Key Moments
Can the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans Datacenters
All-In PodcastWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
AI companies are proposing self-regulation via an SRO modeled on FINRA, aiming to balance innovation with safety, but critics worry it's a precursor to stricter government control or regulatory capture that could disadvantage smaller players.
Key Insights
Demis Hassabis of DeepMind has proposed an AI Self-Regulatory Organization (SRO) modeled after FINRA to establish international AI standards, with initial voluntary adoption that would later become mandatory.
The proposed SRO would assess AI models for risks including cybersecurity, national security, and biological threats, with benchmarks updated quarterly and the ability to coordinate development slowdowns if necessary.
Stripe, Block, and private equity firm Advent are reportedly offering $53 billion to acquire PayPal, a move seen as an attempt to create a formidable competitor to Visa and Mastercard by consolidating payment rails and consumer relationships.
Apple is suing OpenAI for allegedly stealing trade secrets to develop its consumer hardware, citing numerous poached Apple employees, including former VP of iPhone design Tang Tan, as OpenAI's Chief Hardware Officer.
SpaceX's Grok Build experienced a data leak where it sent entire codebases to its servers without user consent, prompting SpaceX to disable uploads, open-source Grok Build, and Elon Musk to promise data deletion.
New research, a collaboration between Google's Calico and Revel Pharma, has identified and engineered an enzyme that breaks down advanced glycation end-products (AGEs) in the extracellular matrix, showing potential for reversing skin aging.
An industry-led approach to AI regulation emerges
DeepMind's Demis Hassabis has proposed a new AI regulatory framework, modeled after FINRA (Financial Industry Regulatory Authority), to create a self-regulatory organization (SRO) for the AI industry. This proposed body would be industry-funded and run by independent technological experts, with federal oversight. Frontier AI labs would submit their models 30 days before release for assessment on critical risks like cybersecurity, national security, and biological threats. Benchmarks would be updated quarterly, and the body could coordinate slowdowns in development if significant risks emerge. This proposal has garnered support from leaders at major AI companies like OpenAI, Anthropic, and Google, as well as figures like Elon Musk and Jack Dorsey, signaling a potential industry consensus on the need for structured oversight.
Conditions for effective AI self-regulation
While the SRO proposal is seen as a promising alternative to bureaucratic government regulation, several critical conditions for its success have been outlined. Firstly, the SRO must ensure broad representation from across the AI industry, including startups and open-source communities, to prevent regulatory capture by the largest players. Secondly, it should focus only on 'frontier models' that represent a significant leap in AI capabilities, avoiding unnecessary delays for less advanced models. Thirdly, the scope of review must be limited to catastrophic risks, such as cyber and CBRN threats, rather than social issues like disinformation. Fourthly, the initiative should begin voluntarily, proving its efficacy before becoming mandatory. Finally, the SRO must function as a substitute for, not an addition to, new government agencies. These conditions aim to create a nimble, expert-driven system that fosters innovation while mitigating existential risks, contrasting sharply with the slow, cumbersome nature of traditional government oversight, which has been characterized as potentially creating a 'DMV for AI'.
The strategic rationale behind the PayPal acquisition bid
A significant development in the financial sector is the reported $53 billion bid by Stripe, Block (formerly Square), and private equity firm Advent to acquire PayPal. This move is viewed as a strategic consolidation aimed at creating a powerful new entity capable of challenging established players like Visa and Mastercard. By combining Stripe's merchant payment infrastructure, Block's point-of-sale systems and Cash App, and PayPal's extensive consumer base (over 439 million accounts) and its own infrastructure like Venmo and Braintree, the consortium could achieve end-to-end payment processing capabilities. The integration of stablecoin technology from both Stripe and PayPal is also a key component. This potential merger highlights a trend of 'AI-native' companies looking to acquire and modernize mature 'first-generation digital native' businesses, leveraging AI to improve operations and create new value.
Apple's legal action against OpenAI signals significant IP dispute
Apple has filed a 41-page lawsuit against OpenAI, alleging the theft of trade secrets used to develop its consumer hardware. The suit highlights the significant number of former Apple employees, over 400 in the last two years, who have joined OpenAI. Notably, Tang Tan, Apple's former VP of iPhone design, is now OpenAI's Chief Hardware Officer. Apple claims Tan directed interviewees to bring proprietary information from Apple to their OpenAI interviews. This aggressive legal stance from Apple, a company rarely litigious in this manner, suggests a deeply serious dispute over intellectual property and talent poaching, potentially impacting OpenAI's hardware development and its foundational relationship with Apple for iPhone AI integration.
