Best AI Writing Tools

Best AI Writing Tools — hands-on reviews, top picks, pricing, pros and cons and a practical how-to guide on Aizhi.

Subpixel rendering

Subpixel rendering is a method used to increase the effective resolution of a color display device. It utilizes the composition of each pixel, which consists of three subpixels of which are red, green, and blue that can each be individually addressable on the display matrix. Subpixel rendering is primarily used for text rendering on standard DPI displays. Despite the inherent color anomalies, it can also be used to render general graphics. == History == The origin of subpixel rendering as used today remains controversial. Apple Inc., IBM, and Microsoft patented various implementations that differed in technical details owing to the different purposes for which their technologies were intended. Microsoft held several patents in the United States for subpixel rendering technology used in text rendering on RGB Stripe layouts. The patents 6,219,025; 6,239,783; 6,307,566; 6,225,973; 6,243,070; 6,393,145; 6,421,054; 6,282,327; and 6,624,828 were filed between October 7, 1998, and October 7, 1999, and expired on July 30, 2019. Analysis of the patent by FreeType indicates that the patent does not cover the idea of subpixel rendering, but rather the actual filter used as a last step to balance the color. Microsoft's patent describes the smallest possible filter that distributes each subpixel value equally among the R, G, and B pixels. Any other filter will either be blurrier or will introduce color artifacts. Apple was able to use it in Mac OS X due to a patent cross-licensing agreement. == Characteristics == A single pixel on a color display is made of several subpixels, typically three arranged left-to-right as red, green, and blue (RGB). The components are readily visible with a small magnifying glass, such as a loupe. These pixel components appear as a single color to the human eye because of blurring by optics and spatial integration by nerve cells in the eye. However, the eye is much more sensitive to the location. Therefore, turning on the G and B of one pixel and the R of the next pixel to the right will produce a white dot, but it will appear to be 1/3 of a pixel to the right of the white dot that would be seen from the RGB of only the first pixel. Subpixel rendering leverages this to provide three times the horizontal resolution of the rendered image. However, it has to blur this image to produce the correct color by ensuring the same amount of red, green, and blue are turned on as when no subpixel rendering is being done. Subpixel rendering does not necessitate the use of antialiasing. It gives a smoother result regardless of whether antialiasing is used or not since it artificially increases the resolution. However, it introduces color aliasing since subpixels are colored. Subsequent filtering applied to remove the color artifacts is a form of antialiasing, although its purpose is not smoothing jagged shapes as in conventional antialiasing. Subpixel rendering requires the software to know the layout of the subpixels. The most common reason it is wrong is monitors that can be rotated 90 (or 180) degrees, though monitors are manufactured with other arrangements of the subpixels, such as BGR or in triangles, or with 4 colors like RGBW squares. On any such display the result of incorrect subpixel rendering will be worse than if no subpixel rendering was done at all (it will not produce color artifacts, but it will produce noisy edges). == Implementations == === Apple II === Steve Gibson has claimed that the Apple II, introduced in 1977, supports an early form of subpixel rendering in its high-resolution (280×192) graphics mode. The Wozniak patent only used 2 "sub-pixels". The bytes that comprise the Apple II high-resolution screen buffer contain seven visible bits (each corresponding directly to a pixel) and a flag bit used to select between purple/green or blue/orange color sets. Each pixel, since it is represented by a single bit, is either on or off; there are no bits within the pixel itself for specifying color or brightness. Color is instead created as an artifact of the NTSC color encoding scheme, determined by horizontal position: pixels with even horizontal coordinates are always purple (or blue, if the flag bit is set), and odd pixels are always green (or orange). Two lit pixels next to each other are always white, regardless of whether the pair is even/odd or odd/even, and irrespective of the value of the flag bit. This is an approximation, but it is what most programmers of the time would have in mind while working with the Apple's high-resolution mode. Gibson's example claims that because two adjacent bits form a white block, there are, in fact, two bits per pixel: one that activates the pixel's purple left half and the other that activates its green right half. If the programmer instead activates the green right half of a pixel and the purple left half of the next pixel, the result is a white block 1/2 pixel to the right, which is indeed an instance of subpixel rendering. However, it is not clear whether any programmers of the Apple II have considered the pairs of bits as pixels—instead calling each bit a pixel. The flag bit in each byte affects color by shifting pixels half a pixel-width to the right. This half-pixel shift was exploited by some graphics software, such as HRCG (High-Resolution Character Generator), an Apple utility that displayed text using the high-resolution graphics mode, to smooth diagonals. === ClearType === Microsoft announced its subpixel rendering technology, called ClearType, at COMDEX in 1998. Microsoft published a paper in May 2000, Displaced Filtering for Patterned Displays, describing the filtering behind ClearType. It was then made available in Windows XP. Still, it was not activated by default until Windows Vista, while Windows XP OEMs could and did change the default setting. === FreeType === FreeType, the library used by most current software on the X Window System, contains two open source implementations. The original implementation uses the ClearType antialiasing filters and carries the following notice: "The colour filtering algorithm of Microsoft's ClearType technology for subpixel rendering is covered by patents; for this reason, the corresponding code in FreeType is disabled by default. Note that subpixel rendering per se is prior art; using a different colour filter thus easily circumvents Microsoft's patent claims." FreeType offers a variety of color filters. Since version 2.6.2, the default filter is light, a filter that is both normalized (value sums up to 1) and color-balanced (eliminate color fringes at the cost of resolution). Since version 2.8.1, a second implementation exists, called Harmony, that "offers high quality LCD-optimized output without resorting to ClearType techniques of resolution tripling and filtering". This is the method enabled by default. When using this method, "each color channel is generated separately after shifting the glyph outline, capitalizing on the fact that the color grids on LCD panels are shifted by a third of a pixel. This output is indistinguishable from ClearType with a light 3-tap filter." Since the Harmony method does not require additional filtering, it is not covered by the ClearType patents. === CoolType === Adobe created their own subpixel renderer called CoolType, allowing them to display documents the same way across various operating systems: Windows, MacOS, Linux etc. When it was launched around the year 2001, CoolType supported a wider range of fonts than Microsoft's ClearType, which at the time was limited to TrueType fonts. In contrast, Adobe's CoolType also supported PostScript fonts (and their OpenType equivalents). === macOS === Mac OS X (later OS X, now macOS) also used subpixel rendering, as part of Quartz 2D. However, it was removed after the introduction of Retina displays. Unlike Microsoft's implementation, which favors a tight fit to the grid (font hinting) to maximize legibility, Apple's implementation prioritizes the shape of the glyphs as set out by their designer.
Read more →
Thomas Bolander

