinction" they meant that if at time t {\displaystyle t} , any inhibitory synapse fires on a neuron i {\displaystyle i} , then θ i ( t + j ) = θ i ( 0 ) + b j {\displaystyle \theta _{i}(t+j)=\theta _{i}(0)+b_{j}} for j = 1 , 2 , 3 , … {\displaystyle j=1,2,3,\dots } , until the next time an inhibitory synapse fires on i {\displaystyle i} . It is required that b j = 0 {\displaystyle b_{j}=0} for all large j {\displaystyle j} . Theorem 4 and 5 state that these are equivalent. They considered three forms of excitation: spatial summation, temporal summation, and facilitation. The definition above is spatial summation (which they pictured as having multiple synapses placed close together, so that the effect of their firing sums up). By "temporal summation" they meant that the total incoming signal is ∑ τ = 0 T ∑ j = 1 n w i j ( t ) N j ( t − τ ) {\displaystyle \sum _{\tau =0}^{T}\sum _{j=1}^{n}w_{ij}(t)N_{j}(t-\tau )} for some T ≥ 1 {\displaystyle T\geq 1} . By "facilitation" they meant the same as extinction, except that b j ≤ 0 {\displaystyle b_{j}\leq 0} . Theorem 6 states that these are equivalent. They considered neural networks that do not change, and those that change by Hebbian learning. That is, they assume that at t = 0 {\displaystyle t=0} , some excitatory synaptic connections are not active. If at any t {\displaystyle t} , both N i ( t ) = 1 , N j ( t ) = 1 {\displaystyle N_{i}(t)=1,N_{j}(t)=1} , then any latent excitatory synapse between i , j {\displaystyle i,j} becomes active. Theorem 7 states that these are equivalent. === Logical expressivity === They considered "temporal propositional expressions" (TPE), which are propositional formulas with one free variable t {\displaystyle t} . For example, N 1 ( t ) ∨ N 2 ( t ) ∧ ¬ N 3 ( t ) {\displaystyle N_{1}(t)\vee N_{2}(t)\wedge \neg N_{3}(t)} is such an expression. Theorem 1 and 2 together showed that neural nets without circles are equivalent to TPE. For neural nets with loops, they noted that "realizable P r {\displaystyle Pr} may involve reference to past events of an indefinite degree of remoteness". These then encodes for sentences like "There was some x such that x was a ψ" or ( ∃ x ) ( ψ x ) {\displaystyle (\exists x)(\psi x)} . Theorems 8 to 10 showed that neural nets with loops can encode all first-order logic with equality and conversely, any looped neural networks is equivalent to a sentence in first-order logic with equality, thus showing that they are equivalent in logical expressiveness. As a remark, they noted that a neural network, if furnished with a tape, scanners, and write-heads, is equivalent to a Turing machine, and conversely, every Turing machine is equivalent to some such neural network. Thus, these neural networks are equivalent to Turing computability and Church's lambda-definability. == Context == === Previous work === The paper built upon several previous strands of work. In the symbolic logic side, it built on the previous work by Carnap, Whitehead, and Russell. This was contributed by Walter Pitts, who had a strong proficiency with symbolic logic. Pitts provided mathematical and logical rigor to McCulloch’s vague ideas on psychons (atoms of psychological events) and circular causality. In the neuroscience side, it built on previous work by the mathematical biology research group centered around Nicolas Rashevsky, of which McCulloch was a member. The paper was published in the Bulletin of Mathematical Biophysics, which was founded by Rashevsky in 1939. During the late 1930s, Rashevsky's research group was producing papers that had difficulty publishing in other journals at the time, so Rashevsky decided to found a new journal exclusively devoted to mathematical biophysics. Also in the Rashevsky's group was Alston Scott Householder, who in 1941 published an abstract model

  Read more →
• 

  Natural Language Toolkit

  The Natural Language Toolkit, or more commonly NLTK, is a suite of libraries and programs for symbolic and statistical natural language processing (NLP) for English written in the Python programming language. It supports classification, tokenization, stemming, tagging, parsing, and semantic reasoning functionalities. It was developed by Steven Bird and Edward Loper in the Department of Computer and Information Science at the University of Pennsylvania. NLTK includes graphical demonstrations and sample data. It is accompanied by a book that explains the underlying concepts behind the language processing tasks supported by the toolkit, plus a cookbook. NLTK is intended to support research and teaching in NLP or closely related areas, including empirical linguistics, cognitive science, artificial intelligence, information retrieval, and machine learning. NLTK has been used successfully as a teaching tool, as an individual study tool, and as a platform for prototyping and building research systems. == Library highlights == Discourse representation Lexical analysis: Word and text tokenizer n-gram and collocations Part-of-speech tagger Tree model and Text chunker for capturing Named-entity recognition

