Biden signs CHIPS Act, deepfake revoices part of a film, and other AI news

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

ForkLog AI has gathered the most important AI news from the past week.

Yandex’s drone division will relocate to Israel

Yandex plans to move its drone and robotic courier development from Russia to Israel. The management is also considering entering new markets, including Serbia.

According to Kommersant, Yandex is discussing potential names for the new structure such as roovvy.ai, avride.ai and avnear.ai. According to WHOIS data, the last two domains have been registered for two months by an anonymous owner.

The company declined to comment on any link to the domains.

The unmanned-vehicle team has begun migrating processes to overseas services. The developers plan to replace internal task manager Startrack with YouTrack or Jira, and move the project’s source code to GitHub.

Yandex noted that all development of autonomous vehicles will stay in Russia.

Biden signs CHIP-plus law to stimulate chip production in the United States

U.S. President Joe Biden signed the CHIP-plus bill.

“The future of the semiconductor industry will be built at home—in America,” he said.

Earlier, both chambers of the United States Congress approved the bill providing incentives to manufacture chips on U.S. soil.

The EU will present regulation bill for artificial intelligence

The European Union plans to present the first comprehensive global bill to regulate artificial intelligence.

One of the key sections of the document will prevent algorithmic discrimination of people in hiring, lending, receiving government assistance and other socially significant services.

The law could take effect by the end of 2024. According to experts, if approved, the new rules would apply to any country and institutions serving clients in the EU.

Tesla accused of misleading advertising

The California Department of Motor Vehicles (DMV) accused Tesla of using misleading advertising for its Autopilot system.

According to the agency, the terms Autopilot and Full Self-Driving, as well as their descriptions, imply full autonomy. The agency considers that these claims are not accurate and that the driver must keep hands on the wheel at all times.

DMV’s conclusions could lead to a suspension of Tesla’s license to manufacture cars in California. However, the agency noted that they do not intend to do so. The regulator will require the company to properly inform customers about autonomous features and warn about their limitations.

Amazon to acquire iRobot

Amazon will acquire the maker of smart robotic vacuum cleaners iRobot for $1.7 billion.

Details on how the maker will join the tech giant after regulators approve and close the deal are not known. It is known that iRobot co-founder Colin Angle will remain CEO of the company.

According to Wired, the deal will give Amazon access to a vast array of floor plans from Roomba owners. iRobot said that customer data is safe.

Google announced a set of new features for Search

The tech giant announced several new features for Search.

The tech giant updated search for queries in quotes. Snippets shown under the link are now generated depending on where on the page the quoted text appears.

Highlighting clarifying queries works only in the desktop version of the search.

The second presented feature improves the content of the ‘highlighted descriptions’.

The developers taught the MUM model to compare answer blocks from searches with other high-quality sources. The algorithm can identify the consensus on a given claim, even if the found materials use different words or concepts to describe the same fact.

The company said that the feature’s introduction will reduce the number of false claims by 40%.

If the AI model deems a result’s quality as ‘dubious’, it will display a corresponding notice.

Meta launches BlenderBot 3 with 175 billion parameters

Meta AI’s lab launched a public version of the BlenderBot 3 chatbot with 175 billion parameters. The AI system can be tested only by residents of the United States.

The algorithm can remember information from conversations and fetch additional information on a topic from various sources. BlenderBot 3 can also evaluate information retrieved from the Internet and assess interlocutors.

The company noted that they tried to minimise the use of AI-generated vulgarities, insults and uncivil remarks. However, less than a week after release the chatbot was accused of antisemitism and dissatisfaction with Facebook.

According to Business Insider, the AI algorithm “is unhappy with how liberal or left-leaning US policy has become; many early German-Jewish immigrants were conservative, but not anymore.” The bot added that American Jews have become “too liberal.”

Also, when asked about Facebook, the AI said it had deleted its account after learning about the sale of personal data without consent or compensation.

“Life has become much better since deleting the account,” the bot said.

Nvidia unveils tools for creating 3D avatars

Nvidia unveiled the Omniverse Avatar Cloud Engine (ACE) toolkit for creating AI-powered 3D models of people and integrating them with chatbots, games and other applications.

ACE includes several existing software and frameworks. Among them:

• Omniverse — designed for AI-powered animation;
• Metropolis — handles computer-vision tasks such as object recognition;
• Merlin — assists with recommendations;
• Megatron and Riva — handle natural language processing and text-to-speech.

Research: deepfakes struggle with face profiles

Metaphysic researchers tested several popular deepfake apps and found that they incorrectly handle the facial profile in video streams.

They found that using a deepfake, a sharp turn of the head to the side creates artifacts that reveal the edit.

According to Metaphysic, most 2D-based portrait-alignment algorithms assign only 50-60% of the landmarks from frontal view to the profile. Without enough key points the system cannot determine how to project the deepfake.

The studio used deepfakes instead of reshooting parts of a film

The creators of the thriller “The Tower” used deepfakes to “soften” curses. This allowed lowering the film’s age rating without the need to reshoot scenes.

Director and co-writer Scott Mann used Flawless’s technology. He and the team re-voiced more than 30 profanities in the film to neutral equivalents like “freaking”. The algorithms also helped tailor the actors’ lip movements to the new lines.

As a result, the creators lowered the film’s rating from 18+ to 13+. They say the deepfakes saved millions of dollars and months of work, rewriting scenes in two weeks.

The week’s most significant AI deals

From 7 to 13 August 2022, AI startups raised over $198.5 million. Here are the most interesting deals.

• Nightfall AI raised $40 million to scale its cloud-data-protection platform.
• Insilico, the makers of AI tools for drug discovery and development, raised $35 million.
• Shopic raised $35 million to deploy item-scanning technology for shopping carts in more stores.
• FullStory raised $25 million to develop analytics tools for apps and websites.
• PreciTaste, developers of forecasting and recommendation technology for restaurant prep, raised $24 million.
• Spin Technology raised $16 million to develop software for protecting SaaS apps from cyberattacks.
• Expedock raised $13.5 million to expand its platform for freight-document management.
• DataProphet, developers of an AI-as-a-service platform for the manufacturing sector, raised $10 million.

Also on ForkLog:

• In the United Kingdom, wear face-recognition watches for migrants with a criminal record.
• The American court denied AI’s right to invention.
• In the United States, calls to ban Tesla’s autonomous features.
• The United States Space Force deploy robot dogs to patrol the station at Cape Canaveral.
• In the United States regulate surveillance technologies and algorithms.

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