Thomas Bolander is a Danish professor at DTU Compute, Technical University of Denmark, where he studies logic and artificial intelligence. Most of his studies focus on the social aspect of artificial intelligence, and how we can make future AI able to navigate in social interactions. Thomas Bolander also sits in different commissions, expert panels and boards, among these he is a member of the Siri Commission, the TeckDK Commission, a member of the editorial board of the journal Studia Logica and co-organizer of Science and Cocktails. Bolander is known for his dissemination of science. In 2019 he was awarded the H. C. Ørsted Medal. Which he was the first to achieve after a break of three years.
Read more →
ELVIS Act

The ELVIS Act or Ensuring Likeness Voice and Image Security Act, signed into law by Tennessee Governor Bill Lee on March 21, 2024, marked a significant milestone in the area of regulation of artificial intelligence and public sector policies for artists in the era of artificial intelligence (AI) and AI alignment. It was noted as the first enacted legislation in the United States specifically designed to protect musicians from the unauthorized use of their voices through artificial intelligence technologies and against audio deepfakes and voice cloning. This legislation distinguishes itself by adding penalties for copying a performer's voice. == Origin and advocacy == The inception of the ELVIS Act has been attributed to Gebre Waddell, founder of Sound Credit, who initially conceptualized a framework in 2023 that later evolved into the legislation. Representative Justin J. Pearson acknowledged Waddell's pivotal role during the March 4 House Floor Session on the bill. Leading Tennessee musicians supported the ELVIS Act. Tennessee Governor Bill Lee endorsed it as a Governor's Bill, and it was introduced in the Tennessee Legislature as House Bill 2091 by William Lamberth (R-44) and Senate Bill 2096 by Jack Johnson (R-27). The ELVIS Act is an amendment to a 1984 law that was the result of the Elvis Presley estate litigation for controlling how his likeness could be used after death. == Lobbying from the recording industry == The legislative journey of the ELVIS Act included a broad coalition of music industry stakeholders, including: These organizations, led by the Recording Academy and the RIAA, played roles in drafting the legislation, advocating for passage, and rallying support among the industry and legislators. The act gained momentum through discussions that bridged industry concerns with legislative action. This collaborative process led to a proposal that specifically targets the use of AI to create unauthorized reproductions of artists' voices and images. == Opposition == The ELVIS Act saw industry opposition from the Motion Picture Association, including testimony in the House Banking & Consumer Affairs Subcommittee, including remarks that the law risks "interference with our members’ ability to portray real people and events." TechNet, representing companies such as OpenAI, Google and Amazon, expressed their opposition in the hearing to the bill as drafted, asserting that the language was too broadly written and could have unintended consequences. Other concerns included its potential application to cover bands, but lawmakers assured people that this was not the intention. The bill passed the Tennessee House and Senate with a unanimous, bi-partisan vote including 93 ayes and 0 Noes in the House, and 30 ayes and 0 noes in the Senate. == Passage == By explicitly addressing AI impersonation, the ELVIS Act originated a legal approach to safeguarding personal rights, in the context of digital and technological advancements. It extends protections to an artist's voice and likeness, areas vulnerable to exploitation with the proliferation of AI technologies that occurred in 2023. The legislation received widespread support from the music industry, signaling a significant step forward in the ongoing effort to balance innovation with the protection of individual rights and creative integrity. It was reported as underscoring Tennessee's commitment to its musical heritage and showed the state as a leader in adapting copyright and privacy protections to the modern technological landscape. Artists including Chris Janson and Luke Bryan appeared at the signing ceremony hosted at Robert's Western World to support the new law and commemorate its passing. == Legal precedent == The ELVIS Act was reported as representing a development in the discourse surrounding AI, intellectual property, and personal rights. It was hoped by proponents to set a precedent for future legislative efforts both within and beyond Tennessee, offering a model for how states and potentially the federal government could address similar challenges. As AI technology continues to evolve, the act represents a foundational framework for protecting the authenticity and rights of artists, ensuring contributions remain protected. The act prohibits usage of AI to clone the voice of an artist without consent and can be criminally enforced as a Class A misdemeanor. This legislation's success was hoped by its supporters to inspire similar actions in other states, contributing to a unified approach to copyright and privacy in the digital age. Such a national response would reinforce the importance of safeguarding artists' rights against unauthorized use of their voices and likenesses.
Read more →
Retrieval-based Voice Conversion