  Read more →
• 

  VACUUM

  VACUUM is a set of normative guidance principles for achieving training and test dataset quality for structured datasets in data science and machine learning. The garbage-in, garbage out principle motivates a solution to the problem of data quality but does not offer a specific solution. Unlike the majority of the ad-hoc data quality assessment metrics often used by practitioners VACUUM specifies qualitative principles for data quality management and serves as a basis for defining more detailed quantitative metrics of data quality. VACUUM is an acronym that stands for: valid accurate consistent uniform unified model

  Read more →
• 

  Language engineering

  Language engineering involves the creation of natural language processing systems, whose cost and outputs are measurable and predictable. It is a distinct field contrasted to natural language processing and computational linguistics. A recent trend of language engineering is the use of Semantic Web technologies for the creation, archiving, processing, and retrieval of machine processable language data. Meta-Language Engineering is a proposed extension of Language Engineering first recorded in 2025, associated with the work of Delyone de Paula Canedo Filho. The term is used to designate an approach that, in addition to natural language processing, encompasses the symbolic, cognitive, and epistemological structuring of language systems.

  Read more →
• 

  Biohybrid microswimmer

  A biohybrid microswimmer also known as biohybrid nanorobot, can be defined as a microswimmer that consist of both biological and artificial constituents, for instance, one or several living microorganisms attached to one or various synthetic parts. In recent years nanoscopic and mesoscopic objects have been designed to collectively move through direct inspiration from nature or by harnessing its existing tools. Small mesoscopic to nanoscopic systems typically operate at low Reynolds numbers (Re ≪ 1), and understanding their motion becomes challenging. For locomotion to occur, the symmetry of the system must be broken. In addition, collective motion requires a coupling mechanism between the entities that make up the collective. To develop mesoscopic to nanoscopic entities capable of swarming behaviour, it has been hypothesised that the entities are characterised by broken symmetry with a well-defined morphology, and are powered with some material capable of harvesting energy. If the harvested energy results in a field surrounding the object, then this field can couple with the field of a neighbouring object and bring some coordination to the collective behaviour. Such robotic swarms have been categorised by an online expert panel as among the 10 great unresolved group challenges in the area of robotics. Although investigation of their underlying mechanism of action is still in its infancy, various systems have been developed that are capable of undergoing controlled and uncontrolled swarming motion by harvesting energy (e.g., light, thermal, etc.). Over the past decade, biohybrid microrobots, in which living mobile microorganisms are physically integrated with untethered artificial structures, have gained growing interest to enable the active locomotion and cargo delivery to a target destination. In addition to the motility, the intrinsic capabilities of sensing and eliciting an appropriate response to artificial and environmental changes make cell-based biohybrid microrobots appealing for transportation of cargo to the inaccessible cavities of the human body for local active delivery of diagnostic and therapeutic agents. == Background == Biohybrid microswimmers can be defined as microswimmers that consist of both biological and artificial constituents, for instance, one or several living microorganisms attached to one or various synthetic parts. The pioneers of this field, ahead of their time, were Montemagno and Bachand with a 1999 work regarding specific attachment strategies of biological molecules to nanofabricated substrates enabling the preparation of hybrid inorganic/organic nanoelectromechanical systems, so called NEMS. They described the production of large amounts of F1-ATPase from the thermophilic bacteria Bacillus PS3 for the preparation of F1-ATPase biomolecular motors immobilized on a nanoarray pattern of gold, copper or nickel produced by electron beam lithography. These proteins were attached to one micron microspheres tagged with a synthetic peptide. Consequently, they accomplished the preparation of a platform with chemically active sites and the development of biohybrid devices capable of converting energy of biomolecular motors into useful work. One of the most fundamental questions in science is what defines life. Collective motion is one of the hallmarks of life. This is commonly observed in nature at various dimensional levels as energized entities gather, in a concerted effort, into motile aggregated patterns. These motile aggregated events can be noticed, among many others, as dynamic swarms; e.g., unicellular organisms such as bacteria, locust swarms, or the flocking behaviour of birds. Ever since Newton established his equations of motion, the mystery of motion