Sam Altman was wrong about AI | Eli the Computer Guy
Summary
Isaac and Eli the Computer Guy argue that OpenAI is failing due to Sam Altman’s rigid "bigger is better" thesis and unsustainable infrastructure contracts. They contrast this with Anthropic’s focus on practical tools and a growing trend of companies like Thomson Reuters building bespoke AI models on proprietary data. The discussion suggests the future of AI lies in specialized, fine-tuned solutions rather than general frontier models, potentially leaving OpenAI as mere infrastructure while Anthropic becomes a utility provider.
Topics discussed
Sponsors: Liquid IV and Midas Tire Guarantee
Intro: OpenAI's future and the rise of open-weight models
OpenAI vs. Anthropic: Frontier models vs. practical utility
Sam Altman's thesis: Why 'bigger is better' may be a mistake
First mover advantage myth and OpenAI spawning competitors
Thinking Machines: Customizable base models and fine-tuning
Thomson Reuters: Building a fiduciary-grade LLM with proprietary data
Business viability: Revenue vs. massive infrastructure contracts
Specialized AI: Fidelity and IBM using niche datasets
Data value shift: From selling data to building internal products
Orchestration layers and the difficulty of licensing quality data
Routing systems: Directing queries to specialized models
OpenAI's instability: Executive turnover and IPO concerns
Anthropic's valuation and potential as a utility provider
Future roles: Model developers as cloud infrastructure
Outro and final sponsor messages
Listen ad-free on Castria