Retrieval-based Voice Conversion (RVC) is an open source voice conversion AI algorithm that enables realistic speech-to-speech transformations, accurately preserving the intonation and audio characteristics of the original speaker. == Overview == In contrast to text-to-speech systems such as ElevenLabs, RVC differs by providing speech-to-speech outputs instead. It maintains the modulation, timbre and vocal attributes of the original speaker, making it suitable for applications where emotional tone is crucial. The algorithm enables both pre-processed and real-time voice conversion with low latency. This real-time capability marks a significant advancement over previous AI voice conversion technologies, such as So-vits SVC. Its speed and accuracy have led many to note that its generated voices sound near-indistinguishable from "real life", provided that sufficient computational specifications and resources (e.g., a powerful GPU and ample RAM) are available when running it locally and that a high-quality voice model is used. == Technical foundation == Retrieval-based Voice Conversion (RVC) utilizes a hybrid approach that integrates feature extraction with retrieval-based synthesis. Instead of directly mapping source speaker features to the target speaker using statistical models, RVC retrieves relevant segments from a target speech database, aiming to enhance the naturalness and speaker fidelity of the converted speech. At a high level, the RVC system typically comprises three main components: (1) a content feature extractor, such as a phonetic posteriorgram (PPG) encoder or self-supervised models like HuBERT; (2) a vector retrieval module that searches a target voice database for the most similar speech units; and (3) a vocoder or neural decoder that synthesizes waveform output from the retrieved representations. The retrieval-based paradigm aims to mitigate the oversmoothing effect commonly observed in fully neural sequence-to-sequence models, potentially leading to more expressive and natural-sounding speech. Furthermore, with the incorporation of high-dimensional embeddings and k-nearest-neighbor search algorithms, the model can perform efficient matching across large-scale databases without significant computational overhead. Recent RVC frameworks have incorporated adversarial learning strategies and GAN-based vocoders, such as HiFi-GAN, to enhance synthesis quality. These integrations have been shown to produce clearer harmonics and reduce reconstruction errors. == Research developments == Research on RVC has recently explored the use of self-supervised learning (SSL) encoders such as wav2vec 2.0 and HuBERT to replace hand-engineered features like MFCCs. These encoders improve content preservation, especially when source and target speakers have dissimilar speaking styles or accents. Moreover, modern RVC models leverage vector quantization methods to discretize the acoustic space, improving both synthesis accuracy and generalization across unseen speakers. For example, retrieval-augmented VQ models can condition the synthesis stage on quantized speech tokens, which enhances controllability and style transfer. Despite its strengths, RVC still faces limitations related to database coverage, especially in real-time or few-shot settings. Inadequate diversity in the target voice corpus may lead to suboptimal retrieval or unnatural prosody. These advances demonstrate the viability of RVC as a strong alternative to conventional deep learning VC systems, balancing both flexibility and efficiency in diverse voice synthesis applications. == Training process == The training pipeline for retrieval-based voice conversion typically includes a preprocessing step where the target speaker's dataset is segmented and normalized. A pitch extractor such as librosa or DDSP-DDC may be used to obtain fundamental frequency (F0) features. During training, the model learns to map content features from the source speaker to the acoustic representation of the target speaker while maintaining pitch and prosody. The training objective often combines reconstruction loss with feature consistency loss across intermediate layers, and may incorporate cycle consistency loss to preserve speaker identity. Fine-tuning on small datasets is feasible due to the use of pre-trained models, particularly for the SSL encoder and content extractor components. This approach allows transfer learning to be applied effectively, enabling the model to converge faster and generalize better to unseen inputs. Most open implementations support batch training, gradient accumulation, and mixed-precision acceleration (e.g., FP16), especially when utilizing NVIDIA CUDA-enabled GPUs. == Real-time deployment == RVC systems can be deployed in real-time scenarios through WebUI interfaces and streaming audio frameworks. Optimizations include converting the inference graph to ONNX or TensorRT formats, reducing latency. Audio buffers are typically processed in chunks of 0.2–0.5 seconds to ensure minimal delay and seamless conversion. Cross-platform compatibility with tools such as OBS Studio and Voicemeeter enables integration into live streaming, video production, or virtual avatar environments. == Applications and concerns == The technology enables voice changing and mimicry, allowing users to create accurate models of others using only a negligible amount of minutes of clear audio samples. These voice models can be saved as .pth (PyTorch) files. While this capability facilitates numerous creative applications, it has also raised concerns about potential misuse as deepfake software for identity theft and malicious impersonation through voice calls. == Ethical and legal considerations == As with other deep generative models, the rise of RVC technology has led to increasing debate about copyright, consent, and authorship. While some jurisdictions may allow parody or fair use in creative contexts, impersonating living individuals without permission may infringe upon privacy and likeness rights. As a result, some platforms have begun issuing takedown notices against AI-generated voice content that closely mimics celebrities or musicians. === In pop culture === RVC inference has been used to create realistic depictions of song covers, such as replacing original vocals with characters like Twilight Sparkle and Mordecai to have them sing duets of popular music like "Airplanes" and "Somebody That I Used to Know." These AI-generated covers, which can sound strikingly similar to the voice imitated, have gained popularity on platforms like YouTube as humorous memes.
Read more →
Blanking (video)