on the microscale has emerged frequently in scientific history, as famously demonstrated by a couple of articles that should be discussed briefly. First, an essential concept, popularized by Osborne Reynolds, is that the relative importance of inertia and viscosity for the motion of a fluid depends on certain details of the system under consideration. The Reynolds number Re, named in his honor, quantifies this comparison as a dimensionless ratio of characteristic inertial and viscous forces: R e = ρ u l μ {\displaystyle \mathrm {Re} ={\frac {\rho ul}{\mu }}} Here, ρ represents the density of the fluid; u is a characteristic velocity of the system (for instance, the velocity of a swimming particle); l is a characteristic length scale (e.g., the swimmer size); and μ is the viscosity of the fluid. Taking the suspending fluid to be water, and using experimentally observed values for u, one can determine that inertia is important for macroscopic swimmers like fish (Re = 100), while viscosity dominates the motion of microscale swimmers like bacteria (Re = 10−4). The overwhelming importance of viscosity for swimming at the micrometer scale has profound implications for swimming strategy. This has been discussed memorably by E. M. Purcell, who invited the reader into the world of microorganisms and theoretically studied the conditions of their motion. In the first place, propulsion strategies of large scale swimmers often involve imparting momentum to the surrounding fluid in periodic discrete events, such as vortex shedding, and coasting between these events through inertia. This cannot be effective for microscale swimmers like bacteria: due to the large viscous damping, the inertial coasting time of a micron-sized object is on the order of 1 μs. The coasting distance of a microorganism moving at a typical speed is about 0.1 angstroms (Å). Purcell concluded that only forces that are exerted in the present moment on a microscale body contribute to its propulsion, so a constant energy conversion method is essential. Microorganisms have optimized their metabolism for continuous energy production, while purely artificial microswimmers (microrobots) must obtain energy from the environment, since their on-board-storage-capacity is very limited. As a further consequence of the continuous dissipation of energy, biological and artificial microswimmers do not obey the laws of equilibrium statistical physics, and need to be described by non-equilibrium dynamics. Mathematically, Purcell explored the implications of low Reynolds number by taking the Navier-Stokes equation and eliminating the inertial terms: μ ∇ 2 u − ∇ p = 0 {\displaystyle {\begin{aligned}\mu \nabla ^{2}\mathbf {u} -{\boldsymbol {\nabla }}p&={\boldsymbol {0}}\\\end{aligned}}} where u {\displaystyle \mathbf {u} } is the velocity of the fluid and ∇ p {\displaystyle {\boldsymbol {\nabla }}p} is the gradient of the pressure. As Purcell noted, the resulting equation — the Stokes equation — contains no explicit time dependence. This has some important consequences for how a suspended body (e.g., a bacterium) can swim through periodic mechanical motions or deformations (e.g., of a flagellum). First, the rate of motion is practically irrelevant for the motion of the microswimmer and of the surrounding fluid: changing the rate of motion will change the scale of the velocities of the fluid and of the microswimmer, but it will not change the pattern of fluid flow. Secondly, reversing the direction of mechanical motion will simply reverse all velocities in the system. These properties of the Stokes equation severely restrict the range of feasible swimming strategies. Recent publications of biohybrid microswimmers include the use of sperm cells, contractive muscle cells, and bacteria as biological components, as they can efficiently convert chemical energy into movement, and additionally are capable of performing complicated motion depending on environmental conditions. In this sense, biohybrid microswimmer systems can be described as the combination of different functional components: cargo and carrier. The cargo is an element of interest to be moved (and possibly released) in a customized way. The carrier is the component responsible for the movement of the biohybrid, transporting the desired cargo, which is linked to its surface. The great majority of these systems rely on biological motile propulsion for the transportation of synthetic cargo for targeted drug delivery/ There are also examples of the opposite case: artificial microswimmers with biological cargo systems. Over the past decade, biohybrid microrobots, in which living mobile microorganisms are physically integrated with untethered artificial structures, have gained growing interest to enable the active locomotion and cargo delivery to a target destination. In addition to the motility, the intrinsic capabilities of sensing and eliciting an appropriate response to artificial and environmental changes make cell-based biohybrid microrobots appealing for transportation of cargo to the inaccessible cavities of the human body for local active delivery of diagnostic and therapeutic agents. Active locomotion, targeting and steering of concentrated therape

