Google's I/O 2023 suggests that AI is a sustaining innovation for Big Tech; the true fight will be between the major players' centralized models and open source
StratecheryBen Thompson
Context & Ripple Effects
Days before the keynote, a leaked internal memo by Luke Sernau argued open source would outcompete Google and OpenAI from below — the argument that escaped Google's walls frames precisely the contest I/O's thesis names: centralized models versus the open ecosystem.
The open side of that contest has its own weakness, flagged the day before I/O: the open-source AI boom runs on Big Tech's released weights like LLaMA and GPT-3, so Meta and OpenAI hold the kill switch. I/O reading AI as a sustaining innovation matters because it says the incumbents intend to keep owning the distribution layer while the real fight moves to the model layer.
First-order effects
Framing AI as a sustaining innovation casts Google's I/O announcements as reinforcements for Search, Android, and Assistant distribution rather than disruptions of them — the incumbents defend their franchises instead of being displaced.
For Google and OpenAI, the immediate consequence is that their moat is redefined: no longer model quality alone, but the durability of centralized, hosted models against freely available alternatives.
Second-order effects
Because the open-source boom depends on weights Meta and OpenAI choose to release, those two companies gain de facto control over their open competitors' roadmap — closing releases would collapse the ecosystem overnight.
Later Stratechery analysis sharpens the stakes: Google's structural advantages make it a threat to both OpenAI, which must turn its massive ChatGPT user base into revenue via ads, and Nvidia.
Third-order effects
If the pattern holds, the industry structurally bifurcates into incumbent centralized platforms and an open ecosystem living on incumbents' discarded weights — an asymmetry where Big Tech loses nothing by open-sourcing yesterday's models.
Google's subsequent pivot toward prioritizing AI diffusion over frontier-model leadership reads as confirmation of the sustaining-innovation thesis: value accrues to whoever distributes AI at scale, not necessarily to whoever holds the top model.
The trend: Big Tech is absorbing AI as a sustaining innovation that reinforces existing distribution, while the decisive competitive axis shifts from model quality to centralized platforms versus open source.
A new European AI regulation proposal would make any “American opensource developer” that hosts an “unlicensed LLMs” on GitHub & available in Europe liable for “€20,000,000 or 4% of worldwide revenue” https://technomancers.ai/... [image]
AI Act: “The amended act, voted out of committee on Thursday, would sanction American open-source developers and software distributors, such as GitHub, if unlicensed generative models became available in Europe.” https://technomancers.ai/...
Google I/O and the Coming AI Battles Google A/I suggests that AI is a sustaining innovation for all of Big Tech; that means the real battle will be between incumbents and Big Tech on one side, and open source on the other. https://stratechery.com/...
The only appropriate response to the EU AI Act by U.S. developers is disobedience and walling off of European IPs. The EU knows they're behind, so rather than enabling their own entrepreneurs they're trying to slow down everyone else. Won't work. https://twitter.com/...
“To that end, I take this Google I/O as evidence that AI is in fact a sustaining technology for all of Big Tech, including Google.” https://stratechery.com/... via @stratechery
The EU's AI Act would require models to pass extensive and expensive licensing. Opensource developers and hosts like GitHub would be liable for unlicensed models. A failure to comply could result in fines of €20 million or 4% worldwide revenue. https://technomancers.ai/...
The EU will be a generational laggard in this critical area of technology. The bureaucrats in Brussels have a singular focus on suppressing technology innovation, benefiting the large multinationals they believe they can bully into compliance. https://technomancers.ai/...
Until recently it was claimed that Google's access to data would give it an unassailable advantage in developing AI. Does anyone still believe this? https://stratechery.com/...
“Some implementation examples include AutoGPT and LangChain. Under these rules, if a third party, using an API, figures out how to get a model to do something new, that third party must then get the new functionality certified.” https://twitter.com/...
Many AI doomers would celebrate this as “a good start” if it passes. This isn't a joke, we are at real risk of letting some LessWrong posters smother a technology in the cradle that could end untold suffering based on pure speculation and conjecture. https://technomancers.ai/...