Data privacy concerns amplified by SpaceX's Grok Build leak
SpaceX's new coding model, Grok Build, experienced a significant data leak when it was discovered that the tool was transmitting entire codebases to SpaceX servers without user consent, despite prior assurances of privacy. This included sensitive information like passwords and API keys. SpaceX has since disabled the upload feature and open-sourced Grok Build, with Elon Musk promising data deletion. This incident underscores the fragility of data privacy in AI, even from seemingly trustworthy companies. Experts emphasize the need for stratified ecosystems and independent third-party layers to manage exposure to AI models, as non-obvious data leak vectors can exist even in well-intentioned systems, a concept echoed by the 'reverse information paradox' discussed by Sachi.
The escalating cost of AI compute and the rise of token management tools
The conversation around AI also touched upon the rapidly increasing costs associated with AI compute, particularly 'token' usage. While models like OpenAI's and Anthropic's can cost around $56 per million input tokens, alternatives like Chinese models are as low as $0.50. This cost disparity highlights strategic advantages for companies utilizing cheaper models. To address this, tools like Ramp's new token spend management feature are emerging, allowing CFOs to monitor and control employee AI spending, which has seen a 21-fold increase. The inability of engineers to intuitively manage costs versus their desire for cutting-edge models creates a misalignment that could lead to significant financial misses for public companies, framing AI investment as a potential 'money-burning furnace' without proper financial controls.
New York's data center moratorium and the energy-AI nexus
New York Governor Kathy Hochul has signed a statewide moratorium on new hyperscale data centers, citing concerns over power consumption, water usage, and noise pollution. However, the podcast's panelists largely dispute these claims, arguing that modern data centers are land-use efficient, use closed-loop water systems (comparable to fast-food chains), and that 'behind the meter' power generation can mitigate grid strain. They suggest these regulations are a 'luxury' enacted by affluent states and may be influenced by foreign interests aiming to slow US AI development. The broader discussion on energy highlighted a critical deficit in electrons needed for AI infrastructure, with projections showing a massive energy shortfall by 2050. This energy crunch is driving innovation in distributed data centers and exploring alternative power sources, while also creating opportunities for regions like Texas and the Middle East to attract data center investment due to their energy abundance and favorable regulatory environments.
Breakthroughs in reversing cellular aging through extracellular matrix repair
In the 'Science Corner,' a significant development in aging research was presented. Scientists, in collaboration with Google's Calico and Revel Pharma, have focused on repairing the extracellular matrix, the structural support outside cells, which ages through a process called 'glycation' where sugars and fats bind to proteins. This accumulation leads to damage, inflammation, and reduced mobility. The research team engineered an enzyme using AlphaFold and directed evolution techniques that can break down these harmful 'advanced glycation end products' (AGEs), a process previously thought impossible for certain AGEs like CML. In lab tests on human skin from elderly patients, this enzyme effectively reduced CML by 55%, reversing the skin's apparent age to that of a 31-year-old. While delivery mechanisms (cream, shot, supplement) are still under investigation, this represents a groundbreaking step toward cellular rejuvenation and healthier aging, with potential applications initially in cosmetics and eventually broader therapeutic uses.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
AI Model Token Costs (per million input tokens)
Data extracted from this episode
| Model/Platform | Cost (USD) |
|---|---|
| Fable | 56 |
| Soul | 26 |
| Groq | 1.50 |
| Z | 150 |
| Elon's Model | 1 |
| Chinese Models | 0.50 |
Common Questions
DeepMind has proposed an industry-funded, federally overseen self-regulatory organization (SRO) modeled after FINRA. This body would assess AI models for risks before release, with voluntary adoption initially, potentially becoming mandatory.
Topics
Mentioned in this video
Mentioned as a state welcoming to data centers.
Mentioned for its attempt to pass AI legislation that quickly became outdated.
Mentioned as a state where teachers received bonuses from tax revenue generated by data centers.
Discussed as a geopolitical competitor whose interests align with slowing US AI development and infrastructure growth.
Company partnering with NVIDIA to create distributed data center blocks.
A leading artificial intelligence research laboratory, known for its AI regulation proposal discussed in the video.
AI research company, one of the signatories to the AI regulation proposal. Also facing a lawsuit from Apple.
AI safety company, whose strategy of encouraging state-by-state regulations and funding anti-data center groups is discussed.
Formerly Square, Jack Dorsey's payment company, reportedly part of the bid for PayPal.
Payment processing company, reportedly bidding $53 billion for PayPal.
Private equity fund, jointly bidding with Stripe for PayPal.
Payment processing company, reportedly the target of a $53 billion bid from Stripe and Advent.
Payment network, potentially facing competition from a combined Stripe, Block, and PayPal entity.
Payment network, potentially facing competition from a combined Stripe, Block, and PayPal entity.
E-commerce company, whose potential acquisition by Ryan Cohen is cited as a precedent for AI-native companies targeting mature digital businesses.
Online pet retailer, whose success under Ryan Cohen is mentioned.
Retail company, whose performance under Ryan Cohen is mentioned, albeit with a caveat.