  Read more →
• 

  Facial age estimation

  Facial age estimation is the use of artificial intelligence to estimate the age of a person based on their facial features. Computer vision techniques are used to analyse the facial features in the images of millions of people whose age is known and then deep learning is used to create an algorithm that tries to predict the age of an unknown person. The key use of the technology is to prevent access to age-restricted goods and services. Examples include restricting children from accessing internet pornography, checking that they meet a mandatory minimum age when registering for an account on social media, or preventing adults from accessing websites, online chat or games designed only for use by children. The technology is distinct from facial recognition systems as the software does not attempt to uniquely identify the individual. Researchers have applied neural networks for age estimation since at least 2010. == Evaluation == An ongoing study by the National Institute of Standards and Technology (NIST) entitled 'Face Analysis Technology Evaluation' seeks to establish the technical performance of prototype age estimation algorithms submitted by academic teams and software vendors including Brno University of Technology, Czech Technical University in Prague, Dermalog, IDEMIA, Incode Technologies Inc, Jumio, Nominder, Rank One Computing, Unissey and Yoti. == Public sector use == The UK government has explored using facial age estimation at the UK border as an alternative to bone X-rays and MRI scans when determining child status of asylum seekers. == Commercial use == Commercial users of facial age estimation include Instagram and OnlyFans. In January 2025, John Lewis & Partners announced that had started using the technology to check the age of people shopping for knives on its website, to comply with UK legislation to limit knife crime. In the UK, several supermarket chains have taken part in Home Office trials of the technology to automate the checking of a customer's age when buying age-restricted goods such as alcohol. UK legislation introduced in January 2025 mandates robust forms of age verification hosting adult content viewable in the UK by July 2025. Allowable methods include facial age estimation. == Criticism == Adam Schwartz, a lawyer for the Electronic Frontier Foundation, criticized the use of facial age estimation software, noting its inaccuracy especially in cases of minorities and women, as was found in NIST's 2024 report. Twenty organisations jointly under European Digital Rights called the practice a "systematic and invasive processing of young people's data" that risks discriminatory profiling.

  Read more →
• 

  BERT (language model)

