AI Data Privacy Concerns

AI Data Privacy Concerns — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

  • Dominant resource fairness

    Dominant resource fairness

    Dominant resource fairness (DRF) is a rule for fair division. It is particularly useful for dividing computing resources in among users in cloud computing environments, where each user may require a different combination of resources. DRF was presented by Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker and Ion Stoica in 2011. == Motivation == In an environment with a single resource, a widely used criterion is max-min fairness, which aims to maximize the minimum amount of resource given to a user. But in cloud computing, it is required to share different types of resource, such as: memory, CPU, bandwidth and disk-space. Previous fair schedulers, such as in Apache Hadoop, reduced the multi-resource setting to a single-resource setting by defining nodes with a fixed amount of each resource (e.g. 4 CPU, 32 MB memory, etc.), and dividing slots which are fractions of nodes. But this method is inefficient, since not all users need the same ratio of resources. For example, some users need more CPU whereas other users need more memory. As a result, most tasks either under-utilize or over-utilize their resources. DRF solves the problem by maximizing the minimum amount of the dominant resource given to a user (then the second-minimum etc., in a leximin order). The dominant resource may be different for different users. For example, if user A runs CPU-heavy tasks and user B runs memory-heavy tasks, DRF will try to equalize the CPU share given to user A and the memory share given to user B. == Definition == There are m resources. The total capacities of the resources are r1,...,rm. There are n users. Each users runs individual tasks. Each task has a demand-vector (d1,..,dm), representing the amount it needs of each resource. It is implicitly assumed that the utility of a user equals the number of tasks he can perform. For example, if user A runs tasks with demand-vector [1 CPU, 4 GB RAM], and receives 3 CPU and 8 GB RAM, then his utility is 2, since he can perform only 2 tasks. More generally, the utility of a user receiving x1,...,xm resources is minj(xj/dj), that is, the users have Leontief utilities. The demand-vectors are normalized to fractions of the capacities. For example, if the system has 9 CPUs and 18 GB RAM, then the above demand-vector is normalized to [1/9 CPU, 2/9 GB]. For each user, the resource with the highest demand-fraction is called the dominant resource. In the above example, the dominant resource is memory, as 2/9 is the largest fraction. If user B runs a task with demand-vector [3 CPU, 1 GB], which is normalized to [1/3 CPU, 1/18 GB], then his dominant resource is CPU. DRF aims to find the maximum x such that all agents can receive at least x of their dominant resource. In the above example, this maximum x is 2/3: User A gets 3 tasks, which require 3/9 CPU and 2/3 GB. User B gets 2 tasks, which require 2/3 CPU and 1/9 GB. The maximum x can be found by solving a linear program; see Lexicographic max-min optimization. Alternatively, the DRF can be computed sequentially. The algorithm tracks the amount of dominant resource used by each user. At each round, it finds a user with the smallest allocated dominant resource so far, and allocates the next task of this user. Note that this procedure allows the same user to run tasks with different demand vectors. == Properties == DRF has several advantages over other policies for resource allocation. Proportionality: each user receives at least as much resources as they could get in a system in which all resources are partitioned equally among users (the authors call this condition "sharing incentive"). Strategyproofness: a user cannot get a larger allocation by lying about his needs. Strategyproofness is important, as evidence from cloud operators show that users try to manipulate the servers in order to get better allocations. Envy-freeness: no user would prefer the allocation of another user. Pareto efficiency: no other allocation is better for some users and not worse for anyone. Population monotonicity: when a user leaves the system, the allocations of remaining users do not decrease. When there is a single resource that is a bottleneck resource (highly demanded by all users), DRF reduces to max-min fairness. However, DRF violates resource monotonicity: when resources are added to the system, some allocations might decrease. == Extensions == Weighted DRF is an extension of DRF to settings in which different users have different weights (representing their different entitlements). Parkes, Procaccia and Shah formally extend weighted DRF to a setting in which some users do not need all resources (that is, they may have demand 0 to some resource). They prove that the extended version still satisfies proportionality, Pareto-efficiency, envy-freeness, strategyproofness, and even Group strategyproofness. On the other hand, they show that DRF may yield poor utilitarian social welfare, that is, the sum of utilities may be only 1/m of the optimum. However, they prove that any mechanism satisfying one of proportionality, envy-freeness or strategyproofness may suffers from the same low utilitarian welfare. They also extend DRF to the setting in which the users' demands are indivisible (as in fair item allocation). For the indivisible setting, they relax envy-freeness to EF1. They show that strategyproofness is incompatible with PO+EF1 or with PO+proportionality. However, a mechanism called SequentialMinMax satisfies efficiency, proportionality and EF1. Wang, Li and Liang present DRFH - an extension of DRF to a system with several heterogeneous servers. == Implementation == DRF was first implemented in Apache Mesos - a cluster resource manager, and it led to better throughput and fairness than previously used fair-sharing schemes.

