There's some really cool work being done using Zero Knowledge proofs to write a succinct proof that output from a given model is correct. This is going to be in

There's some really cool work being done using Zero Knowledge proofs to write a succinct proof that output from a given model is correct. This is going to be increasingly important not just for these types of distributed systems, but even for things like ChatGPT to make sure that you're actually getting results from the model that you're paying to use. Imagine a world where OpenAI or some other large API provider gets taken over by someone who wants to make money, so they start quietly using smaller, weaker models to respond to API calls, even for customers who are paying for the highest end model. Maybe this is just done at first to survive under massive load, but then someone realizes how many millions of dollars they can save by just forgetting to switch back to the more expensive models. Here's a great primer: https://0xparc.org/blog/zk-mnist https://0xparc.org/blog/zk-mnist