In analog video, blanking occurs between horizontal lines and between frames. In raster scan equipment, an image is built up by scanning an electron beam from left to right across a screen to produce a visible trace of one scan line, reducing the brightness of the beam to zero (horizontal blanking), moving it back as fast as possible to the left of the screen at a slightly lower position (the next scan line), restoring the brightness, and continuing until all the lines have been displayed and the beam is at the bottom right of the screen. Its intensity is then reduced to zero again (vertical blanking), and it is rapidly moved to the top left to start again, creating the next frame. In television, in particular, the vertical blanking interval is long to accommodate the slow equipment available at the time the standard was set. Fast modern electronics allows digital information to be encoded into the signal during the vertical blanking interval; it is not displayed on screen as the beam is blanked, but can be processed by appropriate circuitry.
Read more →
Juergen Pirner

Juergen Pirner (born 1956) is the German creator of Jabberwock, a chatterbot that won the 2003 Loebner prize. Pirner created Jabberwock modelling the Jabberwocky from Lewis Carroll's poem of the same name. Initially, Jabberwock would just give rude or fantasy-related answers; but over the years, Pirner has programmed better responses into it. As of 2007 he has taught it 2.7 million responses. Pirner lives in Hamburg, Germany.
Read more →
Chris Olah

Christopher Olah (born 1992 or 1993) is a Canadian machine learning researcher and a co-founder of Anthropic. He is known for his work on neural network interpretability, particularly mechanistic interpretability, and for research and tools that visualise internal representations in neural networks. In 2025, Forbes reported he had become a billionaire due to his ownership in Anthropic. == Early life and education == Olah was born in Canada. According to Wired, he left university at age 18 without earning a degree and later received a Thiel Fellowship, which supported him in pursuing independent work. == Career == Olah has worked on interpretability research at Google Brain, OpenAI, and Anthropic. Time called him one of the pioneers of mechanistic interpretability and noted that he pursued this research line first at Google, then at OpenAI, and later at Anthropic, which he co-founded. Wired reported that Olah was involved in neural network visualisation work including DeepDream in 2015, as part of efforts to better understand what neural networks learn. Later coverage linked him to more structured interpretability approaches such as "activation atlases". The Verge covered activation atlases as a collaboration between Google and OpenAI researchers to help inspect neural network representations. At Anthropic, Olah has been identified in major press coverage as leading interpretability work aimed at mapping internal "features" in large language models and relating interpretability findings to AI safety. Quanta Magazine has also quoted Olah in reporting on interpretability and the internal structure of modern language models. Time included Olah in its TIME100 AI list in 2024. === Vatican address on AI ethics === On May 25, 2026, Olah spoke at the Vatican during the official presentation of Magnifica Humanitas, the first encyclical of Pope Leo XIV, which addresses artificial intelligence and human dignity. Olah said AI could lead to large-scale displacement of human labor and exacerbate global inequality. He said the commercial and geopolitical incentives driving frontier AI labs often conflict with the public good, and described AI systems as "grown" rather than strictly engineered. Olah called for external moral oversight from religious institutions, scholars, and civil society to hold the technology sector accountable.
Read more →
Knowledge Engineering Environment

Knowledge Engineering Environment (KEE) is a frame-based development tool for expert systems. It was developed and sold by IntelliCorp, and was first released in 1983. It ran on Lisp machines, and was later ported to Lucid Common Lisp with the CLX library, an X Window System (X11) interface for Common Lisp. This version was available on several different UNIX workstations. On KEE, several extensions were offered: Simkit, a frame-based simulation library KEEconnection, database connection between the frame system and relational databases In KEE, frames are called units. Units are used for both individual instances and classes. Frames have slots and slots have facets. Facets can describe, for example, a slot's expected values, its working value, or its inheritance rule. Slots can have multiple values. Behavior can be implemented using a message passing model. KEE provides an extensive graphical user interface (GUI) to create, browse, and manipulate frames. KEE also includes a frame-based rule system. In the KEE knowledge base, rules are frames. Both forward chaining and backward chaining inference are available. KEE supports non-monotonic reasoning through the concepts of worlds. Worlds allow providing alternative slot-values of frames. Through an assumption-based truth or reason maintenance system, inconsistencies can be detected and analyzed. ActiveImages allows graphical displays to be attached to slots of Units. Typical examples are buttons, dials, graphs, and histograms. The graphics are also implemented as Units via KEEPictures, a frame-based graphics library.
Read more →
Screen space directional occlusion

Screen space directional occlusion (SSDO) is a computer graphics technique enhancing screen space ambient occlusion (SSAO) by taking direction into account to sample the ambient light (both the light coming directly at an object, as well as the light reflected off of the object directly behind it), to better approximate global illumination. SSDO was introduced by Tobias Ritschel, Thorsten Grosch, and Hans-Peter Seidel in their 2009 ACM Symposium on Interactive 3D Graphics and Games paper Approximating dynamic global illumination in image space, which describes it as extending SSAO to directional occlusion with one diffuse indirect bounce of light; later literature notes that SSDO still suffers from common screen-space artifacts such as noise and banding. == Method == The original SSDO paper describes a two-pass screen-space approach, with one pass for direct lighting and a second pass for indirect bounces. Later literature describes SSDO as assuming a general shadowing direction that allows color bleeding and a single light bounce.
Read more →
Cerebellar model articulation controller