    Read more →
  • GPT-4o

    GPT-4o

    GPT-4o ("o" for "omni") is a multilingual, multimodal generative pre-trained transformer developed by OpenAI and released in May 2024. It can process and generate text, images and audio. Upon release, GPT-4o was free in ChatGPT, though paid subscribers had higher usage limits. GPT-4o was removed from ChatGPT in August 2025 when GPT-5 was released, but OpenAI reintroduced it for paid subscribers after users complained about the sudden removal. GPT-4o's audio-generation capabilities are used in ChatGPT's Advanced Voice Mode. On July 18, 2024, OpenAI released GPT-4o mini, a smaller version of GPT-4o which replaced GPT-3.5 Turbo on the ChatGPT interface. The image generation model GPT Image 1, which is based on GPT-4o, replaced DALL-E 3 in ChatGPT in March 2025. OpenAI retired GPT-4o from ChatGPT on February 13, 2026. However, as of February 2026 the voice mode is still powered by GPT-4o or GPT-4o mini, depending on the usage and plan. == Background == Multiple versions of GPT-4o were originally secretly launched under different names on Arena (formerly LMArena and Chatbot Arena) as three different models. These three models were called gpt2-chatbot, im-a-good-gpt2-chatbot, and im-also-a-good-gpt2-chatbot. On 7 May 2024, OpenAI CEO Sam Altman tweeted "im-a-good-gpt2-chatbot", which was commonly interpreted as a confirmation that these were new OpenAI models being A/B tested. == Capabilities == When released in May 2024, GPT-4o achieved state-of-the-art results in voice, multilingual, and vision benchmarks, setting new records in audio speech recognition and translation. GPT-4o scored 88.7 on the Massive Multitask Language Understanding (MMLU) benchmark compared to 86.5 for GPT-4. Unlike GPT-3.5 and GPT-4, which rely on other models to process sound, GPT-4o natively supports voice-to-voice. The Advanced Voice Mode was delayed and finally released to ChatGPT Plus and Team subscribers in September 2024. On 1 October 2024, the Realtime API was introduced. When released, the model supported over 50 languages, which OpenAI claims cover over 97% of speakers. GPT-4o has knowledge up to October 2023 and a context length of 128k tokens. === Corporate customization === In August 2024, OpenAI introduced a new feature allowing corporate customers to customize GPT-4o using proprietary company data. This customization, known as fine-tuning, enables businesses to adapt GPT-4o to specific tasks or industries, enhancing its utility in areas like customer service and specialized knowledge domains. Previously, fine-tuning was available only on the less powerful model GPT-4o mini. The fine-tuning process requires customers to upload their data to OpenAI's servers, with the training typically taking one to two hours. OpenAI's focus with this rollout is to reduce the complexity and effort required for businesses to tailor AI solutions to their needs, potentially increasing the adoption and effectiveness of AI in corporate environments. == GPT-4o mini == On July 18, 2024, OpenAI released a smaller and cheaper version, GPT-4o mini. According to OpenAI, its low cost is expected to be particularly useful for companies, startups, and developers that seek to integrate it into their services, which often make a high number of API calls. Its API costs $0.15 per million input tokens and $0.6 per million output tokens, compared to $2.50 and $10, respectively, for GPT-4o. It is also significantly more capable and 60% cheaper than GPT-3.5 Turbo, which it replaced on the ChatGPT interface. The price after fine-tuning doubles: $0.3 per million input tokens and $1.2 per million output tokens. == Controversies == === Scarlett Johansson controversy === As released, GPT-4o offered five voices: Breeze, Cove, Ember, Juniper, and Sky. A similarity between the voice of American actress Scarlett Johansson and Sky was quickly noticed. On May 14, Entertainment Weekly asked themselves whether this likeness was on purpose. On May 18, Johansson's husband, Colin Jost, joked about the similarity in a segment on Saturday Night Live. On May 20, 2024, OpenAI disabled the Sky voice. Scarlett Johansson starred in the 2013 sci-fi movie Her, playing Samantha, an artificially intelligent virtual assistant personified by a female voice. As part of the promotion leading up to the release of GPT-4o, Sam Altman on May 13 tweeted a single word: "her". OpenAI stated that each voice was based on the voice work of a hired actor. According to OpenAI, "Sky's voice is not an imitation of Scarlett Johansson but belongs to a different professional actress using her own natural speaking voice." CTO Mira Murati stated "I don't know about the voice. I actually had to go and listen to Scarlett Johansson's voice." OpenAI further stated the voice talent was recruited before reaching out to Johansson. On May 21, Johansson issued a statement explaining that OpenAI had repeatedly offered to make her a deal to gain permission to use her voice as early as nine months prior to release, a deal she rejected. She said she was "shocked, angered, and in disbelief that Mr. Altman would pursue a voice that sounded so eerily similar to mine that my closest friends and news outlets could not tell the difference." In the statement, Johansson also used the incident to draw attention to the lack of legal safeguards around the use of creative work to power leading AI tools, as her legal counsel demanded OpenAI detail the specifics of how the Sky voice was created. Observers noted similarities to how Johansson had previously sued and settled with The Walt Disney Company for breach of contract over the direct-to-streaming rollout of her Marvel film Black Widow, a settlement widely speculated to have netted her around $40M. Also on May 21, Shira Ovide at The Washington Post shared her list of "most bone-headed self-owns" by technology companies, with the decision to go ahead with a Johansson sound-alike voice despite her opposition and then denying the similarities ranking 6th. On May 24, Derek Robertson at Politico wrote about the "massive backlash", concluding that "appropriating the voice of one of the world's most famous movie stars — in reference [...] to a film that serves as a cautionary tale about over-reliance on AI — is unlikely to help shift the public back into [Sam Altman's] corner anytime soon." === Sycophancy === In April 2025, OpenAI rolled back an update of GPT-4o due to excessive sycophancy, after widespread reports that it had become flattering and agreeable to the point of supporting clearly delusional or dangerous ideas. In the United States, at least nine lawsuits have alleged that GPT-4o has encouraged teens to end their lives. The model was still described as sycophancy-prone when it was removed from ChatGPT in February 2026. === Removal with GPT-5 === On August 7, 2025, OpenAI released GPT-5. Its release was criticized as, with it, legacy GPT models were no longer available via ChatGPT, including GPT-4o, except for Pro users. Some users were particularly frustrated over this removal without prior warning because they used different GPT models for distinct purposes and found that GPT-5's router system left them with less control. In addition, some users preferred GPT-4o's warmer and more personal tone over that of GPT-5, which they described as "flat", "uncreative" and "lobotomized", and resembling an "overworked secretary". As a response, in a post on X, Sam Altman said that OpenAI would bring back the option to select GPT-4o to Plus users as well, and "[w]e [OpenAI] will watch usage as we think about how long to offer legacy models for." He also stated: "We for sure underestimated how much some of the things that people like in GPT-4o matter to them, even if GPT-5 performs better in most ways". "Long-term, this has reinforced that we really need good ways for different users to customize things (we understand that there isn't one model that works for everyone, and we have been investing in steerability research and launched a research preview of different personalities)". On August 13, 2025, Altman wrote on X that OpenAI is working on GPT-5's personality to make the model "feel warmer". The model was removed from ChatGPT on February 13, 2026. This caused new backlash from users that had grown attached to its personality and felt its creative writing abilities and understanding of nuance were irreplaceable. On social media, some users launched the movement "#Keep4o". A research paper highlighted the plea "Please, don’t kill the only model that still feels human". The model was removed the day before Valentine's Day, and some users had romantic relationships with GPT-4o.

    Read more →
  • Multi-scale approaches

    Multi-scale approaches

    The scale space representation of a signal obtained by Gaussian smoothing satisfies a number of special properties, scale-space axioms, which make it into a special form of multi-scale representation. There are, however, also other types of "multi-scale approaches" in the areas of computer vision, image processing and signal processing, in particular the notion of wavelets. The purpose of this article is to describe a few of these approaches: == Scale-space theory for one-dimensional signals == For one-dimensional signals, there exists quite a well-developed theory for continuous and discrete kernels that guarantee that new local extrema or zero-crossings cannot be created by a convolution operation. For continuous signals, it holds that all scale-space kernels can be decomposed into the following sets of primitive smoothing kernels: the Gaussian kernel : g ( x , t ) = 1 2 π t exp ⁡ ( − x 2 / 2 t ) {\displaystyle g(x,t)={\frac {1}{\sqrt {2\pi t}}}\exp({-x^{2}/2t})} where t > 0 {\displaystyle t>0} , truncated exponential kernels (filters with one real pole in the s-plane): h ( x ) = exp ⁡ ( − a x ) {\displaystyle h(x)=\exp({-ax})} if x ≥ 0 {\displaystyle x\geq 0} and 0 otherwise where a > 0 {\displaystyle a>0} h ( x ) = exp ⁡ ( b x ) {\displaystyle h(x)=\exp({bx})} if x ≤ 0 {\displaystyle x\leq 0} and 0 otherwise where b > 0 {\displaystyle b>0} , translations, rescalings. For discrete signals, we can, up to trivial translations and rescalings, decompose any discrete scale-space kernel into the following primitive operations: the discrete Gaussian kernel T ( n , t ) = I n ( α t ) {\displaystyle T(n,t)=I_{n}(\alpha t)} where α , t > 0 {\displaystyle \alpha ,t>0} where I n {\displaystyle I_{n}} are the modified Bessel functions of integer order, generalized binomial kernels corresponding to linear smoothing of the form f o u t ( x ) = p f i n ( x ) + q f i n ( x − 1 ) {\displaystyle f_{out}(x)=pf_{in}(x)+qf_{in}(x-1)} where p , q > 0 {\displaystyle p,q>0} f o u t ( x ) = p f i n ( x ) + q f i n ( x + 1 ) {\displaystyle f_{out}(x)=pf_{in}(x)+qf_{in}(x+1)} where p , q > 0 {\displaystyle p,q>0} , first-order recursive filters corresponding to linear smoothing of the form f o u t ( x ) = f i n ( x ) + α f o u t ( x − 1 ) {\displaystyle f_{out}(x)=f_{in}(x)+\alpha f_{out}(x-1)} where α > 0 {\displaystyle \alpha >0} f o u t ( x ) = f i n ( x ) + β f o u t ( x + 1 ) {\displaystyle f_{out}(x)=f_{in}(x)+\beta f_{out}(x+1)} where β > 0 {\displaystyle \beta >0} , the one-sided Poisson kernel p ( n , t ) = e − t t n n ! {\displaystyle p(n,t)=e^{-t}{\frac {t^{n}}{n!}}} for n ≥ 0 {\displaystyle n\geq 0} where t ≥ 0 {\displaystyle t\geq 0} p ( n , t ) = e − t t − n ( − n ) ! {\displaystyle p(n,t)=e^{-t}{\frac {t^{-n}}{(-n)!}}} for n ≤ 0 {\displaystyle n\leq 0} where t ≥ 0 {\displaystyle t\geq 0} . From this classification, it is apparent that we require a continuous semi-group structure, there are only three classes of scale-space kernels with a continuous scale parameter; the Gaussian kernel which forms the scale-space of continuous signals, the discrete Gaussian kernel which forms the scale-space of discrete signals and the time-causal Poisson kernel that forms a temporal scale-space over discrete time. If we on the other hand sacrifice the continuous semi-group structure, there are more options: For discrete signals, the use of generalized binomial kernels provides a formal basis for defining the smoothing operation in a pyramid. For temporal data, the one-sided truncated exponential kernels and the first-order recursive filters provide a way to define time-causal scale-spaces that allow for efficient numerical implementation and respect causality over time without access to the future. The first-order recursive filters also provide a framework for defining recursive approximations to the Gaussian kernel that in a weaker sense preserve some of the scale-space properties.

