Meta loses $730 billion in value, scientists create an ‘invisibility sweater’, and other AI news

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

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

In the United Kingdom, officials warn against emotion-recognition systems

The Information Commissioner’s Office (ICO) warned against using emotion-analysis algorithms for anything more serious than “children’s party games.”

According to the ICO’s deputy commissioner Stephen Bonner, the technology remains “immature,” and the risks associated with it outweigh any potential benefits. He added that misanalysis of data could lead to discrimination.

Bonner also described attempts to use biometrics to detect emotions as “pseudo-scientific.”

TSMC halted shipments of chips to a Chinese AI accelerator maker

TSMC halted shipments of 7‑nm processors to Chinese AI accelerator maker Biren Technology.

According to Bloomberg, the Taiwanese chipmaker took this step amid U.S. sanctions. TSMC does not know whether the Chinese company’s products fall under U.S. restrictions, so it decided to suspend shipments “just in case.”

Earlier Biren Technology unveiled the BR100 accelerator, designed for computer vision and natural language processing tasks. Its closest competitor is Nvidia’s A100.

Gmail accused of routing the Republican Party’s emails to spam

The Republican National Committee (RNC) accused Gmail of deliberately routing the party’s emails to spam. The organisation filed suit against Google.

The tech giant says it does not block emails on political grounds.

According to Google representatives, the mail service’s spam filters are trained on user actions, where users themselves mark similar emails as spam. On that basis, the system could automatically block the RNC’s mailings.

Reports: Microsoft to increase investment in OpenAI

Microsoft plans to increase its investment in OpenAI.

Details of the deal, and the funding amount, have not yet been disclosed. According to reports, representatives from both companies have held several meetings.

Both sides did not comment on the possibility of collaboration.

Meta has fallen out of the top 20 most valuable U.S. companies

Meta is no longer among the top 20 largest U.S. companies.

In October 2022, after the quarterly results were published, the stock fell 23%. Since September 2021, Meta has shed 74% of its value, or $730 billion.

Experts say the core advertising business continues to shrink for a third straight quarter. Investors are not convinced that a pivot toward the metaverse will substantially change the situation.

Meta’s troubles are linked to macroeconomic conditions and privacy changes by Apple on iOS. Under the new rules, users can opt out of data collection in apps, affecting ad targeting.

Sony unveils a camera with a dedicated AI chip

Sony unveiled the A7R V camera with a dedicated AI processor for object identification and tracking.

The device is equipped with a 61‑megapixel sensor. The AI chip enables autofocus to detect and differentiate objects such as cars, trains, planes, animals and insects. It can also assess human poses at 20 key points on the body and anticipate eye position.

The new model will go on sale in mid-December 2022. The price is $3,899.

Mobileye goes public

Intel conducted an IPO for its autonomous-vehicle unit Mobileye.

On the first trading day, the company’s shares rose 37%. The firm’s market capitalization rose from $17 billion to $23 billion.

Intel retained a controlling stake in Mobileye and kept Amnon Shashua as CEO.

Neuralink postpones presentation to November 30

Elon Musk moved Neuralink’s presentation to November 30, 2022.

Originally, the event was slated for October 31. Musk did not explain the reason for the postponement.

The last time Neuralink showcased its work was in April 2021. Back then the company demonstrated a monkey playing ping-pong with the power of thought.

Labels express concern over the popularity of music AI services

Recording companies expressed concern about the spread of music-generation algorithms.

The American Recording Industry Association (RIAA) says AI threatens not only labels but artists as well. According to the organization, such algorithms are trained on data that may be protected by copyright.

RIAA also said that AI services such as Acapella-extractor and Remove-Vocals enable unauthorized copies of musical works.

Dating site opens access to AI nudity detector

The Bumble dating service opened the source code of its AI tool that helps detect inappropriate images.

Private Detector blurs naked bodies, giving users a choice whether to view the photo. According to developers, the tool’s accuracy stands at 98%.

The project’s source code is available on GitHub. Bumble says the tool will allow small companies to integrate Private Detector into their services and protect users from unwanted content.

Engineers create a sweater that fools recognition systems

Researchers at the University of Maryland have created a sweater that hides people from recognition systems.

Developers exploited vulnerabilities in algorithms trained to identify specific objects. Using a large dataset, they generated a pattern with an image of people and adjusted it until the computer vision system stopped recognizing the person. They then applied the final print to a regular sweater.

As a result, the recognition system could not identify the person in the camera’s field of view.

The week’s top AI deals

From October 23 to 29, 2022, AI startups raised more than $150 million. Here are the most notable deals.

• The cashier-less store tech developer Trigo attracted $100 million.
• AI chip maker Axelera AI raised $27 million to commercialize its hardware.
• Navina raised $22 million to support its platform for processing and analyzing medical data.

Also on ForkLog:

• Israel deployed AI weapon near a refugee camp.
• Shutterstock will start selling AI-generated images.
• Study: Facebook’s advertising algorithms distinguish race, gender and age in photographs.
• Media: U.S. DOJ investigates Tesla’s advertising promises about autopilot.
• The drone startup Argo AI is being liquidated.
• In the United Kingdom, calls have been made to ban police use of facial recognition .

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