The cerebellar model arithmetic computer (CMAC) is a type of neural network based on a model of the mammalian cerebellum. It is also known as the cerebellar model articulation controller. It is a type of associative memory. The CMAC was first proposed as a function modeler for robotic controllers by James Albus in 1975 (hence the name), but has been extensively used in reinforcement learning and also as for automated classification in the machine learning community. The CMAC is an extension of the perceptron model. It computes a function for n {\displaystyle n} input dimensions. The input space is divided up into hyper-rectangles, each of which is associated with a memory cell. The contents of the memory cells are the weights, which are adjusted during training. Usually, more than one quantisation of input space is used, so that any point in input space is associated with a number of hyper-rectangles, and therefore with a number of memory cells. The output of a CMAC is the algebraic sum of the weights in all the memory cells activated by the input point. A change of value of the input point results in a change in the set of activated hyper-rectangles, and therefore a change in the set of memory cells participating in the CMAC output. The CMAC output is therefore stored in a distributed fashion, such that the output corresponding to any point in input space is derived from the value stored in a number of memory cells (hence the name associative memory). This provides generalisation. == Building blocks == In the adjacent image, there are two inputs to the CMAC, represented as a 2D space. Two quantising functions have been used to divide this space with two overlapping grids (one shown in heavier lines). A single input is shown near the middle, and this has activated two memory cells, corresponding to the shaded area. If another point occurs close to the one shown, it will share some of the same memory cells, providing generalisation. The CMAC is trained by presenting pairs of input points and output values, and adjusting the weights in the activated cells by a proportion of the error observed at the output. This simple training algorithm has a proof of convergence. It is normal to add a kernel function to the hyper-rectangle, so that points falling towards the edge of a hyper-rectangle have a smaller activation than those falling near the centre. One of the major problems cited in practical use of CMAC is the memory size required, which is directly related to the number of cells used. This is usually ameliorated by using a hash function, and only providing memory storage for the actual cells that are activated by inputs. == One-step convergent algorithm == Initially least mean square (LMS) method is employed to update the weights of CMAC. The convergence of using LMS for training CMAC is sensitive to the learning rate and could lead to divergence. In 2004, a recursive least squares (RLS) algorithm was introduced to train CMAC online. It does not need to tune a learning rate. Its convergence has been proved theoretically and can be guaranteed to converge in one step. The computational complexity of this RLS algorithm is O(N3). == Hardware implementation infrastructure == Based on QR decomposition, an algorithm (QRLS) has been further simplified to have an O(N) complexity. Consequently, this reduces memory usage and time cost significantly. A parallel pipeline array structure on implementing this algorithm has been introduced. Overall by utilizing QRLS algorithm, the CMAC neural network convergence can be guaranteed, and the weights of the nodes can be updated using one step of training. Its parallel pipeline array structure offers its great potential to be implemented in hardware for large-scale industry usage. == Continuous CMAC == Since the rectangular shape of CMAC receptive field functions produce discontinuous staircase function approximation, by integrating CMAC with B-splines functions, continuous CMAC offers the capability of obtaining any order of derivatives of the approximate functions. == Deep CMAC == In recent years, numerous studies have confirmed that by stacking several shallow structures into a single deep structure, the overall system could achieve better data representation, and, thus, more effectively deal with nonlinear and high complexity tasks. In 2018, a deep CMAC (DCMAC) framework was proposed and a backpropagation algorithm was derived to estimate the DCMAC parameters. Experimental results of an adaptive noise cancellation task showed that the proposed DCMAC can achieve better noise cancellation performance when compared with that from the conventional single-layer CMAC. == Summary ==
Read more →
Mojo (programming language)

Mojo is an in-development proprietary programming language based on Python available for Linux and macOS. Mojo aims to combine the usability of a high-level programming language, specifically Python, with the performance of a system programming language such as C++, Rust, and Zig. As of October 2025, the Mojo compiler is closed source with an open source standard library. Modular, the company behind Mojo, has stated an intent to open source the Mojo language, committing to open-source Mojo in "fall 2026". Mojo builds on the Multi-Level Intermediate Representation (MLIR) compiler software framework, instead of directly on the lower level LLVM compiler framework like many languages such as Julia, Swift, C++, and Rust. MLIR is a newer compiler framework that allows Mojo to exploit higher level compiler passes unavailable in LLVM alone, and allows Mojo to compile down and target more than only central processing units (CPUs), including producing code that can run on graphics processing units (GPUs), Tensor Processing Units (TPUs), application-specific integrated circuits (ASICs) and other accelerators. It can also often more effectively use certain types of CPU optimizations directly, like single instruction, multiple data (SIMD) with minor intervention by a developer, as occurs in many other languages. According to Jeremy Howard of fast.ai, Mojo can be seen as "syntax sugar for MLIR" and for that reason Mojo is well optimized for applications like artificial intelligence (AI). == Origin and development history == The Mojo programming language was created by Modular Inc, which was founded by Chris Lattner, the original architect of the Swift programming language and LLVM, and Tim Davis, a former Google employee. The intention behind Mojo is to bridge the gap between Python’s ease of use and the fast performance required for cutting-edge AI applications. According to public change logs, Mojo development goes back to 2022. In May 2023, the first publicly testable version was made available online via a hosted playground. By September 2023 Mojo was available for local download for Linux and by October 2023 it was also made available for download on Apple's macOS. In March 2024, Modular open sourced the Mojo standard library and started accepting community contributions under the Apache 2.0 license. == Features == Mojo was created for an easy transition from Python. The language has syntax similar to Python's, with inferred static typing, and allows users to import Python modules. It uses LLVM and MLIR as its compilation backend. The language also intends to add a foreign function interface to call C/C++ and Python code. The language is not source-compatible with Python 3, only providing a subset of its syntax, e.g. missing the global keyword, list and dictionary comprehensions, and support for classes. Further, Mojo also adds features that enable performant low-level programming: fn for creating typed, compiled functions and "struct" for memory-optimized alternatives to classes. Mojo structs support methods, fields, operator overloading, and decorators. The language also provides a borrow checker, an influence from Rust. Mojo def functions use value semantics by default (functions receive a copy of all arguments and any modifications are not visible outside the function), while Python functions use reference semantics (functions receive a reference on their arguments and any modification of a mutable argument inside the function is visible outside). The language is not currently open source, but it is planned to be made open source in the future. Modular has since committed to open-sourcing the Mojo language in "fall 2026". == Programming examples == In Mojo, functions can be declared using both fn (for performant functions) or def (for Python compatibility). Basic arithmetic operations in Mojo with a def function: and with an fn function: The manner in which Mojo employs var and let for mutable and immutable variable declarations respectively mirrors the syntax found in Swift. In Swift, var is used for mutable variables, while let is designated for constants or immutable variables. Variable declaration and usage in Mojo:
Read more →
Leabra