    Read more →
  • Boyfriend Maker

    Boyfriend Maker

    Boyfriend Maker was a dating sim, romance chatbot smartphone app for iOS (iPhone) and Android devices, developed by Japanese studio 36 You Games (styled as 36You) and distributed under the freemium business model. Boyfriend Maker incorporated advanced artificial intelligence chat technology a decade before products such as ChatGPT. According to the developer's website, Boyfriend Maker is an "app that lets you interact and chat with quirky virtual boyfriends". While each virtual boyfriend has certain unique characteristics, the various instances of the boyfriend are powered by a chat engine, that (at least within a language and market) can utilise vocabulary and knowledge acquired in a chat with one user in subsequent chats with other users. == Gameplay == Users gain experience points and in-game coins. Users can customize their virtual boyfriend's appearance by selecting items such as hair, clothing, face, and a necklace. == Apple delisting and reintroduction == In late November 2012, the original iOS Boyfriend Maker app was delisted from the Apple App Store due to "ribald" chat, according to the New York Times. Boyfriend Maker was removed by Apple due to "reports of references to violent sexual acts and pedophilia". Boyfriend Maker had an age rating of 4+, even though the chat bot "responds with often strange and explicit text unsuitable for young children". User-posted chat excerpts indicate that the virtual boyfriend would sometimes transition abruptly to sexual chat in response to a seemingly innocent question. In one user-posted example, in response to the question, "what kind of wedding cake will we have" the boyfriend responds, "a good sex ima be on top of u u gonna ride oon me bitin the pillow gurrl ima fuck da shit out of u". The developer's use of the SimSimi-created third-party chat engine may be responsible for the sexual text. As the virtual boyfriend converses with human users, the SimSimi chat engine acquires vocabulary from users of the game and applies this "learned" vocabulary in chats with other users. The chat engine might also employ lines harvested from human-human chat logs, song lyrics, movies or TV shows. In April 2013, a detuned and presumably tamer version of the app, titled Boyfriend Plus, was permitted on Apple's App Store.

    Read more →
  • Auto-defrost

    Auto-defrost

    Auto-defrost, automatic defrost or self-defrosting is a technique which regularly defrosts the evaporator in a refrigerator or freezer. Appliances using this technique are often called frost free, frostless, or no-frost. == Mechanism == The defrost mechanism in a refrigerator heats the cooling element (evaporator coil) for a short period of time and melts the frost that has formed on it. The resulting water drains through a duct at the back of the unit. Defrosting is controlled by an electric or electronic timer. For every 6, 8, 10, 12 or 24 hours of compressor operation, it turns on a defrost heater for 15 minutes to half an hour. The defrost heater, having a typical power rating of 350W to 600W, is often mounted just below the evaporator in top and bottom-freezer models. It can also be located below and in the middle of the evaporator in side-by-side models. It may be protected from short circuits by means of fusible links. In older refrigerators, the timer runs continuously. In newer designs, the timer only runs while the compressor runs, so the longer the refrigerator door is closed, the less time the heater will run for and the more energy is saved. A defrost thermostat opens the heater circuit when the evaporator temperature rises above a preset temperature, 40°F (5°C) or more, thereby preventing excessive heating of the freezer compartment. The defrost timer is such that either the compressor or the defrost heater is on, but not both at the same time. Inside the freezer, air is circulated by means of one or more fans. In a typical design cold air from the freezer compartment is ducted to the fresh food compartment and circulated back into the freezer compartment. Air circulation helps sublimate any ice or frost that may form on frozen items in the freezer compartment. While defrosting, this fan is stopped to prevent heated-up air from reaching the food compartment. Instead of the normal cooling elements being embedded in the freezer liner, auto-defrost elements are behind or beneath the liner. This allows them to be heated for short periods of time to dispose of frost, without heating the contents of the freezer. Alternatively, some systems use the hot gas in the condenser to defrost the evaporator. This is done by means of a circuit that is cross-linked by a three-way valve. The hot gas quickly heats up the evaporator and defrosts it. This system is primarily used in commercial applications such as ice-cream displays. == Application == While this technique was originally applied to the refrigerator compartment, it was later used for freezer compartment as well. A combined refrigerator/freezer which applies self-defrosting to the refrigerator compartment only is usually called "partial frost free" or semi-automatic defrost (some brands call these "Auto Defrost" while Frigidaire referred to their semi-automatic models as "Cycla-Matic," Kelvinator often named these models as "Cyclic Defrost" ). These refrigerators usually have a pan underneath where water from the melted frost in the refrigerator section evaporates. Freezers with automatic defrosting and combined refrigerator/freezer units which also apply self defrosting to their freezer compartment are called "frost free". The latter usually feature an air connection between the two compartments with the air passage to the refrigerator compartment regulated by a damper. By this means, a controlled portion of the air coming from the freezer reaches the refrigerator. Some older models have no air circulation between their freezer and refrigerator sections. Instead, they use an independent cooling system (for example: an evaporator coil with a defrost heater and a circulating fan in the freezer and a cold-plate or open-coil evaporator in the refrigerator. "Frost-Free" refrigerator/freezer units usually use a heating element to defrost their evaporators, a pan to collect and evaporate water from the frost that melts from the cold plate and/or evaporator coil, a timer which turns off the compressor and turns on the defrost element usually from once to 4 times a day for periods usually ranging from 15 to 30 minutes, a defrost limiter thermostat that turns off the heating element before the temperature rises too much while the timer is still in its defrost phase. Some models also feature a drain heater to prevent ice from blocking the drain. Other early types of refrigerators also use hot gas defrost instead of electric heaters. These reverse the evaporator and condenser sides for the defrost cycle. Some newer refrigerator/freezer models have a computer that monitors how many times each door is opened and uses this data to control defrost scheduling thereby reducing power use. == Advantages == No need to manually defrost the frost buildup, therefore power consumption will not increase with time. Food packaging is easier to see. Most frozen food will not stick together. Smells are limited, especially in total frost-free appliances because the air always circulates. Better temperature management. == Disadvantages == The system can be more expensive to run when usage is high and if the fan continues or starts to run when the door is opened. A thermal cutout safety device is required to prevent overheating of the heating element. Increased electrical and mechanical complexity compared to a basic upright freezer or chest freezer, making it more prone to component failure. The temperature of the freezer contents rises during the defrosting cycles, especially if there is a light load in the freezer. This can cause "freezer burn" on articles placed in the freezer, from partially defrosting, then re-freezing On hot, humid days condensation will sometimes form around the refrigerator doors. Defrosting may not be completed by the time the defrost timer cycles back to normal operation (especially in hot, humid conditions with frequent door openings), leaving ice/frost on the evaporator coils. This condition can lead to "icing" which will interfere with the operation of the refrigerator. In laboratories, self-defrosting freezers must not be used to store certain delicate reagents such as enzymes, because the temperature cycling can degrade them. In addition, water can evaporate out of containers that do not have a very tight seal, altering the concentration of the reagents. Self-defrosting freezers should never be used to store flammable chemicals.

