We strive to inform readers not only about events in the bitcoin industry, but also to talk about what is happening in related technological fields - cybersecurity and in the world of artificial intelligence.
GitHub has blocked the accounts of Russian developers.
Tech giants have formed an alliance to fight deepfakes.
Google Lens has learned to recognize text in the PC version of Chrome.
Instagram's algorithms have trained us to "appreciate" original content more.
DeepMind has introduced a new language model with 70 billion parameters.
South Korea will build an underground city with its own data center.
The most important deals from the field of AI over the past week.
GitHub started blocking accounts of Russian developers
The IT project storage service GitHub has started blocking accounts of Russian developers and companies.
According to media reports, from April 13, 2022, the platform suspends without warning the work of the accounts of programmers who now or previously worked for organizations that have fallen under US sanctions.
As of April 15, GitHub blocked dozens of Russian accounts, including Sberbank AI Lab, Sberbank, Alfa-Bank and individual developers.
"When a corporate account is blocked, access to repository data can be maintained for two hours, when a personal account is stopped, it is limited instantly," Habr noted.
Tech giants formed an alliance to counter deepfakes
The C2PA Alliance of IT Companies will develop a standard for verifying the authenticity of images and videos on the Internet. Among the consortium members are Adobe, Microsoft, Intel, Twitter, Sony, Nikon, BBC and ARM.
According to the alliance, the introduction of an open standard will prevent the spread of deepfakes. Many of the authentication mechanisms will appear in 2022, and in a couple of years an ecosystem will be formed to control the veracity of photos and videos at all stages of their life cycle, Adobe said.
The technical standard will ensure that there is no interference with content using distributed ledger technology. Adobe claims that the alliance has a long way to go before all platforms can implement effective mechanisms for verifying the authenticity of materials.
TSMC to release 2nm chips in 2026
In 2026, TSMC will release the first chips based on a 2-nm process technology.
The company's executive director, Xi Xi Wei, said that the N2 production units will rely on isolated gate transistors (GAA). However, he did not provide any specific information about the architecture or the name of the technology.
It is also known that the manufacture of chips will rely on existing photolithography in extreme ultraviolet radiation with a numerical aperture of 0.33.
TSMC plans to develop test chips at the end of 2024, and begin their large-scale production at the end of 2025. This means that the company's customers will receive the first N2-based chips in 2026.
Meta Quest has learned to recognize gestures
The Meta Quest headset was taught to recognize complex gestures like "high five" and clapping.
The company has updated the mixed reality hand tracking system, adding support for more movements. The new version of the Presence Platform API will become publicly available after preliminary testing by several developers. The technology also promises more complex and reliable interactions on the Meta Quest headset.
The Presence Platform update is aimed at improving gesture tracking using the headset's own cameras instead of the default hardware controllers. The system focuses on recognizing movements when a part of the user's hand is out of sight of the tracking device, including palm clapping, "high five" and crossed arms.
According to The Verge, in the future the company plans to improve the identification of raised thumbs up, pinching, grabbing and poking.
Google Lens has added a text recognition function in Chrome for PC
The Google Lens visual search service was taught to work with text on images in the PC version of Chrome.
The user needs to hover the cursor over the image, right-click, click "Find via Google Lens" and select an action in the command bar that opens.
Google Lens in the desktop version of Chrome. Data:Google.
The tool can highlight text on images, copy it, voice it, and translate it. It also allows you to find the source of the image.
The feature is available in Chrome version 100 for Windows, macOS and Chrome OS users.
Instagram has activated algorithms for promoting original content
Instagram's algorithms have trained us to "appreciate" original content more.
According to the head of the social network Adam Mosseri, authors who create something from scratch should receive more recognition than users who have taken posts from other platforms like TikTok. Developers are adjusting content ranking tools to highlight authentic photos and videos.
Mosseri said that the company is already doing everything possible to "predict" the originality of content by algorithms, but there are no guarantees yet that the decisions made by the models are correct.
Zoom has added a gesture recognition function for sending emojis
Gesture recognition has been added to the desktop version of Zoom.
It is enough for the user to raise his hand or thumbs up to send the appropriate reaction to the interlocutors in a video conference.
The feature is available in the Zoom app version 5.10.3 on Mac and Windows. The option is disabled by default, but those who wish can activate it for each user, account or the entire group.
DeepMind has introduced a new language model with 70 billion parameters
The British AI lab DeepMind has presented a predictive optimal Chinchilla language model with 70 billion parameters. According to the company, the algorithm surpasses GPT-3, Gopher, Jurassic-1 and Megatron-Turing NLG in a wide range of tasks.
The developers stated that Chinchilla has an equal computing budget with the Gopher language model previously created by DeepMind for 280 billion parameters, but has four times the amount of data. The algorithm requires significantly less calculations for fine tuning and logical inference, which greatly simplifies further use, they added.
According to MMLU testing results, Chinchilla achieves 67.5% average accuracy in forecasting tasks, which is 7% higher than that of Gopher.
Performance comparison of different language models. Data: DeepMind.
An underwater city with its own data center will appear in South Korea
South Korean scientists have proposed to build an experimental underwater city with its own data center. The $30 million project is designed for 5 years, and 80% of the costs of its creation will be covered by the state.
The concept of an underwater city. Data: KIOST.
The implementation of the underwater settlement is being carried out by the Korean Institute of Ocean Science and Technology (KIOST) with the support of the Ministry of Maritime Affairs and Fisheries, as well as 23 other departments and private companies. The object is planned to be located near the urban agglomeration of Ulsan.
The project involves the development of technologies and equipment for analyzing the marine environment, construction with the help of drones, the creation of underwater energy sources and communications, monitoring the health and safety of city residents.
The data center will help in the development of future underwater energy-efficient and sustainable data centers. The researchers will also test an early warning system for natural disasters and the use of the complex for underwater tourism. In the future, the creation of a separate hotel is being considered.
An engineer has created an interceptor drone for other drones
A Lithuanian engineer has developed a drone with a Drone Interceptor network to intercept other drones.
The system is a quadrocopter with an increased power-to-weight ratio. The device can overtake a suspicious drone and conduct reconnaissance, transmitting video to a ground operator in real time.
If a potentially dangerous object needs to be shot down, the Drone Interceptor is placed under the target and, at the command of the pilot, detaches all four screws from the frame. Moving up, the propellers open a Kevlar net between them, which entangles and knocks down the "enemy" drone.
The main part of the quadcopter, which remains without screws, descends to the ground by parachute, making a sound signal to simplify the search for operators.
The most important AI deals of the last week
From April 17 to April 23, 2022, startups in the field of artificial intelligence raised over $550 million. Here are the most interesting deals.
Jarvis ML received $16 million to develop a platform that helps brands personalize products, services and experiences.
The creators of the Convoy marketplace, which connects cargo senders with drivers, raised $260 million.
AlphaMountain.ai raised $2.7 million to improve the threat analysis system for cybersecurity platforms.
Mutiny has raised $50 million to develop a marketing service for website personalization.
Prezent.ai received $20 million to expand the development team of the presentation creation tool.
Laiye has raised $160 million to develop a platform for automating office tasks of varying degrees of complexity.
Creators of digital twins for clinical trials of Unlearn.AI raised $50 million.