🔮 Open-source AI surge; UBI surprises; OpenAI losses; Murdoch’s empire drama, waste-eating flies & learning ++ #484

An insider’s guide to AI and exponential technologies

Ideas of the week

Meta’s Gambit. Llama 3.1’s release has narrowed the gap between open-source and closed-source AI, matching proprietary models on key benchmarks.¹ Zuckerberg champions open-source AI, displaying altruism while executing a shrewd corporate strategy. Meta employs a “commoditise your complement” approach, a strategy to make LLMs more generic and accessible. Core AI models would be freely available which reduces their market value — and that of their competitors — while positioning Meta to profit from the surrounding ecosystem of tools, platforms and services that rely on these models. But can open-source models sustain this progress? Open-source must innovate. Open-source must scale. Open-source must find sustainable revenue. As AI progresses through two more generations, costs will soar (for more on this, see my essay on the scaling ceiling). Closed-source models can justify these expenses through the promise of direct monetisation, whereas Meta’s open-source strategy rel…