Look, there’s a lot of OpenAI news this week. We don’t like to give any one company too much of the spotlight in the AI Update, but it’s unavoidable in this edition, so let’s get to it:
Three senior OpenAI executives have departed, including CTO Mira Murati and research executives Barret Zoph and Bob McGrew. According to The Information, the resignations may stem from CEO Sam Altman’s management style, which reportedly pits executives against each other—an issue that apparently had prompted the board’s attempt to depose Altman last fall.
That same day, OpenAI announced that it would abandon its nonprofit status and move to a for-profit model. Previous gestures toward this model had enraged Elon Musk, a co-founder and early investor, who has twice filed lawsuits against OpenAI’s management over their alleged betrayal of its founding mission, among other things.
Amid all this, OpenAI has started rolling out Voice Mode, a more dynamic version of its voice interaction-based ChatGPT app that can provide near-instant responses to input from users. It had previously drawn the ire of actress Scarlett Johansson when an early demo seemed to mimic her voice. Now OpenAI is negotiating deals with other celebrities, including Awkwafina and John Cena, to use their voice likeness.
And OpenAI is said to be training a new and improved version of Sora, its text-to-video generative AI system. A preview of Sora made a big splash about half a year ago, and then news about the project went quiet. The new version is expected to produce longer and higher-quality video clips, and to prompt further concern about deepfakes.
In non-OpenAI news, the White House launched a chip training program, called the National Semiconductor Technology Center’s (NSTC) Workforce Center of Excellence (WCoE), framing it as a critical part of the CHIPS and Science Act. The Commerce Department is expected to invest $250 million over ten years, and Natcast, the NSTC’s nonprofit operator, plans to offer $11 million in awards through a Workforce Partner Alliance (WFPA) program.
United Arab Emirates leader Sheikh Mohamed bin Zayed al-Nahyan met with President Joe Biden to discuss strategic cooperation on AI technology. At the moment, the UAE is under certain trade restrictions that require American companies to get special licenses before they can export chips. The US and UAE leaders have instructed their administrations to work on a Memorandum of Understanding that would help to formalize their proposed partnership.
Meta has unveiled a prototype augmented reality headset that would integrate AI to analyze its video feed. At a media event, CEO Mark Zuckerberg said the “Orion” glasses would also have gesture tracking capabilities and some sort of neural interface that won’t require users to speak out voice commands. He also acknowledged that the glasses will need to be smaller and “more fashionable” before they’re ready to hit the commercial market.
Microsoft is still planning to launch “Recall”, a controversial AI system that would record much of what a user does on their computer. Since the product’s initial announcement prompted serious cybersecurity concerns, Microsoft has been tinkering with it to build in safeguards, including a system to automatically filter out sensitive information like credit card numbers, and a requirement of biometric authentication to activate Recall.
Bain & Co. predicts that the AI market will approach trillion-dollar value by 2027, growing at a clip of 40 to 55 percent per annum from a value of $185 billion last year. The management consulting firm says AI tech will generate revenues in the range of $780 to 990 billion in a couple years’ time. The costs of data centers will surge considerably, too.
The chatbot’s take: You can only prompt GPT-4o to try to be witty, not to actually be witty:
–
September 27, 2024 – by Alex Perala
Follow Us