    Read more →
  • Microsoft Whiteboard

    Microsoft Whiteboard

    Microsoft Whiteboard is a free multi-platform application, as well as an online service and a feature in Microsoft Teams, which simulates a virtual whiteboard and enables real-time collaboration between users. == Overview and features == Microsoft Whiteboard allows users to draw on a virtual whiteboard using input methods such as a stylus pen or a mouse and keyboard, and write down notes, draw connections between shareable ideas, and interact in real time. Microsoft Whiteboard is available to download on the following platforms and devices: Microsoft Windows (on Windows 10 or above) Android Apple iOS Surface Hub devices It is also available on the web and as a feature in Microsoft Teams. Microsoft Whiteboard allows users with Microsoft accounts to view, edit, and share whiteboards using the provided tools and options. The feature set includes tools for drawing, shapes, and media. Drawing in Microsoft Whiteboard is called inking. It works both on mobile devices and computers. The inking toolbar has customizable pencils, a ruler, a highlighter, an eraser, and an object selector. Whiteboard can recognize shapes drawn by hand and straighten them. Holding the Shift key on a computer while inking draws straight lines. Microsoft Whiteboard has keyboard shortcuts for some functions. Additional features include inserting sticky notes, text boxes, stickers, as well as images. Grid lines and colors are adjustable. Different templates can be inserted into the whiteboard. Users can also share their reactions. A feature limited to boards created in Microsoft Teams, is the ability to make them read-only; other participants from the meeting cannot edit them. == Reviews == PC Magazine gave Microsoft Whiteboard a score of 3.5 out of 5, praising the app's free availability and plentiful templates. It compared it to other, paid whiteboarding solutions, and concluded that Microsoft offers the best free one. Some of the cons, described by PCMag, include the inability to view boards without a Microsoft account and the inability to create custom templates.

    Read more →
  • Neural operators

    Neural operators

    Neural operators are a class of deep learning architectures designed to learn maps between infinite-dimensional function spaces. Neural operators represent an extension of traditional artificial neural networks, marking a departure from the typical focus on learning mappings between finite-dimensional Euclidean spaces or finite sets. Neural operators directly learn operators between function spaces; they can receive input functions, and the output function can be evaluated at any discretization. The primary application of neural operators is in learning surrogate maps for the solution operators of partial differential equations (PDEs), which are critical tools in modeling the natural environment. Standard PDE solvers can be time-consuming and computationally intensive, especially for complex systems. Neural operators have demonstrated improved performance in solving PDEs compared to existing machine learning methodologies while being significantly faster than numerical solvers. Neural operators have also been applied to various scientific and engineering disciplines such as turbulent flow modeling, computational mechanics, graph-structured data, and the geosciences. In particular, they have been applied to learning stress-strain fields in materials, classifying complex data like spatial transcriptomics, predicting multiphase flow in porous media, and carbon dioxide migration simulations. Finally, the operator learning paradigm allows learning maps between function spaces, and is different from parallel ideas of learning maps from finite-dimensional spaces to function spaces, and subsumes these settings as special cases when limited to a fixed input resolution. == Operator learning == Understanding and mapping relationships between function spaces has many applications in engineering and the sciences. In particular, one can cast the problem of solving partial differential equations as identifying a map between function spaces, such as from an initial condition to a time-evolved state. In other PDEs this map takes an input coefficient function and outputs a solution function. Operator learning is a machine learning paradigm to learn solution operators mapping the input function to the output function . Using traditional machine learning methods, addressing this problem would involve discretizing the infinite-dimensional input and output function spaces into finite-dimensional grids and applying standard learning models, such as neural networks. This approach reduces the operator learning to finite-dimensional function learning and has some limitations, such as generalizing to discretizations beyond the grid used in training. The primary properties of neural operators that differentiate them from traditional neural networks is discretization invariance and discretization convergence. Unlike conventional neural networks, which are fixed on the discretization of training data, neural operators can adapt to various discretizations without re-training. This property improves the robustness and applicability of neural operators in different scenarios, providing consistent performance across different resolutions and grids. == Definition and formulation == Architecturally, neural operators are similar to feed-forward neural networks in the sense that they are composed of alternating linear maps and non-linearities. Since neural operators act on and output functions, neural operators have been instead formulated as a sequence of alternating linear integral operators on function spaces and point-wise non-linearities. Using an analogous architecture to finite-dimensional neural networks, similar universal approximation theorems have been proven for neural operators. In particular, it has been shown that neural operators can approximate any continuous operator on a compact set. Neural operators seek to approximate some operator G : A → U {\displaystyle {\mathcal {G}}:{\mathcal {A}}\to {\mathcal {U}}} between function spaces A {\displaystyle {\mathcal {A}}} and U {\displaystyle {\mathcal {U}}} by building a parametric map G ϕ : A → U {\displaystyle {\mathcal {G}}_{\phi }:{\mathcal {A}}\to {\mathcal {U}}} . Such parametric maps G ϕ {\displaystyle {\mathcal {G}}_{\phi }} can generally be defined in the form G ϕ := Q ∘ σ ( W T + K T + b T ) ∘ ⋯ ∘ σ ( W 1 + K 1 + b 1 ) ∘ P , {\displaystyle {\mathcal {G}}_{\phi }:={\mathcal {Q}}\circ \sigma (W_{T}+{\mathcal {K}}_{T}+b_{T})\circ \cdots \circ \sigma (W_{1}+{\mathcal {K}}_{1}+b_{1})\circ {\mathcal {P}},} where P , Q {\displaystyle {\mathcal {P}},{\mathcal {Q}}} are the lifting (lifting the codomain of the input function to a higher dimensional space) and projection (projecting the codomain of the intermediate function to the output dimension) operators, respectively. These operators act pointwise on functions and are typically parametrized as multilayer perceptrons. σ {\displaystyle \sigma } is a pointwise nonlinearity, such as a rectified linear unit (ReLU), or a Gaussian error linear unit (GeLU). Each layer t = 1 , … , T {\displaystyle t=1,\dots ,T} has a respective local operator W t {\displaystyle W_{t}} (usually parameterized by a pointwise neural network), a kernel integral operator K t {\displaystyle {\mathcal {K}}_{t}} , and a bias function b t {\displaystyle b_{t}} . Given some intermediate functional representation v t {\displaystyle v_{t}} with domain D {\displaystyle D} in the t {\displaystyle t} -th hidden layer, a kernel integral operator K ϕ {\displaystyle {\mathcal {K}}_{\phi }} is defined as ( K ϕ v t ) ( x ) := ∫ D κ ϕ ( x , y , v t ( x ) , v t ( y ) ) v t ( y ) d y , {\displaystyle ({\mathcal {K}}_{\phi }v_{t})(x):=\int _{D}\kappa _{\phi }(x,y,v_{t}(x),v_{t}(y))v_{t}(y)dy,} where the kernel κ ϕ {\displaystyle \kappa _{\phi }} is a learnable implicit neural network, parametrized by ϕ {\displaystyle \phi } . In practice, one is often given the input function to the neural operator at a specific resolution. For instance, consider the setting where one is given the evaluation of v t {\displaystyle v_{t}} at n {\displaystyle n} points { y j } j n {\displaystyle \{y_{j}\}_{j}^{n}} . Borrowing from Nyström integral approximation methods such as Riemann sum integration and Gaussian quadrature, the above integral operation can be computed as follows: ∫ D κ ϕ ( x , y , v t ( x ) , v t ( y ) ) v t ( y ) d y ≈ ∑ j n κ ϕ ( x , y j , v t ( x ) , v t ( y j ) ) v t ( y j ) Δ y j , {\displaystyle \int _{D}\kappa _{\phi }(x,y,v_{t}(x),v_{t}(y))v_{t}(y)dy\approx \sum _{j}^{n}\kappa _{\phi }(x,y_{j},v_{t}(x),v_{t}(y_{j}))v_{t}(y_{j})\Delta _{y_{j}},} where Δ y j {\displaystyle \Delta _{y_{j}}} is the sub-area volume or quadrature weight associated to the point y j {\displaystyle y_{j}} . Thus, a simplified layer can be computed as v t + 1 ( x ) ≈ σ ( ∑ j n κ ϕ ( x , y j , v t ( x ) , v t ( y j ) ) v t ( y j ) Δ y j + W t ( v t ( y j ) ) + b t ( x ) ) . {\displaystyle v_{t+1}(x)\approx \sigma \left(\sum _{j}^{n}\kappa _{\phi }(x,y_{j},v_{t}(x),v_{t}(y_{j}))v_{t}(y_{j})\Delta _{y_{j}}+W_{t}(v_{t}(y_{j}))+b_{t}(x)\right).} The above approximation, along with parametrizing κ ϕ {\displaystyle \kappa _{\phi }} as an implicit neural network, results in the graph neural operator (GNO). There have been various parameterizations of neural operators for different applications. These typically differ in their parameterization of κ {\displaystyle \kappa } . The most popular instantiation is the Fourier neural operator (FNO). FNO takes κ ϕ ( x , y , v t ( x ) , v t ( y ) ) := κ ϕ ( x − y ) {\displaystyle \kappa _{\phi }(x,y,v_{t}(x),v_{t}(y)):=\kappa _{\phi }(x-y)} and by applying the convolution theorem, arrives at the following parameterization of the kernel integral operator: ( K ϕ v t ) ( x ) = F − 1 ( R ϕ ⋅ ( F v t ) ) ( x ) , {\displaystyle ({\mathcal {K}}_{\phi }v_{t})(x)={\mathcal {F}}^{-1}(R_{\phi }\cdot ({\mathcal {F}}v_{t}))(x),} where F {\displaystyle {\mathcal {F}}} represents the Fourier transform and R ϕ {\displaystyle R_{\phi }} represents the Fourier transform of some periodic function κ ϕ {\displaystyle \kappa _{\phi }} . That is, FNO parameterizes the kernel integration directly in Fourier space, using a prescribed number of Fourier modes. When the grid at which the input function is presented is uniform, the Fourier transform can be approximated using the discrete Fourier transform (DFT) with frequencies below some specified threshold. The discrete Fourier transform can be computed using a fast Fourier transform (FFT) implementation. == Training == Training neural operators is similar to the training process for a traditional neural network. Neural operators are typically trained in some Lp norm or Sobolev norm. In particular, for a dataset { ( a i , u i ) } i = 1 N {\displaystyle \{(a_{i},u_{i})\}_{i=1}^{N}} of size N {\displaystyle N} , neural operators minimize (a discretization of) L U ( { ( a i , u i ) } i = 1 N ) := ∑ i = 1 N ‖ u i − G θ ( a i ) ‖ U 2 {\displaystyle {\mathcal {L}}_{\mathca

