As China continues playing catch-up with the US and OpenAI slowly transitioning from non-profit to for-profit, the company is taking the initiative to look to solidify both its own and the US' position in AI. Other key takeaways include:
- Amazon and General Catalyst join up to launch an AI-based global healthcare partnership
- The US government announces its last set of AI chip export regulations under the Biden administration
- A team of researchers hailing from UC Berkeley releases an open-source reasoning model developed on just $450
Join us at AI Tangle as we untangle this week's happenings in AI!
|
Earlier on Monday, OpenAI unveiled a new policy framework, advocating nationwide AI regulations, fair access to AI benefits, and fostering AI development across US regions to curb China's ever-growing pace. The "economic blueprint" proposals include streamlining AI export rules, creating regional AI hubs, supporting public universities with compute resources, and expanding energy infrastructure for data centers - all to further solidify the US' position in the global AI arms race.
Let's elaborate a little more on OpenAI's plan
Aside from what the blueprint is putting emphasis on, OpenAI also wants to push for better international AI standards and balanced copyright laws, arguing that US developers should use publicly available data, including copyrighted material, to train models. Sam Altman has further criticized the fragmented state-level AI regulations and current federal policies, such as the CHIPS Act, claiming they're "not as effective as any of us hoped." OpenAI has spent a fair amount on lobbying efforts and partnerships in defense, and it's clear that the company wants to become a mainstay in molding US AI policy.
|
AWS and venture firm General Catalyst have recently announced a partnership between the two to integrate AI across healthcare to improve patient outcomes and accessibility. The partnership will leverage AWS's AI tools, like Amazon Bedrock, along with teaming up with the likes of Mistral and Anthropic to develop predictive and personalized care solutions, optimize diagnostics, and boost operational efficiency. AWS CEO Matt Garman called it a "bold step" toward transforming global healthcare, though emphasis is also on the US where many young people worry about the associated costs.
Following up on last week's news, the US government announced stricter export regulations for AI chips, dividing the world into "three tiers" to limit access while prioritizing allies. Countries like China, Russia, Iran, and North Korea remain part of the "blocked" tier, while close allies, including Japan and the UK, are exempt. The new rules cap chip exports and restrict global AI model development as a means for the US to maintain its dominance and mitigate national security risks. However, industry leaders like Nvidia criticized the measures as "overreach," warning of unintended consequences for global markets.
In a recent blog post, UC Berkeley's NovaSky team unveiled Sky-T1-32B-Preview, an open-source reasoning AI model competitive with earlier versions of OpenAI's o1 on key benchmarks like MATH500. However, the real kicker for Sky-T1 is that it was developed for under $450, an astronomically small sum in comparison, by leveraging synthetic training data and more efficient computing resources. Though Sky-T1 still fell short on broader scientific benchmarks like GPQA-Diamond in comparison to more fleshed-out models, Sky-T1's incredibly low price-to-train could be a sign of things to come.
NY-based cloud computing firm CoreWeave has recently introduced its first two UK data centers, marking its debut outside the US. The Crawley and London Docklands facilities, powered by Nvidia's H200 GPUs, are part of the company's $1.25 billion investment into AI in the UK, which coincides with an AI growth announcement made by the UK government. Valued at $19 billion, CoreWeave aligns with the UK's AI growth strategy to boost the country's private and public sectors in the AI industry. By 2025, the company plans to expand further into Europe with 10 new data centers, including sites in Norway, Sweden, and Spain.
Qventus, an AI startup that builds tools to automate work across the healthcare field, has recently raised $105 million in a Series D round led by KKR, with participation from Bessemer Venture Partners and Northwestern Medicine. The funding includes $85 million in equity and $20 million in optional debt to give a leg-up to its AI tools for automating tasks such as surgeries, discharges, and check-ups. Though not disclosed, Qventus is estimated to hover around a $400 million valuation and has grown its customer base fourfold as it looks to expand beyond in-patient applications and into broader healthcare operations.
|
Kerlig - Convey your voice better and more time efficiently in Slack, Figma, Gmail, Outlook, LinkedIn, and more with Kerlig, a lightweight AI writing assistant for macOS.
Keytopic - Turn long YouTube videos, tutorials, and podcasts into personalized, actionable summaries tailored to your professional role.
Dymium - Minimize or completely prevent data leaks with Dymium, a platform that boosts security posture across workflows and helps comply with regulatory requirements.
Senuto - Senuto covers a business's SEO needs with its expansive suite of data and content tools to help recognize your strengths and weaknesses in SEO.
|
Why AWS CEO Matt Garman Wants to Bet AWS on AI (68-min read / listen)
AWS CEO Matt Garman recently joined The Verge's editor-in-chief Nilay Patel in an episode of the Decoder podcast. They discussed Garman's rise to AWS chief from the very start as an intern, why he believes that AI is cloud computing's next big step, and how AWS is already seeing huge gains from its AI investments.
|
| What did you think of this newsletter? Let us know! |
|
|
|
|
|