AI: Weekly Summary. RTZ #416

Jul 13, 2024

OpenAI lays out AGI Levels: OpenAI is cleverly redefining the raging ‘time to AGI’ debate, laying out 5 levels to AGI. The Levels go from L1 ‘Chatbots’ (here now), to L2 ‘Reasoners’, L3 ‘Agents’, L4 ‘Innovators’ and L5 ‘Organizations’. No timing is specified, although as Ben’s Bites notes, OpenAI is intimating it is likely at the threshold of L2 presumably with GPT-5 later this year. This reframing will now likely be responded to by all other LLM AI companies, from Google to Meta to Anthropic and others. The 5 level rating system of course echoes the Government’s L5 rating system on self-driving cars, a topic I’ve discussed in depth. It also fits in with OpenAI Founder/CEO Sam Altman’s desire to transition to an ‘iterative’ approach to AI product and service innovation, and less a ‘Big Bang’ to AGI (Artificial General Intelligence), that has gripped the industry and regulators. Notable that ‘Agentic Workflows’ blend both L2 reasoning and L3 smart agents, and is a core focus for most AI companies already, as discussed here.

Microsoft and Apple opt out of OpenAI Board Observer Seats: Both Microsoft and Apple opted out of board Observer seats at OpenAI this week, almost concurrently. The trend is notable especially after the recent moves by big tech to do non-traditional AI ‘Acqui-hires’, with Microsoft/Inflection and Amazon/Adept being recent examples. Of course increased US Big Tech regulatory scrutiny is a likely driver, as well as the ‘Frenemies’ nature of the interactions amongst the companies. Trend likely to continue with other examples in the tech/AI industry. More here.

AI ‘Digital Twins’ race towards mainstream: Axios summarizes the latest AI ‘Digital Twin’ boom, outlining how both big tech and startups are focusing on the consumer and enterprise opportunities around ‘AI Digital Twins’. Also unique is the opportunity to use these innovations to generate Synthetic data for new AI training and inference. On the consumer side, VC/Tech Luminary Reid Hoffman’s experiments with a ‘virtual Reid’ point to the direction of the mainstream creator and consumer economies. Broader Media industry applications here also of note. More here.

AI Spend vs Returns debate returns: VC and Wall Street debates continue on AI capex spend vs timing of eventual returns. Both Sequoia, Goldman Sachs and others have detailed reports worth going through. Timing of investments vs returns seldom sync in big tech waves. AI likely no exception especially with unique headwinds of (a) Unprecedented early Market Fears of AI (b) Regulatory concerns globally (c) Multi-year AI GPU/data center/Power/Talent input shortages, (d) Extend times needed for Companies and Enterprises to find ‘Product-Market-Fit’, and (e) Unique bottleneck to source, secure, and scale new Data sources. This debate is likely to continue the rest of the year and beyond. Background on this debate from last year here.

EU extends regulatory focus to X/Twitter, beyond Apple, Nvidia, Meta et al: The EU continues its laser focus on US Big Tech companies, with its latest focus on x/Twitter, adding to its regulatory tussles with Meta, Google, Apple and others. Stratechery has a useful piece doing a deep dive, providing broader historical context. EU authorities ultimately ask for penalties from 6-10%+ of GLOBAL revenues in these cases. Observers point out that in the case of Meta, Apple and others, EU revenues are typically less than 10% of their global revenues. This trend continues to be a unique headwind for US big tech deploying next generation AI products and services in Europe. And an additional burden for smaller startups, or ‘Little Tech Agenda’ as articulated by VC a16z.

AI: Reset to Zero