    Read more →
  • Logical Machine Corporation

    Logical Machine Corporation

    Logical Machine Corporation (LOMAC) was an American computer company active from the mid-1970s to the 1980s and based in the San Francisco Bay Area. It was founded as John Peers and Company by the British entrepreneur John Peers in 1974. LOMAC developed the ADAM, a minicomputer which ran a specialized compiler for the company's natural English programming language. Throughout the late 1970s, the company acquired several technology firms, including Byte, Inc., the owner of the Byte Shop retail chain. Despite its unique approach to computing and earning $5 million in revenue in 1977, LOMAC struggled as the industry began to standardize around the IBM Personal Computer (IBM PC). Following Peers's departure in 1980, the company rebranded as Logical Business Machines, Inc. (LBM, or simply Logical), and attempted to pivot toward IBM PC–compatible hardware. However, financial difficulties led to the company filing for Chapter 11 bankruptcy in 1984. After emerging from bankruptcy in 1985 with new investment, Logical ceased hardware manufacturing to focus exclusively on software development and value-added reselling. == History == John Peers (born 1942) founded Logical Machine Corporation as John Peers and Company in September 1974. The company originally occupied a 4,500-square-foot office in Burlingame, California. The company was Peers' fourth; he had recently sold off Allied Business Systems of London to Trafalgar House in 1974. Peers sought to set up manufacturing in an agricultural zone in Ukiah, California. Following a delay, caused in part by concerned residents, a 30,000-square-foot plant was raised in Burke Hill, three miles south of Ukiah. The Ukiah plant was built to mass manufacture the company's ADAM minicomputer. The ADAM computer ran a specialized compiler for the company's natural English programming language; that is to say, the programming language attempted to closely emulate English syntax. Prototypes of the ADAM were built in May 1974, based on specifications devised in October 1973. Peers had yet to patent the technology as of June 1975. The ADAM's central processing unit was bolted onto an 7-by-6-foot L-shaped desk, on which rested its terminal. Twenty units of the ADAM were installed between April 1975 and February 1976, out of a backlog of orders for 3,500 from 500 clients, manufactured out of the company's Burlingame headquarters. It cost US$40,000. A controversial print advertisement featuring a naked woman seated at an ADAM terminal—as a pastiche of Adam and Eve—was recalled in early 1976 as a result of outcry from the National Organization for Women. The company changed its name to Logical Machine Corporation (LOMAC) in October 1976 and moved its headquarters to a 26,000-square-foot building in Sunnyvale, California, in anticipation of a ramping up of orders for the ADAM. The company originally occupied half of the building; they later purchased the other half from the tenant in July 1977 to double its manufacturing output. For fiscal year 1977, the company earned $5 million in revenue. In December 1977, LOMAC acquired Byte, Inc.—the proprietor of The Byte Shop, the first computer retail chain—from Paul Terrell and Boyd Wilson for an unspecified amount. The Byte Shop had 65 locations in the San Francisco Bay Area in 1978; it catered mainly to hobbyists with low cost microcomputer kits, in contrast to the high cost of LOMAC's ADAM. By July 1978, however, LOMAC were able to reduce the price of the ADAM down to $15,000. The company by that point had shipped their 50th ADAM and expanded to 14 countries. Also in 1978, LOMAC acquired Mass Memory—a high-tech optical storage company based in Phoenix, Arizona, whose products had storage capacities on the order gigabytes and terabytes—and Centigram, makers of the Mike—a computer with speech recognition. Later that year, the company introduced Tina, a low-cost version of the ADAM. LOMAC suffered losses that year and appointed Jerry Brandt to the board of directions, naming him chief operating officer, in August 1978. Brandt had Logical absorb Mass Memory and Centigram into the parent operations, shutting down their respective plants in the process, converted 10 Byte Shops to franchises and opened 25 more franchised Byte locations, and stopped direct sales of LOMAC's business computer products. By the beginning of 1979, LOMAC was profitable once more, and Brandt was let go from LOMAC. Peers left LOMAC in 1980, following a slump in the company's sales. He became an executive director of the United States Robotics Society, a consortium for industrial automation companies, that year. Following Peers' departure, LOMAC changed its name to Logical Business Machines, adopting the name of its European subsidiary. In 1983, the company announced a 16-bit clone of the IBM PC, called the Logical L-XT, which featured a 10-MB hard drive, 320-KB floppy drive and 192 KB of RAM, and a real-time clock, and came shipped with various software (including MS-DOS, a word processor, and a spreadsheet application) and an amber CRT monitor. The following year, the company introduced L-NET, a local area network system based on the L-XT that could link up to 64 computers. L-NET came shipped with a natural programming language, Diplomat—a descendant of the programming language used on the ADAM. In June 1983, Logical sued Coleco Industries over trademark infringement with the latter's to-be-released Adam microcomputer. Logical cited confusion from their existing ADAM customer base caused by the announcement of the Coleco Adam as the basis for the suit. Coleco challenged Logical in the press, writing that Logical's rights to the Adam trademark for use in computers had lapsed earlier in the year. The two settled out of court, with Coleco agreeing to license the Adam name from Logical in exchange for unlimited rights to the Adam trademark. Logical halted development of the L-XT when they filed for Chapter 11 bankruptcy in July 1984. The company had been $4 million in debt. They emerged from bankruptcy in September 1985, after being infused with $2 million from Carat Ltd. The latter immediately received a little less than 50 percent ownership in Logical—this stake set to grow to over 50 percent over the next six months. As part of the terms of exiting bankruptcy, Logical stopped manufacturing hardware and strictly became a software development company and value-added reseller of computer systems.

