Plus: AI infrastructure firm CoreWeave secures $7.5B in debt financing.
| For May 17, 2024 |  |
| Here’s a look at today's AI briefing: - OpenAI disbands its "superalignment" team.
- Gannett using AI to generate news summaries.
- Microsoft will offer AMD AI chips.
- CoreWeave secures $7.5B in debt financing.
Beth
p/beth-duckett | |
| 1 | | OpenAI has disbanded its "superalignment" team for managing long-term AI risks following the resignations of its two leaders. OpenAI co-founder Ilya Sutskever and Jan Leike announced their departures earlier this week. More: - OpenAI formed the superalignment team last year to lead “scientific and technical breakthroughs to steer and control AI systems much smarter than us.”
- The team will no longer operate as a standalone entity.
- Instead, OpenAI is integrating the group's members across its research to bolster its safety initiatives.
Zoom out: - In a post on X, Leike wrote today that OpenAI's “safety culture and processes have taken a backseat to shiny products.”
- Leike said he joined OpenAI to lead in research, but longstanding disagreements with leadership on priorities led to "a breaking point."
- He acknowledged that problems of superintelligent AI are "hard to get right" and he is concerned OpenAI is not "on a trajectory to get there."
- CEO Sam Altman responded to Leike’s post. “He's right we have a lot more to do,” Altman wrote. “We are committed to doing it.”
|     | |
| 2 | | U.S. media giant Gannett is using AI technology to generate summaries of news articles. The "key points" feature has started appearing directly below the stories' headlines. More: - Gannett, the largest U.S. newspaper publisher by daily circulation, owns USA Today and over 100 other daily newspapers.
- Its new AI-generated bullet points are now active on some USA Today online stories.
- The articles with the summaries include a disclaimer reading: "The Key Points at the top of this article were created with the assistance of Artificial Intelligence (AI) and reviewed by a journalist before publication."
- The disclaimer also notes that no other parts of the story were generated with AI.
Zoom out: - The rollout comes after Gannett reversed its decision to experiment with an AI sports writing tool.
- The company last year paused the "LedeAI" tool to write recaps of high school sports games after readers complained they were poorly written.
| | | |
| 3 | | Microsoft will offer AMD's AI chips through its cloud service, providing an alternative to Nvidia’s GPUs in the data center AI chip market. 
The news: Demand for GPUs to power AI services continues to outpace supply. Microsoft announced plans to sell AMD's MI300X AI chips through its Azure cloud computing service, offering another option besides Nvidia’s in-demand H100 GPUs. What the numbers show: The $17.7B data-center AI chip market has NVIDIA leading with a 65% share, followed by Intel at 22%, and AMD at 11%. What's next: Microsoft said it will reveal more about the AMD service at its Build Conference May 21-23, when Microsoft is also expected to reveal other new AI offerings for consumers, developers, and businesses. | | | |
| 4 | | AI infrastructure firm CoreWeave secured $7.5B in debt financing to expand its cloud data centers. The announcement comes shortly after CoreWeave finalized a $1.1B funding round that values the AI cloud provider at around $19B. More: - CoreWeave rents out Nvidia GPUs to companies for running AI models, competing with Amazon, Google, and other cloud infrastructure operators.
- Blackstone and Magnetar co-led the latest round, one of the largest private debt funding rounds to date, with participation from Carlyle Group and BlackRock.
- CoreWeave will use the funding to buy more GPUs and build out its new data centers to meet growing demand from enterprise and AI customers.
Zoom out: - Nvidia is also an investor in CoreWeave, which has supplied Microsoft and Mistral AI with GPU access.
- The New Jersey startup has 14 data centers, all in the U.S., with plans to expand to 28 by year's end.
| | | |
| 5 | | European lawmakers are investigating Meta's Instagram and Facebook for potentially failing to protect children. EU regulators said they are concerned Meta's algorithms "stimulate behavioral addictions in children." More: - The European Commission, the EU’s executive body, opened formal proceedings this week to determine if Meta violated the bloc's Digital Services Act (DSA).
- The law requires most online platforms to take more action to tackle illegal and harmful content.
- The EU has expressed concerns about Meta's algorithms leading to the "rabbit-hole effect," when a user views harmful content and is recommended similar and more extreme content.
- The European Commission is also concerned about Meta's age verification methods for children related to accessing inappropriate content.
- In response, Meta said it has developed over 50 tools and policies to protect children online, adding that it will share details of its work with the EU.
| | | |
| |
| |
 | AI and technology writer | | Beth is a contributing editor and writer of Inside's AI and Tech newsletters. She has written for publications including USA Today, the Arizona Business Gazette, and The Arizona Republic, where she received recognition with a Pulitzer Prize nomination and a First Amendment Award for collaborative reporting on state pension cost increases. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Beth Duckett | |
| |
| |
RE:
RE ALL:
FWD:
