AI News Roundup: Hugging Face Partners with Google Cloud

AI Business brings you the latest news and insights from across the AI world.

To keep up-to-date with coverage of all things AI, subscribe to the AI Business newsletter to get content straight to your inbox and follow the AI Business Podcast on Apple and Spotify.

Google teams up with Hugging Face

Google has struck a partnership with Hugging Face to collaborate on open source AI.

The agreement will enable Google Cloud clients to train and deploy Hugging Face models in Google Kubernetes Engine and Vertex AI while accessing Google’s TPU and GPU suite to allow for faster training.

Hugging Face users will be able to deploy their models for production on Google Cloud.

Hugging Face CEO Clem Delangue said on LinkedIn that his startup is the only one “who has commercial collaborations with all the major cloud providers and hardware providers.”

Neuralink finally implants a human

Elon Musk-owned startup Neuralink has completed its first implant into a human brain.

Musk confirmed the news on X (Twitter), saying the initial results showed promising neuron spike detection without further elaboration. He added that the patient was “recovering well."

Neuralink implants a coin-sized, wireless device into a person’s brain to let them control electronics with their minds. The startup’s chips are designed to help quadriplegics regain mobility.

Neuralink received FDA approval last May to conduct human trials of its brain implant.

Microsoft forms team to make conversational AI cheaper

Microsoft has formed a team tasked with creating conversational AI that uses less computing power, making them cheaper, according to The Information.

Several of the company’s leading AI developers have been put on the project, under the name ‘GenAI team.’

Microsoft is already working on developing smaller models, including its Phi system, but now looks to expand to cover conversational systems, like chatbots.

The new GenAI team will be led by Corporate Vice President Misha Bilenko, who will report to CTO Kevin Scott.

X lifts ban on Taylor Swift searches

X (Twitter) has lifted its ban on searches for Taylor Swift’s explicit deepfakes on its platform.

These deepfakes were flooding X so the platform had to halt searches. Users looking for these deepfakes were hit with an error message.

Joe Benarroch, head of business operations at X, told The Guardian the ban was “a temporary action” to “prioritize safety” - but that did not stop one of the images being viewed 47 million times before the account was suspended.

The explicit deepfakes had sparked outrage, with White House press secretary, Karine Jean-Pierre, calling the pictures “alarming.”

Following the news, a bipartisan bill was introduced in the U.S. Senate that would allow victims of explicit deepfakes to demand a civil penalty against individuals attempting to distribute them. The Defiance Act would apply to internet users who “produced or possessed the forgery with intent to distribute.”

“Victims are unable to get justice and the problem is increasing due to a lack of consequences,” said Omny Miranda Martone, founder and CEO of the Sexual Violence Prevention Association. “The Defiance Act will address nonconsensual deepfake pornography by providing victims a civil right of action to seek justice. This will prevent nonconsensual deepfake pornography in the future.”

Amazon scraps iRobot purchase

Amazon has called off its proposed purchase of Roomba maker iRobot as EU regulators said they would block the deal.

The acquisition was announced in August 2022, with Amazon set to pay between $1.4 billion and $1.7 billion for the autonomous vacuum company.

However, the deal came under intense scrutiny. The FTC began to probe the deal just a month after it was announced, while the EU’s antitrust arm raised a formal objection last year.

Regulators fear the deal could restrict competition by giving Amazon the potential to reduce the visibility of other robot vacuum cleaners on its platform.

David Zapolsky, Amazon’s senior vice-president, expressed disappointment over the decision. “This outcome will deny consumers faster innovation and more competitive prices, which we’re confident would have made their lives easier and more enjoyable.”

ElevenLabs user behind fake Biden robocall

Synthetic speech startup ElevenLabs is the platform that was used to create an audio deepfake of President Biden ahead of the New Hampshire presidential primary election, according to Bloomberg.

Fake audio of Biden in phone calls to New Hampshire voters last week told them to stay at home instead of voting in the Democratic primary. The fake Biden told them to save their votes for the national presidential election.

ElevenLabs has now suspended the account of the creator of the deepfake. It also banned other users from creating audio messages of the president.

The startup recently snagged $80 million in a series B funding raise.

Alibaba woos enterprises for its cloud services

Alibaba’s cloud arm has unveiled a serverless version of its Platform for AI to offer enterprises a cost-efficient solution for model deployment and inferencing.

Showcased during its AI & Big Data Summit in Singapore, the platform allows users to tap into computing resources as needed. Users can access a variety of large language models and build customized generative AI applications while only being billed for the computing resources they use.

Alibaba said it plans to expand the serverless solution to give users access to more open source LLMs from the company’s AI model community, ModelScope.

Samsung is using Baidu’s Ernie AI in Chinese phones

Samsung’s new Galaxy S24 smartphones sold in China will include AI features powered by Baidu’s Ernie model.

The Korean handset maker’s add-on mobile services are not available in China, so it has enlisted the help of Baidu as a local alternative.

Ernie will power features like text summarization and translation, as well as Google Android's ‘Circle to Search’ where users can search the web by simply circling or highlighting objects and text.

Ernie is a Chinese attempt to challenge ChatGPT. It was launched last March to a lackluster reception and Baidu has since made a series of updates, claiming Ernie can rival ChatGPT.