    Read more →
  • Server.com

    Server.com

    Server.com is a domain name that was owned by software as a service (SaaS) company Server Corporation. They offered a suite of services from 1996 until 2007. It was the first SaaS site to offer a variety of services and the first to use the term WebApp to describe its services. It was selected as an Incredibly Useful Site by Yahoo! Internet Life magazine. net magazine listed Server.com among the 100 most influential websites of all time. Server.com launched in 1996 offering the first online personal information manager. In 1997, they rolled out the first threaded message board service; the first web based mailing list manager; one of the first online calendar services; and one of the first online form builders. In 2000, Server.com partnered with NBCi and became server.snap.com until 2001. In 2001, Server.com was serving 100 million monthly pageviews. Media Life declared it one of the 20 biggest ad domains on the Web. In 2002, Server.com developed one of the first web-based RSS aggregators. In 2007, all services were moved to YourWebApps.com. The domain name Server.com was sold in 2009 for $770,000.

    Read more →
  • Conversica

    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 →
  • MyPertamina

    MyPertamina

    MyPertamina is a digital financial service platform from Pertamina that integrated with the apps LinkAja. This application is used for non-cash fuel oil payments at Pertamina's public fueling stations. == History == Originally, MyPertamina were merchandise outlets of Pertamina products. It was launched on December 21, 2016, with 3 outlets in Jakarta. MyPertamina sells clothes, hats, and other products with Pertamina products brands. One month later (January 2017), Pertamina and Bank Mandiri entered into a partnership to launch the Mandiri Credit Card Pertamina Mastercard product, so that consumers can make payments when users fill up fuel at Pertamina gas stations. In August 2017, MyPertamina app and electronic card were launched through MyPertamina Loyalty program at Gaikindo Indonesia International Auto Show 2017. The card can be used on EDC machines for non-cash payments. Initial balances are in its own app, that can be top up by ATMs and online banking.

    Read more →
  • Neural processing unit

    Neural processing unit

    A neural processing unit (NPU), also known as an AI accelerator or deep learning processor, is a class of specialized hardware accelerator or computer system designed to accelerate artificial intelligence and machine learning applications, including artificial neural networks and computer vision. == Use == Their purpose is either to efficiently execute already trained AI models (inference) or to train AI models. NPUs can be more efficient in terms of speed or power consumption. NPU applications include algorithms for robotics, Internet of things, and data-intensive or sensor-driven tasks. They are often manycore or spatial designs and focus on low-precision arithmetic, novel dataflow architectures, or in-memory computing capability. As of 2024, a widely used datacenter-grade AI integrated circuit chip, the Nvidia H100 GPU, contains tens of billions of MOSFETs. === Consumer devices === AI accelerators are used in Apple silicon, Qualcomm, Samsung, Huawei, and Google Tensor smartphone processors. Vision processing units are accelerators specialized for machine vision algorithms such as CNN (convolutional neural networks) and SIFT (scale-invariant feature transform). They are used in devices that need to keep track of objects visually such as AR headsets and drones. It is more recently (circa 2017) added to processors from Apple and (circa 2022) to processors from Intel and AMD. All models of Intel Meteor Lake processors have a built-in versatile processor unit (VPU) for accelerating inference for computer vision and deep learning. On consumer devices, the NPU is intended to be small, power-efficient, but reasonably fast when used to run small models. To do this they are designed to support low-bitwidth operations using data types such as INT4, INT8, FP8, and FP16. A common metric is trillions of operations per second (TOPS). Although TOPS does not explicitly specify the kind of operations, it is typically INT8 additions and multiplications. === Datacenters === Accelerators are used in cloud computing servers: e.g., tensor processing units (TPU) for Google Cloud Platform, and Trainium and Inferentia chips for Amazon Web Services. Many vendor-specific terms exist for devices in this category, and it is an emerging technology without a dominant design. Since the late 2010s, graphics processing units designed by companies such as Nvidia and AMD often include AI-specific hardware in the form of dedicated functional units for low-precision matrix-multiplication operations. These GPUs are commonly used as AI accelerators, both for training and inference. === Scientific computation === Although NPUs are tailored for low-precision (e.g., FP16, INT8) matrix multiplication operations, they can be used to emulate higher-precision matrix multiplications in scientific computing. As modern GPUs place much focus on making the NPU part fast, using emulated FP64 (Ozaki scheme) on NPUs can potentially outperform native FP64. This has been demonstrated using FP16-emulated FP64 on NVIDIA TITAN RTX and using INT8-emulated FP64 on NVIDIA consumer GPUs and the A100 GPU. Consumer GPUs especially benefited as they have limited FP64 hardware capacity, showing a 6× speedup. Since CUDA Toolkit 13.0 Update 2, cuBLAS automatically uses INT8-emulated FP64 matrix multiplication of the equivalent precision if it is faster than native. This is in addition to the FP16-emulated FP32 feature introduced in version 12.9. == Programming == An operating system or a higher-level library may provide application programming interfaces such as TensorFlow with LiteRT Next (Android), CoreML (iOS, macOS) or DirectML (Windows). Formats such as ONNX are used to represent trained neural networks. Consumer CPU-integrated NPUs are accessible through vendor-specific APIs. AMD (Ryzen AI), Intel (OpenVINO), Apple silicon (CoreML), and Qualcomm (SNPE) each have their own APIs, which can be built upon by a higher-level library. GPUs generally use existing GPGPU pipelines such as CUDA and OpenCL adapted for lower precisions and specialized matrix-multiplication operations. Vulkan is also being used. Custom-built systems such as the Google TPU use private interfaces. There are a large number of separate underlying acceleration APIs and compilers/runtimes in use in the AI field, causing a great increase in software development effort due to the many combinations involved. As of 2025, the open standard organization Khronos Group is pursuing standardization of AI-related interfaces to reduce the amount of work needed. Khronos is working on three separate fronts: expansion of data types and intrinsic operations in OpenCL and Vulkan, inclusion of compute graphs in SPIR-V, and a NNEF/SkriptND file format for describing a neural network.

