Ads in Bing Chat, the prominence of Clearview AI and other AI industry news

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

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

Microsoft added advertising in Bing chat

Microsoft added advertising to Bing chat for a limited number of users.

Bing Chat now has Ads!

It’s going to be fascinating to see how the unit economics of Ads in language models will unfold and affect search advertising.

1/3 pic.twitter.com/o5YjRjikOP

— Deedy (@debarghya_das) March 29, 2023

The system embeds relevant sponsored links in responses to queries. Such ads are marked with a footnote “Ad”.

Microsoft Corporate Vice President of Search and Devices Yusuf Mehdi confirmed that the company is experimenting with advertising in Bing chat. More such ads will appear in the search engine, he said.

Mehdi also spoke about the tech giant’s efforts to drive traffic to publishers and web pages from which Bing sources information.

According to him, Microsoft is developing a feature that would allow hovering over an advertiser’s ad to see more links from their site. The company is also exploring adding extended captions to Start’s news-feed content alongside Bing’s AI-generated answers.

Study: ChatGPT will replace text annotators for AI training

Researchers from the University of Zurich believe that ChatGPT is capable of replacing crowdworkers who label text for AI training.

The researchers tasked the chatbot with selecting 2,382 tweets and classifying them by several criteria, including relevance, topic, and the framing of the problem or solution.

According to them, AI proved more accurate and efficient than workers hired via the Amazon Mechanical Turk platform and some specially trained annotators. Also, using ChatGPT cost 20 times less.

However, experts noted that it is too early to speak of human labor being replaced by AI algorithms. The experiment demonstrates the technology’s potential for a specific task, but further research is required to identify the system’s capabilities in data annotation, they added.

Baidu cancelled public demonstration of Ernie Bot

The Baidu company cancelled the public demonstration of its Ernie Bot chatbot.

On 27 March 2023 the company planned to present the technology live, marketing it as “open to media and the public.” Instead, Baidu held a closed briefing for the first group of companies testing Ernie Bot.

Baidu said the format change was necessary to meet the “high demand” from 120,000 organisations that had applied to trial the chatbot. It is reported to be the first of many closed meetings.

Media: Neuralink seeks partners for brain-implant testing

Neuralink has begun seeking partners to test neural interfaces in humans.

According to Reuters, the company is in talks with the Barrow Neurological Institute for clinical trials of implants after receiving approval from U.S. regulators. Earlier, the FDA refused Neuralink’s testing of neural chips in humans.

Sources say the company is also discussing partnerships with other organisations.

Microsoft unveils Security Copilot for cybersecurity professionals

Microsoft has unveiled an AI assistant, Security Copilot, to detect current cyberattacks and forecast future threats.

The system is built on the GPT-4 language model. It trains in real time on enormous data sets, including trillions of security signals the company receives daily. The AI can also learn from users.

It is expected to help security professionals identify connections between different parts of a breach. For example, a suspicious email, a malware file and a compromised component of the system.

The tool is currently available to only a handful of enterprise clients. It remains unclear whether the system will be opened more broadly.

Google to reorganise its AI team to focus on Bard

Google reorganises its team responsible for the voice assistant to concentrate on Bard’s development.

According to an internal memo, Vice President of AI assistant development Jianchan Mao will leave the company for personal reasons. He will be replaced by Piyush Randjan, the CEO of Google Pay.

Vice President of the Assistant Creation Amar Subramanyan will lead the chatbot development team. His predecessor Trevor Stroman will continue as Bard’s local technical lead.

US police used Clearview AI almost 1 million times

American police have deployed the Clearview AI facial recognition system nearly a million times. The company’s founder Hoan Ton-That said hundreds of law enforcement agencies use the startup’s technology nationwide.

Police have not confirmed the information. However, Miami law enforcement agencies admitted using Clearview AI in investigations of all kinds of crimes.

Ton-That also noted the database now contains about 30 billion images drawn from various platforms like Facebook. The company has faced criticism for collecting photos without user consent.

UK regulators urged to develop AI rules

The UK government has urged regulators to develop rules governing the creation and use of AI.

The Department for Science, Innovation and Technology (DSIT) published a white paper with recommendations for the AI industry. It will be presented to Parliament.

The document sets out five principles for companies to follow:

• Safety, security and reliability;
• Transparency and explainability;
• Fairness;
• Accountability and governance;
• Controllability and redress.

Rather than new regulatory mandates, authorities urged regulators to apply existing rules and inform companies of their obligations in line with the white paper.

DSIT tasked the Health and Safety Executive, the Equality and Human Rights Commission, and the Competition and Markets Authority with developing “adapted and context-dependent approaches appropriate to how AI is actually used in their sectors.”

Levi’s to deploy AI models to showcase clothing

Levi’s engaged AI models to showcase clothing and diversify the online shopping experience.

The company has signed a contract with Lalaland, a digital fashion studio. Models created through the partnership will cover a range of body types, ages, sizes and skin tones. They are expected to help customers select items shown on a similarly built and looking person.

The company said it does not view AI as the “only solution” for promoting clothing and does not seek to replace people. The technology will augment models and help create a more personalised shopping experience.

It is not yet clear on which platforms the technology will launch or whether users will be able to tailor samples to themselves.

Levi’s plans to roll out AI models by the end of 2023.

Cosplayer used AI to design makeup

The blogger and cosplayer DeLa Doll used AI to create makeup looks.

She used the Dream by Wombo app to generate makeup ideas over five days. When recreating the results, the artist sometimes tweaked the makeup due to the lack of the right cosmetics or “mood” constraints.

According to the cosplayer, the AI struggles with generating fingers and teeth, but the makeup turns out to be “top-notch.”

The most important AI deals of the past week

From March 26 to April 1, 2023, AI startups attracted more than $181 million. Here are the most notable deals.

• DataDome received $42 million for global deployment of a platform to detect bot-driven cyberattacks.
• Blockchain startup Fetch AI raised $40 million at a $250 million valuation.
• Creators of the conversational AI Perplexity AI raised $25.6 million.
• The robotics company 1X raised $23.5 million from the OpenAI-backed startup fund, Tiger Global and a group of Norwegian investors.
• Parloa received $21 million to develop products for automating contact centres.
• Oscilar raised $20 million to build a risk-management platform for financial institutions.
• Deep Render, the creators of AI technology for compressing videos online, raised $9 million.

Also on ForkLog:

• ChatGPT payment information of subscribers leaked in a outage.
• Twitter algorithms will show only verified accounts in the For You tab.
• Europol warned about the dangers of ChatGPT.
• U.S. authorities will review antitrust rules in AI.
• Hundreds of experts called for pausing GPT-4’s successor training.
• Study: Generative AI threatens 300 million jobs.
• Google accused of using ChatGPT to train Bard.
• Experts criticised the idea of pausing AI research.
• OpenAI accused of violating trade laws.

What to read this weekend?

The ForkLog editors tested an AI model, explored its capabilities and compared it with GPT-3.5.