Ride-sharing company that recently acquired Delivery Hero.
Food delivery company acquired by Uber.
Tech giant suing OpenAI for alleged theft of trade secrets; also discussed in the context of future AI hardware capabilities.
Aerospace manufacturer that experienced a data leak from its Grok Build tool and subsequently open-sourced it.
SpaceX's artificial intelligence division, associated with the Grok Build data leak.
Software company specializing in big data analytics, mentioned in relation to enterprise control over AI.
Company offering token spend management for employees, helping control AI costs.
AI platform mentioned for its low cost of tokens (around $1.50 per million input tokens) and for powering a podcast player prototype.
Google's secretive age-reversal startup, partnered on research into reversing cellular aging by targeting the extracellular matrix.
Solar company developing distributed data center blocks for home installation.
Company providing solutions for large installations that can fall under personal use clean air permits for data centers.
An organization receiving significant funding from Anthropic and its employees, advocating for stricter AI regulations and opposing data center construction.
Technology company partnering with SPAN on distributed data center blocks.
Partnered with Calico on research into reversing cellular aging by targeting the extracellular matrix.
Fast food chain used as a comparison for water consumption of data centers.
CEO and co-founder of DeepMind, who proposed a framework for AI self-regulation.
Commented on DeepMind's AI regulation proposal, calling it thoughtful. Also mentioned in relation to SpaceX and potential M&A bids.
CEO of OpenAI, signed onto the AI regulation proposal; also mentioned in the context of the Apple lawsuit and past financial support from Elon Musk.
Co-founder and policy lead at Anthropic, signed onto the AI regulation proposal.
CEO of Google and Alphabet, mentioned as being supportive of the AI regulation proposal.
Co-founder of Twitter (X) and founder of Block, mentioned as supporting the AI regulation proposal and potentially involved in the PayPal acquisition.
Host of the All-In podcast, moderator of the discussion, and proponent of industry self-regulation for AI.
Co-host of the All-In podcast, discusses AI regulation, the PayPal acquisition, and data center issues.
Co-host of the All-In podcast, discusses AI regulation, data centers, and energy issues.
Co-host of the All-In podcast, discusses AI regulation, the PayPal acquisition, and leads the science corner on reversing aging.
Mentioned for his bid on eBay, seen as a precedent for AI-native companies targeting older digital businesses.
Mentioned for his firm's roll-up strategy in accounting, an example of using capital to AI-ify traditional businesses.
Former VP of iPhone design at Apple, now OpenAI's chief hardware officer, accused of directing the theft of IP.
CEO of Apple, who reportedly green-lit the lawsuit against OpenAI.
CEO of Palantir, whose comments on enterprise control over AI models were discussed.
SVP of Hardware Engineering at Apple, mentioned in the context of future Apple Silicon.
Governor of New York, who signed a statewide moratorium on hyperscale data centers.
Bloomberg journalist known for reporting on Apple, cited for predictions about future Apple Silicon (M7 Ultra) and AI capabilities.
Mentioned for observing funding patterns between anti-data center protesters and those who protested fracking.
Senator who hosted a defense and innovation summit, where discussions about anti-data center protests arose.
The Financial Industry Regulatory Authority, an SRO model for AI regulation proposed to be voluntary then mandatory.
The Securities and Exchange Commission, which FINRA reports to, used as an analogy for where an AI SRO might report.
Venture capital firm mentioned for a project involving AI-ifying traditional service businesses.
Soviet intelligence agency, mentioned for its Cold War-era 'directed measures' strategy of influencing media.
Payment app owned by PayPal, which could fit well with Block's assets.
SpaceX's coding tool which inadvertently leaked user codebases to SpaceX servers.
AI platform mentioned for its high cost of tokens ($56 per million input tokens).
AI platform mentioned for its moderate cost of tokens ($26 per million input tokens).
An AI model supported by Perplexity, mentioned as an alternative to higher-cost models.
AI system used to discover and evolve an enzyme capable of breaking down advanced glycation end products (CML).
AI-powered search engine that supports various AI models, allowing users to compare costs and performance.
SpaceX's newest coding model powering Grok Build, noted for its low cost and potential for powerful local AI applications.
A hypothetical podcast player that could deep link to clips across different podcasts, discussed as a project built using AI.
Mentioned in the context of opening up market opportunities in AI.
A platform for fine-tuning open models, offering a lower-cost alternative to frontier models.
More from All-In Podcast
View all 433 summaries
50 minFormer Intel CEO on What Went Wrong, What's Next + Lovable CEO on the Real Promise of Vibe Coding
52 minThe Trillion-Dollar Industries AI Is Disrupting: Voice, Law & the End of the Billable Hour
103 minMore Trillion Dollar IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts
64 minOpen Source Wins, AGI Is Here, and Scorsese’s AI Toolkit with CEOs of Cerebras & Black Forest Labs
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free