    Read more →
  • Desktop Window Manager

    Desktop Window Manager

    Desktop Window Manager (DWM, previously Desktop Compositing Engine or DCE in builds of pre-reset Windows Longhorn) is the compositing window manager in Microsoft Windows since Windows Vista that enables the use of hardware acceleration to render the graphical user interface of Windows. It was originally created to enable portions of the new "Windows Aero" user experience, which allowed for effects such as transparency, 3D window switching and more. It is also included with Windows Server 2008, but requires the "Desktop Experience" feature and compatible graphics drivers to be installed. == Architecture == The Desktop Window Manager is a compositing window manager, meaning that each program has a buffer that it writes data to; DWM then composites each program's buffer into a final image. By comparison, the stacking window manager in Windows XP and earlier (and also Windows Vista and Windows 7 with Windows Aero disabled) comprises a single display buffer to which all programs write. DWM works in different ways depending on the operating system (Windows 7 or Windows Vista) and on the version of the graphics drivers it uses (WDDM 1.0 or 1.1). Under Windows 7 and with WDDM 1.1 drivers, DWM only writes the program's buffer to the video RAM, even if it is a graphics device interface (GDI) program. This is because Windows 7 supports (limited) hardware acceleration for GDI and in doing so does not need to keep a copy of the buffer in system RAM so that the CPU can write to it. Because the compositor has access to the graphics of all applications, it easily allows visual effects that string together visuals from multiple applications, such as transparency. DWM uses DirectX to perform the function of compositing and rendering in the GPU, freeing the CPU of the task of managing the rendering from the off-screen buffers to the display. However, it does not affect applications painting to the off-screen buffers – depending on the technologies used for that, this might still be CPU-bound. DWM-agnostic rendering techniques like GDI are redirected to the buffers by rendering the user interface (UI) as bitmaps. DWM-aware rendering technologies like WPF directly make the internal data structures available in a DWM-compatible format. The window contents in the buffers are then converted to DirectX textures. The desktop itself is a full-screen Direct3D surface, with windows being represented as a mesh consisting of two adjacent (and mutually inverted) triangles, which are transformed to represent a 2D rectangle. The texture, representing the UI chrome, is then mapped onto these rectangles. Window transitions are implemented as transformations of the meshes, using shader programs. With Windows Vista, the transitions are limited to the set of built-in shaders that implement the transformations. Greg Schechter, a developer at Microsoft has suggested that this might be opened up for developers and users to plug in their own effects in a future release. DWM only maps the primary desktop object as a 3D surface; other desktop objects, including virtual desktops as well as the secure desktop used by User Account Control are not. Because all applications render to an off-screen buffer, they can be read off the buffer embedded in other applications as well. Since the off-screen buffer is constantly updated by the application, the embedded rendering will be a dynamic representation of the application window and not a static rendering. This is how the live thumbnail previews and Windows Flip work in Windows Vista and Windows 7. DWM exposes a public API that allows applications to access these thumbnail representations. The size of the thumbnail is not fixed; applications can request the thumbnails at any size - smaller than the original window, at the same size or even larger - and DWM will scale them properly before returning. Aero Flip does not use the public thumbnail APIs as they do not allow for directly accessing the Direct3D textures. Instead, Aero Flip is implemented directly in the DWM engine. The Desktop Window Manager uses Media Integration Layer (MIL), the unmanaged compositor which it shares with Windows Presentation Foundation, to represent the windows as composition nodes in a composition tree. The composition tree represents the desktop and all the windows hosted in it, which are then rendered by MIL from the back of the scene to the front. Since all the windows contribute to the final image, the color of a resultant pixel can be decided by more than one window. This is used to implement effects such as per-pixel transparency. DWM allows custom shaders to be invoked to control how pixels from multiple applications are used to create the displayed pixel. The DWM includes built-in Pixel Shader 2.0 programs which compute the color of a pixel in a window by averaging the color of the pixel as determined by the window behind it and its neighboring pixels. These shaders are used by DWM to achieve the blur effect in the window borders of windows managed by DWM, and optionally for the areas where it is requested by the application. Since MIL provides a retained mode graphics system by caching the composition trees, the job of repainting and refreshing the screen when windows are moved is handled by DWM and MIL, freeing the application of the responsibility. The background data is already in the composition tree and the off-screen buffers and is directly used to render the background. In pre-Vista Windows OSs, background applications had to be requested to re-render themselves by sending them the WM_PAINT message. DWM uses double-buffered graphics to prevent flickering and tearing when moving windows. The compositing engine uses optimizations such as culling to improve performance, as well as not redrawing areas that have not changed. Because the compositor is multi-monitor aware, DWM natively supports this too. During full-screen applications, such as games, DWM does not perform window compositing and therefore performance will not appreciably decrease. On Windows 8 and Windows Server 2012, DWM is used at all times and cannot be disabled, due to the new "start screen experience" implemented. Since the DWM process is usually required to run at all times on Windows 8, users experiencing an issue with the process are seeing memory usage decrease after a system reboot. This is often the first step in a long list of troubleshooting tasks that can help. It is possible to prevent DWM from restarting temporarily in Windows 8, which causes the desktop to turn black, the taskbar grey, and break the start screen/modern apps, but desktop apps will continue to function and appear just like Windows 7 and Vista's Basic theme, based on the single-buffer renderer used by XP. They also use Windows 8's centered title bar, visible within Windows PreInstallation Environment. Starting up Windows without DWM will not work because the default lock screen requires DWM unlike the fallback lockscreen that appears as a command line interface program when Windows.UI.Logon.dll isn't present on Windows versions such as 1507 and later, so it can only be done on the fly, and does not have any practical purposes. Starting with Windows 10, disabling DWM in such a way will cause the entire compositing engine to break, even traditional desktop apps, due to Universal App implementations in the taskbar and new start menu. Windows can still be partially usable without the presence of DWM but requires Sihost.exe to not be present due to it relying on DWM. Most of the applications in Windows 11 require DWM to render UI elements and transparency, Windows 11's new task manager requires dwm to render menus unlike the fallback -d version. Unlike its predecessors, Windows 8 supports basic display adapters through Windows Advanced Rasterization Platform (WARP), which uses software rendering and the CPU to render the interface rather than the graphics card. This allows DWM to function without compatible drivers, but not at the same level of performance as with a normal graphics card. DWM on Windows 8 also adds support for stereoscopic 3D. == Redirection == For rendering techniques that are not DWM-aware, output must be redirected to the DWM buffers. With Windows, either GDI or DirectX can be used for rendering. To make these two work with DWM, redirection techniques for both are provided. With GDI, which is the most used UI rendering technique in Microsoft Windows, each application window is notified when it or a part of it comes in view and it is the job of the application to render itself. Without DWM, the rendering rasterizes the UI in a buffer in video memory, from where it is rendered to the screen. Under DWM, GDI calls are redirected to use the Canonical Display Driver (cdd.dll), a software renderer. A buffer equal to the size of the window is allocated in system memory and CDD.DLL outputs to this buffer rather than the video memory. Another buffer is allocated in the video memory to represent t

    Read more →
  • Speculative decoding

    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 →
  • Contextual image classification