Leabra stands for local, error-driven and associative, biologically realistic algorithm. It is a model of learning which is a balance between Hebbian and error-driven learning with other network-derived characteristics. This model is used to mathematically predict outcomes based on inputs and previous learning influences. Leabra is heavily influenced by and contributes to neural network designs and models, including emergent. == Background == It is the default algorithm in emergent (successor of PDP++) when making a new project, and is extensively used in various simulations. Hebbian learning is performed using conditional principal components analysis (CPCA) algorithm with correction factor for sparse expected activity levels. Error-driven learning is performed using GeneRec, which is a generalization of the recirculation algorithm, and approximates Almeida–Pineda recurrent backpropagation. The symmetric, midpoint version of GeneRec is used, which is equivalent to the contrastive Hebbian learning algorithm (CHL). See O'Reilly (1996; Neural Computation) for more details. The activation function is a point-neuron approximation with both discrete spiking and continuous rate-code output. Layer or unit-group level inhibition can be computed directly using a k-winners-take-all (KWTA) function, producing sparse distributed representations. A feedforward and feedback (FFFB) form of inhibition has now replaced the KWTA form of inhibition. FFFB inhibition can be efficiently implemented by using the average excitatory input and activity levels in a given layer. The net input is computed as an average, not a sum, over connections, based on normalized, sigmoidally transformed weight values, which are subject to scaling on a connection-group level to alter relative contributions. Automatic scaling is performed to compensate for differences in expected activity level in the different projections. Documentation about this algorithm can be found in the book "Computational Explorations in Cognitive Neuroscience: Understanding the Mind by Simulating the Brain" published by MIT press. and in the Emergent Documentation Archived 2009-04-16 at the Wayback Machine == Overview of the leabra algorithm == The pseudocode for Leabra is given here, showing exactly how the pieces of the algorithm described in more detail in the subsequent sections fit together. Iterate over minus and plus phases of settling for each event. o At start of settling, for all units: - Initialize all state variables (activation, v_m, etc.). - Apply external patterns (clamp input in minus, input & output in plus). - Compute net input scaling terms (constants, computed here so network can be dynamically altered). - Optimization: compute net input once from all static activations (e.g., hard-clamped external inputs). o During each cycle of settling, for all non-clamped units: - Compute excitatory netinput (g_e(t), aka eta_j or net) -- sender-based optimization by ignoring inactives. - Compute kWTA inhibition for each layer, based on g_i^Q: Sort units into two groups based on g_i^Q: top k and remaining k+1 -> n. If basic, find k and k+1th highest If avg-based, compute avg of 1 -> k & k+1 -> n. Set inhibitory conductance g_i from g^Q_k and g^Q_k+1 - Compute point-neuron activation combining excitatory input and inhibition o After settling, for all units, record final settling activations as either minus or plus phase (y^-_j or y^+_j). After both phases update the weights (based on linear current weight values), for all connections: o Compute error-driven weight changes with CHL with soft weight bounding o Compute Hebbian weight changes with CPCA from plus-phase activations o Compute net weight change as weighted sum of error-driven and Hebbian o Increment the weights according to net weight change. == Implementations == Emergent Archived 2015-10-03 at the Wayback Machine is the original implementation of Leabra; its most recent implementation is written in Go. It was written chiefly by Dr. O'Reilly, but professional software engineers were recently hired to improve the existing codebase. This is the fastest implementation, suitable for constructing large networks. Although emergent has a graphical user interface, it is very complex and has a steep learning curve. If you want to understand the algorithm in detail, it will be easier to read non-optimized code. For this purpose, check out the MATLAB version. There is also an R version available, that can be easily installed via install.packages("leabRa") in R and has a short introduction to how the package is used. The MATLAB and R versions are not suited for constructing very large networks, but they can be installed quickly and (with some programming background) are easy to use. Furthermore, they can also be adapted easily. == Special algorithms == Temporal differences and general dopamine modulation. Temporal differences (TD) is widely used as a model of midbrain dopaminergic firing. Primary value learned value (PVLV). PVLV simulates behavioral and neural data on Pavlovian conditioning and the midbrain dopaminergic neurons that fire in proportion to unexpected rewards (an alternative to TD). Prefrontal cortex basal ganglia working memory (PBWM). PBWM uses PVLV to train prefrontal cortex working memory updating system, based on the biology of the prefrontal cortex and basal ganglia.
Read more →
JustWatch

