Nvidia Reveals Progress in Neural Graphics; Elon Musk Planned to Buy DeepMind; and Other AI News

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

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

• In 2013, Elon Musk sought to acquire the AI laboratory DeepMind.
• IBM will pause hiring for roles that AI could potentially take on.
• Nvidia released about 20 research papers on advancing generative AI.
• A German photographer was fined for requesting removal of his works from the LAION-5B dataset.
• Amnesty International used deepfakes to illustrate the Colombian protests.
• The week’s most notable AI deals.

Media: Elon Musk Planned to Buy DeepMind

In 2013, Elon Musk wanted to buy the AI laboratory DeepMind.

According to The Wall Street Journal, the billionaire made a ‘last-minute’ bid to acquire the company. However, in negotiations he was outbid by Google.

Musk planned to lead DeepMind’s research. He told colleagues that he could not trust then-CEO of Google, Larry Page, to oversee the development of powerful AI.

Google Launched AI Courses in Ukraine

The Ministry of Digital Transformation of Ukraine and Google launched a free online course ‘AI Basics’.

Among the main topics:

• What artificial intelligence is and why it is needed;
• Which ready-made AI solutions can be applied across sectors—from marketing to management;
• Strategies for deploying machine learning by companies;
• How to use the technologies without coding experience.

Training will begin on 16 May and run for four weeks. Those wishing to attend the course should register.

IBM Announces Replacement of Some Jobs by Artificial Intelligence

IBM will pause hiring for roles that AI could potentially take on.

According to CEO Arvind Krishna, roughly 26,000 back-office workers who do not interact with customers could be affected. Over five years, 30% of them could be easily automated or performed by AI, he said.

The changes are likely to affect around 7,800 jobs. Krishna said that after the reductions there are no plans to hire for certain roles.

LinkedIn Introduced an AI Tool for Writing Messages

The professional network LinkedIn has launched a generative AI-based feature for automatically drafting messages to recruiters. It is available to Premium subscribers.

According to the company, the tool uses information from the user’s profile and the recruiter’s account. It also analyses the job description and information about the target company. It then composes a ‘highly personalized’ letter.

LinkedIn emphasised that users should ‘double-check’ the generated text and edit it if necessary.

Nvidia Unveils Advances in Generative AI

Nvidia released about 20 research papers on advancing generative AI and neural graphics.

They cover text-to-image transformation algorithms and inverse rendering tools to turn static illustrations into 3D objects. The documents also include information on neurophysiological models that AI uses to simulate complex 3D elements with high realism, and neural rendering technologies.

Nvidia says the research will enable developers and companies to generate synthetic datasets and virtual worlds for training robots and drones. They will also help creators to rapidly produce high-quality effects for storyboarding, previsualization, and production.

Walmart Suppliers Prefer Interacting with Chatbot

Walmart has deployed an AI-based chatbot to close some deals with suppliers.

The system was developed by California-based Pactum AI. When using it, the retailer simply specifies the desired item, the available budget and additional details such as discounts or payment terms. The chatbot will then negotiate with suppliers and secure the best deal for Walmart.

So far, AI has been tasked with negotiating for grocery trolleys and other essentials for stores, not shelf products.

The bot reduces deal-making time from weeks or months to a few days, and the cost-efficiency reaches 3%. Also, according to Walmart, three-quarters of suppliers prefer negotiating with AI rather than a human.

Initially the retailer trialed the technology in Canada. Later, the company began using it in the United States, Chile and South Africa.

Photographer Fined for Demanding Removal of His Works from Dataset

German stock photographer Robert Kneschke received a fine for requesting the removal of his works from the open training dataset LAION-5B.

In February 2023 he learned that his photographs were used to train generative models. He also found a ‘ton of images’ from his portfolio in LAION-5B.

The photographer asked the dataset owner to remove the works and received a reply from the law firm Heidrich Rechtsanwälte. The firm said its client supports the database only with links to online images and does not store pictures. Therefore, ‘there is no violation and nothing to delete,’ the lawyers said.

At the end of March Kneschke sent a request to LAION to cease infringing actions. He again asked to delete the photos and provide information about the extent and timing of their use.

The lawyers at Heidrich Rechtsanwälte responded that there were no violations by the client. They also demanded payment of $979 within 14 days as compensation for a frivolous complaint.

At the end of April, the photographer filed a lawsuit against LAION in the Hamburg District Court in Germany.

Grimes Unveils Software to Imitate Her Voice

Singer Grimes unveiled Elf Tech software for simulating her voice.

HOW TO MAKE MUSIC FEAT GrimesAI
— we can distribute it for you and you can earn royalties from your work: https://t.co/p598CXaXnD

— ?????? (@Grimezsz) April 30, 2023

Users need to upload an audio track and click the ‘Create’ button. The program will sing it in Grimes’ voice. Creators can commercially distribute the result and receive 50% of the master recording royalties.

The artist warned users against generating inappropriate tracks such as a Nazi anthem or songs about killing children.

Grimes also provided stems and samples for training her own AI algorithms. In addition, she announced two new tracks, Music for Machines and I Wanna Be Software.

According to the artist, Elf Tech is still in beta, so the result may not be perfect.

Amnesty International Used Deepfakes to Illustrate Colombian Protests

The human rights organisation Amnesty International used AI to create images timed to the two-year anniversary of Colombia’s protests.

Three illustrations published on Twitter depicted police brutality toward protesters during events in April 2021. Tens of thousands of Colombians protested tax hikes, corruption, and healthcare reform.

AI images by Amnesty International contained artefacts such as misaligned colours on the flag, distorted faces and hands, and scrambled letters on officers’ uniforms.

The organisation explained the use of AI was out of a desire not to “put protesters at risk.” Many people hid their faces due to fear of repression and stigmatisation by state security forces, Amnesty International added.

Experts warned that employing AI could set a troubling precedent and undermine trust in human rights advocates.

AI Ran the Broadcast on a Swiss Radio Station

In Switzerland, AI ran the airwaves of the French-language radio station Couleur 3 all day.

From 6:00 to 19:00 the algorithms selected and created music. They also told jokes and read out fictional news, such as a ban on spaceflight over Geneva’s airspace due to noise complaints.

Preparation for the experiment took three months. The text was written by ChatGPT and other language models, and the voices were supplied by voices cloned by Respeecher, a Ukrainian company, for five presenters.

Every 20 minutes listeners were told that the programme was run by artificial intelligence.

According to Couleur 3, the station received a flood of messages after the experiment. In 90% of them, people said that ‘it turned out cool, but lacked the human element.’

The Week’s Most Significant AI Deals

From May 1 to May 6, AI startups attracted more than $455 million. Here are the most notable deals.

• Cohere, the makers of natural language processing software, raised $250 million.
• Mavenir, a telecommunications company, attracted $100 million to accelerate AI integration into cloud-based automated network software.
• Simpplr raised $70 million for intranet platform development.
• Pando raised $30 million to expand order-management technologies.
• Alison, the creators of a platform for analyzing creative advertising, raised $5.1 million.

Also on ForkLog:

• ChatGPT unblocked in Italy.
• In the United States, there were proposals to ban AI from launching nuclear bombs.
• “Godfather of AI” left Google and warned of the risks.
• Binance: ChatGPT spreads misinformation about Changpeng Zhao.
• AI turn thoughts into text.
• Samsung banned employees from using chatbots.
• WEF: forecasts for enterprise automation did not pan out.
• Microsoft introduced new features for the Bing chatbot.
• The White House discussed AI risks with tech giants.
• Britain’s antitrust authority will study the AI market.