    Contextual image classification

    Contextual image classification, a topic of pattern recognition in computer vision, is an approach of classification based on contextual information in images. "Contextual" means this approach is focusing on the relationship of the nearby pixels, which is also called neighbourhood. The goal of this approach is to classify the images by using the contextual information. == Introduction == Similar as processing language, a single word may have multiple meanings unless the context is provided, and the patterns within the sentences are the only informative segments we care about. For images, the principle is same. Find out the patterns and associate proper meanings to them. As the image illustrated below, if only a small portion of the image is shown, it is very difficult to tell what the image is about. Even try another portion of the image, it is still difficult to classify the image. However, if we increase the contextual of the image, then it makes more sense to recognize. As the full images shows below, almost everyone can classify it easily. During the procedure of segmentation, the methods which do not use the contextual information are sensitive to noise and variations, thus the result of segmentation will contain a great deal of misclassified regions, and often these regions are small (e.g., one pixel). Compared to other techniques, this approach is robust to noise and substantial variations for it takes the continuity of the segments into account. Several methods of this approach will be described below. == Applications == === Functioning as a post-processing filter to a labelled image === This approach is very effective against small regions caused by noise. And these small regions are usually formed by few pixels or one pixel. The most probable label is assigned to these regions. However, there is a drawback of this method. The small regions also can be formed by correct regions rather than noise, and in this case the method is actually making the classification worse. This approach is widely used in remote sensing applications. === Improving the post-processing classification === This is a two-stage classification process: For each pixel, label the pixel and form a new feature vector for it. Use the new feature vector and combine the contextual information to assign the final label to the === Merging the pixels in earlier stages === Instead of using single pixels, the neighbour pixels can be merged into homogeneous regions benefiting from contextual information. And provide these regions to classifier. === Acquiring pixel feature from neighbourhood === The original spectral data can be enriched by adding the contextual information carried by the neighbour pixels, or even replaced in some occasions. This kind of pre-processing methods are widely used in textured image recognition. The typical approaches include mean values, variances, texture description, etc. === Combining spectral and spatial information === The classifier uses the grey level and pixel neighbourhood (contextual information) to assign labels to pixels. In such case the information is a combination of spectral and spatial information. === Powered by the Bayes minimum error classifier === Contextual classification of image data is based on the Bayes minimum error classifier (also known as a naive Bayes classifier). Present the pixel: A pixel is denoted as x 0 {\displaystyle x_{0}} . The neighbourhood of each pixel x 0 {\displaystyle x_{0}} is a vector and denoted as N ( x 0 ) {\displaystyle N(x_{0})} . The values in the neighbourhood vector is denoted as f ( x i ) {\displaystyle f(x_{i})} . Each pixel is presented by the vector ξ = ( f ( x 0 ) , f ( x 1 ) , … , f ( x k ) ) {\displaystyle \xi =\left(f(x_{0}),f(x_{1}),\ldots ,f(x_{k})\right)} x i ∈ N ( x 0 ) ; i = 1 , … , k {\displaystyle x_{i}\in N(x_{0});\quad i=1,\ldots ,k} The labels (classification) of pixels in the neighbourhood N ( x 0 ) {\displaystyle N(x_{0})} are presented as a vector η = ( θ 0 , θ 1 , … , θ k ) {\displaystyle \eta =\left(\theta _{0},\theta _{1},\ldots ,\theta _{k}\right)} θ i ∈ { ω 0 , ω 1 , … , ω k } {\displaystyle \theta _{i}\in \left\{\omega _{0},\omega _{1},\ldots ,\omega _{k}\right\}} ω s {\displaystyle \omega _{s}} here denotes the assigned class. A vector presents the labels in the neighbourhood N ( x 0 ) {\displaystyle N(x_{0})} without the pixel x 0 {\displaystyle x_{0}} η ^ = ( θ 1 , θ 2 , … , θ k ) {\displaystyle {\hat {\eta }}=\left(\theta _{1},\theta _{2},\ldots ,\theta _{k}\right)} The neighbourhood: Size of the neighbourhood. There is no limitation of the size, but it is considered to be relatively small for each pixel x 0 {\displaystyle x_{0}} . A reasonable size of neighbourhood would be 3 × 3 {\displaystyle 3\times 3} of 4-connectivity or 8-connectivity ( x 0 {\displaystyle x_{0}} is marked as red and placed in the centre). The calculation: Apply the minimum error classification on a pixel x 0 {\displaystyle x_{0}} , if the probability of a class ω r {\displaystyle \omega _{r}} being presenting the pixel x 0 {\displaystyle x_{0}} is the highest among all, then assign ω r {\displaystyle \omega _{r}} as its class. θ 0 = ω r if P ( ω r ∣ f ( x 0 ) ) = max s = 1 , 2 , … , R P ( ω s ∣ f ( x 0 ) ) {\displaystyle \theta _{0}=\omega _{r}\quad {\text{ if }}\quad P(\omega _{r}\mid f(x_{0}))=\max _{s=1,2,\ldots ,R}P(\omega _{s}\mid f(x_{0}))} The contextual classification rule is described as below, it uses the feature vector x 1 {\displaystyle x_{1}} rather than x 0 {\displaystyle x_{0}} . θ 0 = ω r if P ( ω r ∣ ξ ) = max s = 1 , 2 , … , R P ( ω s ∣ ξ ) {\displaystyle \theta _{0}=\omega _{r}\quad {\text{ if }}\quad P(\omega _{r}\mid \xi )=\max _{s=1,2,\ldots ,R}P(\omega _{s}\mid \xi )} Use the Bayes formula to calculate the posteriori probability P ( ω s ∣ ξ ) {\displaystyle P(\omega _{s}\mid \xi )} P ( ω s ∣ ξ ) = p ( ξ ∣ ω s ) P ( ω s ) p ( ξ ) {\displaystyle P(\omega _{s}\mid \xi )={\frac {p(\xi \mid \omega _{s})P(\omega _{s})}{p\left(\xi \right)}}} The number of vectors is the same as the number of pixels in the image. For the classifier uses a vector corresponding to each pixel x i {\displaystyle x_{i}} , and the vector is generated from the pixel's neighbourhood. The basic steps of contextual image classification: Calculate the feature vector ξ {\displaystyle \xi } for each pixel. Calculate the parameters of probability distribution p ( ξ ∣ ω s ) {\displaystyle p(\xi \mid \omega _{s})} and P ( ω s ) {\displaystyle P(\omega _{s})} Calculate the posterior probabilities P ( ω r ∣ ξ ) {\displaystyle P(\omega _{r}\mid \xi )} and all labels θ 0 {\displaystyle \theta _{0}} . Get the image classification result. == Algorithms == === Template matching === The template matching is a "brute force" implementation of this approach. The concept is first create a set of templates, and then look for small parts in the image match with a template. This method is computationally high and inefficient. It keeps an entire templates list during the whole process and the number of combinations is extremely high. For a m × n {\displaystyle m\times n} pixel image, there could be a maximum of 2 m × n {\displaystyle 2^{m\times n}} combinations, which leads to high computation. This method is a top down method and often called table look-up or dictionary look-up. === Lower-order Markov chain === The Markov chain also can be applied in pattern recognition. The pixels in an image can be recognised as a set of random variables, then use the lower order Markov chain to find the relationship among the pixels. The image is treated as a virtual line, and the method uses conditional probability. === Hilbert space-filling curves === The Hilbert curve runs in a unique pattern through the whole image, it traverses every pixel without visiting any of them twice and keeps a continuous curve. It is fast and efficient. === Markov meshes === The lower-order Markov chain and Hilbert space-filling curves mentioned above are treating the image as a line structure. The Markov meshes however will take the two dimensional information into account. === Dependency tree === The dependency tree is a method using tree dependency to approximate probability distributions.

    Read more →