  Bidirectional encoder representations from transformers (BERT) is a language model introduced in October 2018 by researchers at Google. It learns to represent text as a sequence of vectors using self-supervised learning. It uses the encoder-only transformer architecture. BERT dramatically improved the state of the art for large language models. As of 2020, BERT is a ubiquitous baseline in natural language processing (NLP) experiments. BERT is trained by masked token prediction and next sentence prediction. With this training, BERT learns contextual, latent representations of tokens in their context, similar to ELMo and GPT-2. It found applications for many natural language processing tasks, such as coreference resolution and polysemy resolution. It improved on ELMo and spawned the study of "BERTology", which attempts to interpret what is learned by BERT. BERT was originally implemented in the English language at two model sizes, BERTBASE (110 million parameters) and BERTLARGE (340 million parameters). Both were trained on the Toronto BookCorpus (800M words) and English Wikipedia (2,500M words). The weights were released on GitHub. On March 11, 2020, 24 smaller models were released, the smallest being BERTTINY with just 4 million parameters. == Architecture == BERT is an "encoder-only" transformer architecture. At a high level, BERT consists of 4 modules: Tokenizer: This module converts a piece of English text into a sequence of integers ("tokens"). Embedding: This module converts the sequence of tokens into an array of real-valued vectors representing the tokens. It represents the conversion of discrete token types into a lower-dimensional Euclidean space. Encoder: a stack of Transformer blocks with self-attention, but without causal masking. Task head: This module converts the final representation vectors into one-shot encoded tokens again by producing a predicted probability distribution over the token types. It can be viewed as a simple decoder, decoding the latent representation into token types, or as an "un-embedding layer". The task head is necessary for pre-training, but it is often unnecessary for so-called "downstream tasks," such as question answering or sentiment classification. Instead, one removes the task head and replaces it with a newly initialized module suited for the task, and finetune the new module. The latent vector representation of the model is directly fed into this new module, allowing for sample-efficient transfer learning. === Embedding === This section describes the embedding used by BERTBASE. The other one, BERTLARGE, is similar, just larger. The tokenizer of BERT is WordPiece, which is a sub-word strategy like byte-pair encoding. Its vocabulary size is 30,000, and any token not appearing in its vocabulary is replaced by [UNK] ("unknown"). The first layer is the embedding layer, which contains three components: token type embeddings, position embeddings, and segment type embeddings. Token type: The token type is a standard embedding layer, translating a one-hot vector into a dense vector based on its token type. Position: The position embeddings are based on a token's position in the sequence. BERT uses absolute position embeddings, where each position in a sequence is mapped to a real-valued vector. Each dimension of the vector consists of a sinusoidal function that takes the position in the sequence as input. Segment type: Using a vocabulary of just 0 or 1, this embedding layer produces a dense vector based on whether the token belongs to the first or second text segment in that input. In other words, type-1 tokens are all tokens that appear after the [SEP] special token. All prior tokens are type-0. The three embedding vectors are added together representing the initial token representation as a function of these three pieces of information. After embedding, the vector representation is normalized using a LayerNorm operation, outputting a 768-dimensional vector for each input token. After this, the representation vectors are passed forward through 12 Transformer encoder blocks, and are decoded back to 30,000-dimensional vocabulary space using a basic affine transformation layer. === Architectural family === The encoder stack of BERT has 2 free parameters: L {\displaystyle L} , the number of layers, and H {\displaystyle H} , the hidden size. There are always H / 64 {\displaystyle H/64} self-attention heads, and the feed-forward/filter size is always 4 H {\displaystyle 4H} . By varying these two numbers, one obtains an entire family of BERT models. For BERT: the feed-forward size and filter size are synonymous. Both of them denote the number of dimensions in the middle layer of the feed-forward network. the hidden size and embedding size are synonymous. Both of them denote the number of real numbers used to represent a token. The notation for encoder stack is written as L/H. For example, BERTBASE is written as 12L/768H, BERTLARGE as 24L/1024H, and BERTTINY as 2L/128H. == Training == === Pre-training === BERT was pre-trained simultaneously on two tasks: Masked language modeling (MLM): In this task, BERT ingests a sequence of words, where one word may be randomly changed ("masked"), and BERT tries to predict the original words that had been changed. For example, in the sentence "The cat sat on the [MASK]," BERT would need to predict "mat." This helps BERT learn bidirectional context, meaning it understands the relationships between words not just from left to right or right to left but from both directions at the same time. Next sentence prediction (NSP): In this task, BERT is trained to predict whether one sentence logically follows another. For example, given two sentences, "The cat sat on the mat" and "It was a sunny day", BERT has to decide if the second sentence is a valid continuation of the first one. This helps BERT understand relationships between sentences, which is important for tasks like question answering or document classification. ==== Masked language modeling ==== In masked language modeling, 15% of tokens would be randomly selected for masked-prediction task, and the training objective was to predict the masked token given its context. In more detail, the selected token is: replaced with a [MASK] token with probability 80%, replaced with a random word token with probability 10%, not replaced with probability 10%. The reason not all selected tokens are masked is to avoid the dataset shift problem. The dataset shift problem arises when the distribution of inputs seen during training differs significantly from the distribution encountered during inference. A trained BERT model might be applied to word representation (like Word2Vec), where it would be run over sentences not containing any [MASK] tokens. It is later found that more diverse training objectives are generally better. As an illustrative example, consider the sentence "my dog is cute". It would first be divided into tokens like "my1 dog2 is3 cute4". Then a random token in the sentence would be picked. Let it be the 4th one "cute4". Next, there would be three possibilities: with probability 80%, the chosen token is masked, resulting in "my1 dog2 is3 [MASK]4"; with probability 10%, the chosen token is replaced by a uniformly sampled random token, such as "happy", resulting in "my1 dog2 is3 happy4"; with probability 10%, nothing is done, resulting in "my1 dog2 is3 cute4". After processing the input text, the model's 4th output vector is passed to its decoder layer, which outputs a probability distribution over its 30,000-dimensional vocabulary space. ==== Next sentence prediction ==== Given two sentences, the model predicts if they appear sequentially in the training corpus, outputting either [IsNext] or [NotNext]. During training, the algorithm sometimes samples two sentences from a single continuous span in the training corpus, while at other times, it samples two sentences from two discontinuous spans. The first sentence starts with a special token, [CLS] (for "classify"). The two sentences are separated by another special token, [SEP] (for "separate"). After processing the two sentences, the final vector for the [CLS] token is passed to a linear layer for binary classification into [IsNext] and [NotNext]. For example: Given "[CLS] my dog is cute [SEP] he likes playing [SEP]", the model should predict [IsNext]. Given "[CLS] my dog is cute [SEP] how do magnets work [SEP]", the model should predict [NotNext]. === Fine-tuning === BERT is meant as a general pretrained model for various applications in natural language processing. That is, after pre-training, BERT can be fine-tuned with fewer resources on smaller datasets to optimize its performance on specific tasks such as natural language inference and text classification, and sequence-to-sequence-based language generation tasks such as question answering and conversational response generation. The original BERT paper published results demonstrating that a small amount of fine

  Read more →