JustWatch is a website that provides information on the availability of films and TV shows on various streaming platforms such as Netflix, HBO Max, Disney+, Hulu, Peacock, Fandango at Home, Apple TV, and Amazon Prime Video, among others. It is also available as a mobile application and smart TV application. JustWatch provides a search engine that allows users to discover which digital platforms host a particular movie or TV series. As of November 2023, JustWatch is available to users in 139 countries. == Features == JustWatch functions as a search engine by aggregating information about the online availability of films and TV series from video-on-demand streaming services. It aggregates information from more than 100 video content libraries, as well providing information about video resolution quality, pricing, and purchase or rental options. The website includes various filters for searching, including genre, price, release date, rating, and popularity. Users are also able to create lists of shows and movies and to share these lists with other users. == History == JustWatch GmbH is an international database company that is privately held and headquartered in Berlin, Germany. The company specializes in the online availability of movies and TV series. In addition to its user-facing website, the company also has an advertising-focused arm, JustWatch Media, that works with corporate clients, using data about what people watch that it gleans from user behavior to help entertainment companies tailor their marketing strategies. Its clients include Universal Pictures, Paramount Pictures, and Sony Pictures, among others. Development of the website began in 2014, and it was launched in the U.S. and Germany in February 2015. In 2018, the company received funding to improve databases within the European Union. In December 2019, the company acquired a rival streaming aggregation service, GoWatchIt, from Plexus Entertainment. JustWatch also used the acquisition to open its first New York office. In 2019, JustWatch had over 30 million users across 38 countries. By 2020, the company's streaming aggregation service was available in over 45 countries. By November 2023, it was available in 139 countries, and had over 40 million monthly users. === Founding === JustWatch was co-founded in 2013 by David Croyé, Cristoph Hoyer, Kevin Hiller, Dominik Raute, Ingke Weimert, and Michael Wilken. In a company blog post from February 2017, Croyé described the group of co-founders as all having previously "worked in leading roles at successful international tech-startups in Berlin." Croyé, who currently holds the title of CEO at JustWatch GmbH, had previously worked as the chief marketing officer at kaufDA, a European location-based mobile coupon and promotion service, and the background of other co-founders included time at the adtech company Trademob and the streaming site MyVideo. Startup capital for the website initially came from the founders themselves. Croyé in particular was able to reinvest funds he had obtained from the sale of kaufDA to Axel Springer, a European media company, in March 2011. Since 2015, the company has had at least one additional round of seed funding, with investors including venture capital groups CG Partners and STS Ventures.
Read more →
Cognitive tutor

A cognitive tutor is a particular kind of intelligent tutoring system that utilizes a cognitive model to provide feedback to students as they are working through problems. This feedback will immediately inform students of the correctness, or incorrectness, of their actions in the tutor interface; however, cognitive tutors also have the ability to provide context-sensitive hints and instruction to guide students towards reasonable next steps. == Introduction == The name of Cognitive Tutor now usually refers to a particular type of intelligent tutoring system produced by Carnegie Learning for high school mathematics based on John Anderson's ACT-R theory of human cognition. However, cognitive tutors were originally developed to test ACT-R theory for research purposes since the early 1980s and they are developed also for other areas and subjects such as computer programming and science. Cognitive Tutors can be implemented into classrooms as a part of blended learning that combines textbook and software activities. The Cognitive Tutor programs utilize cognitive model and are based on model tracing and knowledge tracing. Model tracing means that the cognitive tutor checks every action performed by students such as entering a value or clicking a button, while knowledge tracing is used to calculate the required skills students learned by measuring them on a bar chart called Skillometer. Model tracing and knowledge tracing are essentially used to monitor students' learning progress, guide students to correct path to problem solving, and provide feedback. The Institute of Education Sciences published several reports regarding the effectiveness of Carnegie Cognitive Tutor. A 2013 report concluded that Carnegie Learning Curricula and Cognitive Tutor was found to have mixed effects on mathematics achievement for high school students. The report identified 27 studies that investigate the effectiveness of Cognitive Tutor, and the conclusion is based on 6 studies that meet What Works Clearinghouse standards. Among the 6 studies included, 5 of them show intermediate to significant positive effect, while 1 study shows statistically significant negative effect. Another report published by Institute of Education Sciences in 2009 found that Cognitive Tutor Algebra I to have potentially positive effects on math achievement based on only 1 study out of 14 studies that meets What Works Clearinghouse standards. It should be understood that What Works Clearinghouse standards call for relatively large numbers of participants, true random assignments to groups, and for a control group receiving either no treatment or a different treatment. Such experimental conditions are difficult to meet in schools, and thus only a small percentage of studies in education meet the standards of this clearinghouse, even though they may still be of value. == Theoretical foundations == === Four-component architecture === Intelligent tutoring systems (ITS) have a four-component architecture: a domain model, a student model, a tutoring model and an interface component. The domain model contains the rules, concepts, and knowledge related to the domain to be learned. It helps to evaluate students' performance and detect students' errors by setting a standard of domain expertise. The student model, the central component of an ITS, is expected to contain knowledge about the students: their cognitive and affective states, and their progress as they learn. The function of the student model is threefold: to gather data from and about the learner, to represent the learner's knowledge and learning process, and to perform diagnostics of a student's knowledge and select optimal pedagogical strategies. The tutoring model uses the data gained from the domain model and student model to make decisions about tutoring strategies such as whether or not to intervene, or when and how to intervene. Functions of the tutoring model include instruction delivery and content planning. The interface component reflects the decisions made by the tutoring model in different forms such as Socratic dialogs, feedback and hints. Students interact with the tutor through the learning interface, also known as communication. The interface provides domain knowledge elements. === Cognitive model === A cognitive model replicates the domain knowledge and skills comparable to that of a human expert or an advanced student of the domain. A cognitive model enables intelligent tutoring systems to respond to problem-solving situations in a way similar to a human tutor. A tutoring system adopting a cognitive model is called a cognitive tutor. A cognitive model is an expert system that generates a multitude of solutions to the problems presented to students. The cognitive model is used to trace each student's solution through complex alternative solution paths, enabling the tutor to provide step-by-step feedback and advice, and to maintain a targeted model of the student's knowledge based on student performance. === Cognitive Tutors === Cognitive Tutors provide step-by-step guidance as a learner develops a complex problem-solving skill through practice. Typically, cognitive tutors provide such forms of support as: (a) a problem-solving environment that is designed rich and "thinking visible"; (b) step-by-step feedback on student performance; (c) feedback messages specific to errors; (d) context-specific next-step hints at student's request, and (e) individualized problem selection. Cognitive Tutors accomplish two of the principal tasks characteristic of human tutoring: (1) monitors the student's performance and providing context-specific individual instruction, and (2) monitors the student's learning and selects appropriate problem-solving activities. Both cognitive model and two underlying algorithms, model tracing and knowledge tracing, are used to monitor the student's learning. In model tracing, the cognitive tutor uses the cognitive model in complex problems to follow the student's individual path and provide prompt accuracy feedback and context-specific advice. In knowledge tracing, the cognitive tutor uses a Bayesian Knowledge Tracing method of evaluating the student's knowledge and uses this student model to select appropriate problems for each student. === Cognitive architecture === Cognitive tutor development is guided by ACT-R cognitive architecture, which specifies the underlying framework developing the cognitive model or expert component of a cognitive tutor. ACT-R, a member of the ACT family, is the most recent cognitive architecture, devoted primarily to modelling human behavior. ACT-R includes a declarative memory of factual knowledge and a procedural memory of production rules. The architecture functions by matching productions on perceptions and facts, mediated by the real-valued activation levels of objects, and executing them to affect the environment or alter declarative memory. ACT-R has been used to model psychological aspects such as memory, attention, reasoning, problem solving, and language processing. == Application and utilization == The first real world applications of cognitive tutors were in the 1980s and involved a geometry proof tutor used by high school students and a LISP programming tutor used by college students in a mini course in introductory programming course at Carnegie Mellon University. Since then, cognitive tutors have been used in a variety of scenarios, with a few organizations developing their own cognitive tutor programs. These programs have been used with students spanning elementary school through university level, though primarily in the subject areas of Computer Programming, Mathematics, and Science. One of the first organizations to develop a system for use within the school system was the PACT Center at Carnegie Mellon University. Their aim was to "...develop systems that provide individualized assistance to students as they work on challenging real-world problems in complex domains such as computer programming, algebra and geometry". PACT's most successful product was the Cognitive Tutor Algebra course. Originally created in the early 1990s, this course was in use in 75 schools through the U.S. by 1999, and then its spin-off company, Carnegie Learning, now offers tutors to thousands of schools in the U.S. The Carnegie Mellon Cognitive Tutor has been shown to raise students' math test scores in high school and middle-school classrooms, and their Algebra course was designated one of five exemplary curricula for K-12 mathematics educated by the US Department of Education. There were several research projects conducted by the PACT Center to utilize Cognitive tutor for courses in Excel and to develop an intelligent tutoring system for algebra expression writing, called Ms. Lindquist. Further, in 2005, Carnegie Learning released Bridge to Algebra, a product intended for middle schools that was piloted in over 100 schools. Cognitive tutoring software is continuing to be used.
Read more →
Script theory

