AI Trends This Week: Challenges with Retaining Talent at OpenAI

Greetings, everyone! Here’s your latest update from TechCrunch’s AI-focused newsletter.

A significant event in AI this week is the departure of another founding member from OpenAI.

John Schulman, integral to the creation of ChatGPT, OpenAI’s AI chatbot technology, has moved to Anthropic, a competitor. Schulman shared on X that his move was driven by a wish to dedicate more time to AI alignment — ensuring AI systems perform as expected — and to be more involved in technical projects.

One might speculate if Schulman’s exit, aligned with Greg Brockman, OpenAI’s president, taking a sabbatical until year’s end, was calculated.

On the day Schulman revealed his departure, OpenAI also announced changes to its DevDay format this year, shifting to multiple smaller developer events rather than a large, single-day gathering. A TechCrunch spokesperson indicated no new model would be unveiled at DevDay, hinting at a slowdown in the development of GPT-4o’s successor. Additionally, the anticipated delay in Nvidia’s Blackwell GPUs could further decelerate progress.

Is OpenAI facing challenges? Schulman’s exit may hint at underlying issues. The atmosphere seems less optimistic compared to last year within Sam Altman’s realm.

Ed Zitron, a PR expert and tech commentator, discussed in his newsletter the hurdles OpenAI may encounter moving forward. OpenAI’s journey toward sustaining success is fraught with challenges, notably highlighted by Zitron.

OpenAI is expected to incur a $5 billion loss this year. Facing escalating costs from AI research, training, and service delivery, the firm needs to secure a significant funding round soon. Microsoft, holding a 49% stake in OpenAI and maintaining a collaborative product development relationship despite occasional competition, seems a likely candidate for further investment. Yet, with Microsoft’s own capital expenditures surging, the prospect raises questions about their willingness to make such a hefty, risky investment.

It’s still probable that OpenAI will find the necessary financial support, albeit potentially under less favorable conditions, possibly affecting its profit cap model.

For OpenAI to survive, it may need to stray further from its founding ethos, navigating into less familiar and more uncertain domains. This shift could have been a difficult decision for Schulman and others, given the growing skepticism from investors and the business community toward the AI sector.

News

Limits to Apple Intelligence: Apple introduced its Apple Intelligence capabilities in the recent iOS 18.1 developer beta, but Ivan reports challenges with sensitive content management.

Google Refreshes its Nest Thermostat: Google updates its flagship Nest thermostat after nine years, launching the Nest Learning Thermostat 4 just before the Made by Google 2024 event.

Election Misinformation on X: The chatbot Grok has been disseminating false claims about Vice President Kamala Harris’ eligibility for the 2024 U.S. ballots on X, leading to an open letter from five secretaries of state to Elon Musk.

YouTube Creator Files Lawsuit Against OpenAI: Alleging infringement, a YouTube content creator is suing OpenAI for using YouTube videos to train its AI models without permission.

Increased AI Lobbying Activities: AI lobbying in the U.S. has escalated, with the number of groups lobbying on AI issues rising significantly within the first half of 2024, signifying the ongoing boom and regulatory interests in AI.

Research Paper of the Week

Despite the innovation potential of open-source models like Meta’s Llama, risks remain, primarily concerning misuse. Yet, a new study introduces a model protection technique to limit undesirable behaviors without significantly impacting accuracy, though scalability challenges persist.

Model of the Week

The AI scene welcomes Flux.1, a potent image-generator challenging existing technologies. Developed by Black Forest Labs, it specializes in textually enriched images, a feat for generative AI. Despite its capabilities, the project’s dataset transparency and misuse prevention plans remain vague.

Grab Bag

As generative AI ventures explore copyrighted material for model training, claiming fair use, legal scrutiny intensifies. Nvidia, for instance, faces speculation over its Cosmos project, potentially setting a precedent for AI’s creative liberties versus copyright law.