China tests algorithms on Uyghurs, synthetic people rap and other AI world news

We aim to inform readers not only about events in the bitcoin industry but also about developments in adjacent technology spheres — cybersecurity, and now in the world of artificial intelligence (AI).

ForkLog has gathered the week’s most important AI news.

The Russian government will allocate 6.49 billion rubles to support AI deployment

The Government of the Russian Federation will allocate 5.26 billion rubles in subsidies to companies that will implement AI by 2024.

The funds are planned to be distributed on a competitive basis. They can be used to purchase software and equipment, pay for programmers, rent office premises and other purposes. The grant size for a single company will not exceed 100 million rubles.

Additionally, 1.23 billion rubles will be allocated to support events during which schoolchildren and students can develop their own AI projects. The best initiatives will be helped by experts and mentors to bring the product to market.

Europeans backed the idea of replacing parliamentarians with artificial intelligence

More than half of Europeans would prefer to replace lawmakers with artificial intelligence. This finding came from researchers at the IE Change Management Center, who surveyed 2,769 people across 11 countries.

Researchers asked respondents how they would react to a reduction in the number of national parliamentarians in their country and the allocation of those seats to an AI that would have access to their data.

As a result, 51% supported such a move.

Spain showed the strongest support — 66% of respondents would replace parliamentarians with AI. In the United Kingdom, 69% rejected the idea.

In China about 75% of respondents supported replacing lawmakers with AI, while 60% of American respondents opposed the idea.

Opinions vary sharply by age. Younger people were more open to the idea. More than 60% of Europeans aged 25–34 and 56% aged 34–44 supported it, while the majority of those over 55 did not.

ARM announced the first CPUs on the v9 architecture

The company ARM introduced the first processors and graphics accelerators based on the v9 architecture.

The flagship Cortex-X2 processor is aimed at premium smartphones and laptops. The company said it is 30% more powerful than current systems.

The mid-range Cortex-A710 and energy-efficient Cortex-A510 target smartphones, IoT devices, wearables and other smart devices. The company claims the A510 is the first budget, high‑efficiency core in four years, delivering a 35% performance uplift and a threefold increase in ML throughput.

The flagship GPU Mali-G710, according to ARM, is 35% faster in ML tasks. Budget GPU variants will also offer performance and energy-efficiency gains over the previous generation.

Devices based on the new architecture are expected in the first half of 2022.

Microsoft unveiled the first product based on GPT-3

Microsoft announced the first commercial product based on GPT-3 at Build 2021listed the tool for code autocompletion on the Power Apps low‑code development platform.

GPT-3 runs in Azure, trained via Azure Machine Learning.

The service will be available by the end of June 2021 as part of public testing. Microsoft plans to extend GPT-3 usage to other developer tools in the future.

Facebook developed a self-supervised speech-recognition algorithm

Facebook engineers developed a Wav2vec-U speech-recognition model that does not require annotated datasets.

They have tested it on Swahili, Kyrgyz and Crimean Tatar, languages with limited high-quality datasets. The developers say the system yields 63% fewer errors than the best unsupervised method and that the tool is as accurate as supervised systems a few years ago.

Facebook believes their discovery could spur speech-recognition systems for more languages and dialects worldwide. To accelerate development, the researchers posted the Wav2vec-U source code on GitHub.

Sber presents a fifth‑level autonomous electric car

SberAutoTech, the Sber ecosystem company, unveiled a fifth‑level autonomous electric prototype called FLIP.

The developers say the robo-car uses a suite of cameras, sensors and LiDARs to orient itself, recognise pedestrians and obstacles, and make decisions on acceleration, braking and turning.

External dimensions of the FLIP are the same as a standard passenger car. The autonomous vehicle can be used for passenger transport as well as cargo delivery.

The vehicle is currently being tested on closed sites. When it will appear on city streets is unknown.

Germany allowed driverless cars to operate on public roads

The Bundestag of Germany passed a law permitting driverless vehicles to operate on public roads.

The document covers fourth‑level autonomous cars, meaning the computer handles driving in specific conditions or environments.

The bill’s authors listed initial road‑use applications for autonomous vehicles in Germany, such as public transport, business travel, logistics and corporate shuttles.

Companies wishing to operate commercial robo-cars must ensure the ability to remotely control the vehicle.

The bill now needs to be approved by the upper chamber. If approved, driverless cars could be on German public roads as early as 2022.

Tesla stopped equipping cars with radar sensors and uses cameras to monitor the driver

Tesla began supplying new Model 3 and Model Y cars without radar sensors for the full self‑driving feature.

Some buyers may receive cars with restrictions on Autopilot and ADAS features. The company promised that software updates “in the coming weeks” would re‑enable these features.

Earlier, Elon Musk said the company would drop radar in favour of computer‑vision and deep‑learning algorithms.

Also Tesla activated a camera above the rear‑view mirror in Model 3 and Model Y to monitor the driver while Autopilot is engaged.

Regulators and safety experts have repeatedly urged Tesla to improve driver‑monitoring, including via in‑cab cameras. Musk has rejected these proposals and called the technology ineffective.

YouTube banned collecting data on faces from videos

YouTubeadded a clause to its Terms of Service regarding a ban on collecting data for facial recognition. Company representatives said the ban on collecting data for identifying a person has always existed, but the wording has now been made more precise.

“We decided to fix this, because we take privacy seriously and want users to be confident in the safety of their data,” the company said in a statement.

The changes will take effect on June 1. In addition, YouTube will add ads to all videos on the platform, and creators will begin paying taxes.

WHO presents a chatbot that helps people quit smoking

The World Health Organization (WHO), in partnership with Soul Machines, Amazon Web Services and Google Cloud, has developed an AI chatbot named Florence to help people quit smoking.

The assistant explains proven strategies for stopping smoking and why quitting tobacco is a good idea. Florence can converse by voice, but if she does not understand, the system will switch to text messaging.

The bot is available on the WHO’s official site here.

Startup taught synthetic people to rap

The Replica startup published a video in which synthetic rap artists perform the lyrics of a song. The developers say the experiment demonstrates the potential for integration between their speech‑synthesis tools and MetaHumans in Unreal Engine to create realistic CGI‑person performances.

Replica notes that with these two tools it is possible to create dialogues between synthetic people with realistic lip-sync in video games and films.

Also on ForkLog:

• Hermann Gref: artificial intelligence will force people to change professions.
• Roskomnadzor employed AI for monitoring prohibited content.
• China tested AI on Uyghurs for emotion detection.
• Scientists used neural networks to forecast storms.
• European rights advocates filed complaints against Clearview AI.

Subscribe to ForkLog AI news on Telegram: ForkLog AI — all the AI news!