Script theory is a psychological theory which posits that human behaviour largely falls into patterns called scripts because they function the way a written script does, by providing a program for action. Silvan Tomkins created script theory as a further development of his affect theory, which regards human beings' emotional responses to stimuli as falling into categories called affects: he noticed that the purely biological response of affect may be followed by awareness and by what we cognitively do in terms of acting on that affect, so that more was needed to produce a complete explanation of what he called human being theory. These scripts fall under the larger cognitive concept called schemas, which are organized chunks of information. A schema is a script that has the potential to lack the specificity of the sequence of events. A schema becomes a script is when there is an ordering to it that requires action, such as the process of starting a car (get in, put on the seatbelt, turn the car on, release the emergency brake, etc.). In script theory, the basic unit of analysis is called a scene, defined as a sequence of events linked by the affects triggered during the experience of those events. Tomkins recognized that affective experiences fall into patterns that we may group together according to criteria, such as the types of persons and places involved and the degree of intensity of the effect experienced—the patterns of which constitute scripts that inform behavior in an effort to maximize positive affect and to minimize negative affect. == In artificial intelligence == Roger Schank, Robert P. Abelson and their research group extended Tomkins' scripts and used them in early artificial intelligence work as a method of representing procedural knowledge. In their work, scripts are very much like frames, except the values that fill the slots must be ordered. A script is a structured representation describing a stereotyped sequence of events in a particular context. Scripts are used in natural-language understanding systems to organize a knowledge base in terms of the situations that the system should understand. The classic example of a script involves the typical sequence of events that occur when a person drinks in a restaurant: finding a seat, reading the menu, ordering drinks from the waitstaff, etc. In the script form, these would be decomposed into conceptual transitions, such as MTRANS and PTRANS, which refer to mental transitions [of information] and physical transitions [of things]. Schank, Abelson and their colleagues tackled some of the most difficult problems in artificial intelligence (i.e., story understanding), but ultimately their line of work ended without tangible success. This type of work received little attention after the 1980s, but became very influential in later knowledge representation techniques, such as case-based reasoning. Scripts can be inflexible. To deal with inflexibility, smaller modules called memory organization packets (MOP) can be combined in a way that is appropriate for the situation.
Read more →