Should we have a government-funded “public option” for AI?
Should the government get into making AI? That's what the authors of an interesting new piece in Slate Magazine argue.
💡 The idea: have the government develop AI (or fund its development), and make the resulting tech available to everyone.
Importantly, the authors are NOT proposing the government would maintain sole control over government-funded AI. That would be frightening.
A few selling points:
Could help prevent a future where a small number of powerful companies control AI
Could produce models better aligned to the public interest, rather than corporate interests
Could improve transparency – we would know exactly how these models were trained, and on what data
Could lead to development of better safeguards
Some questions and concerns I have:
Who would decide which models the government would fund?
Would such a centrally planned process be able to keep up with the rapid pace of change in this field?
Who would decide how these models would be trained? Decisions about data usage, training objectives, and alignment to human intentions are not straightforward, and can profoundly affect the resulting models.
Would this become a new battleground in the culture wars, with people accusing the government models of being too woke, or not woke enough?
Would corporate interests end up influencing government choices? Would that be good or bad?
Helping small businesses compete with Big Tech seems good, but how would we prevent this from becoming a new form of corporate welfare?
Some really unconvincing stuff also in the article
The authors also suggest that government-funded AI could be deployed to improve democracy. I find this part of the article ridiculous.
For example, they suggest large language models (LLMs) could be used to summarize online debates, as a kind of new form of polling. While this sounds doable, I don’t think it represents a step forward for democracy.
They also suggest carrying out a pilot project where LLMs moderate local town hall debates about budget proposals. Really?
These suggestions are comical, and in my view they’re a distraction from the main point. Government funding of AI is an idea worth considering, even if we never use AI directly in the democratic process.
What do you think? Is a “public option” a good idea?