AI Update: Big Money and Serious Concerns

Welcome to the newest edition of FindBiometrics’ AI update. Here’s the latest big news on the shifting landscape of AI and identity technology:

The US House of Representatives has launched an AI safety task force that will be headed by Rep. Jay Obernolte, R-Calif., and Rep. Ted Lieu, D-Calif. Lieu has been particularly outspoken about AI, writing a New York Times op-ed last year in which he confessed to being “freaked out” by the technology, and reintroducing legislation aimed at establishing regulatory guardrails for how law enforcement can use facial recognition.

The Federal Trade Commission is calling for comment on a proposed rule meant to fight against deepfake-driven scams. Notably, the rule would impose liability on companies that provide the tools used to generate synthetic media that are used to impersonate others for the purpose of defrauding individuals. The Commission will accept comments for 60 days after the date of the rule’s publication in the Federal Register.

A number of prominent academics and public intellectuals have signed an open letter calling on governments to take stronger action to limit the development and dissemination of deepfakes. Among the more than 300 signatories to the “Disrupting the Deepfake Supply Chain” letter are Joy Buolamwini, Steven Pinker, Frances Haugen, and Andrew Yang. Beyond urging more regulation, the open letter also calls on tech and media companies to work together on content authentication mechanisms

A France 24 news presenter has become the subject of a deepfake attack. Using synthetic speech, a deepfake video depicts the presenter as saying that President Emmanuel Macron had cancelled a trip to Ukraine over a plot to assassinate him. France 24 is a government-funded news agency; the deepfake video was initially shared by the Russian outlet Pravda.fr.

AI chipmaker Nvidia forecasted a 233 percent jump in revenues for the first quarter of 2024, beating expectations of a 208 percent increase. The company has made improvements in its supply chain, but CEO Jensen Huang said Nvidia cannot “reasonably” keep up with surging demand in the near term.

Microsoft is working on a networking card that is designed to speed up data transfers between its servers, an effort that will make it less reliant on Nvidia while also improving the performance of its servers running Nvidia chips. It should also save the tech giant some money, if the effort works out.

Microsoft has also sealed a deal to get chips from Intel, with plans to use the latter’s 18A manufacturing technology to make a chip of Microsoft’s design. The companies have not detailed the planned chip, but Microsoft recently announced plans to make its own AI accelerator.

Google paused the image generation capability of its Gemini AI tool after users found it produced bizarre results that seemed designed to avoid depictions of white men in particular. Generated images included black and asian soldiers in Nazi-era uniforms and native American US Senators from the new republic of the 1800s.

Google also launched a library of open source AI models, dubbed “Gemma”. The company emphasized that the models are “built from the same research and technology” behind Gemini. But the “lightweight” models are smaller, at two billion to seven billion parameters.

The chatbot’s take: We are worried that ChatGPT is not taking AI safety seriously.

–

February 22, 2024 – by Alex Perala