I think we're saying the same thing: you have to be working on something where, for one reason or another, you value the privacy. I think the o1 level flattened out about a year ago because no one is actually buying that level of intelligence anymore. You're either using the smartest frontier model or something that you can run locally, which itself would be smarter than o1 at this point.
I think we are having a moment of really strange economic incentives ahead of the IPO for both OpenAI and Anthropic. They are going to have massive revenues, but also huge liability potential. I think they are attempting to get some form of regulation in place, and in exchange for that regulation, they want limited liability for the bad deeds these models eventually will do. They would take regulation if they also got immunity or limited economic liability. That's why I think they are being very liberal with the disclosures.
@Sentdex At minimum, I'm finding that it does things "as well as" certain LLM use cases, at a fraction of the cost and far more quickly. Not sure that Jev has anything special about it (alternatives already seem to be available), but the idea will end up sticking.
@Jason@grok Based on the best available data, would both Puerto Rico and Washington, D.C. want to become official states? If so, why has this not yet happened? what other U.S. territories would even want to become a U.S. state? Keep your answer brief.
@Sentdex I've had surprisingly good results with "off" for coding tasks using Flash models, particularly if you have some sort of well structured spec before it starts coding. Is that something you've actively varied to see if it changes output quality? They can be extremely chatty.
@Teknium Thanks for sharing. Funny enough, I was so interested in what you’d done, that I had pointed an agent at your repo to write up and reverse engineer your process. Was hoping you’d release something like this. May I DM you with a few questions where answers were non-obvious?
@zainhas It is absolutely cracked, but it also completely redefined the economics of the Pareto frontier. I got curious about this the other day and didn't know how big of a difference it actually made. Crazy good.
https://t.co/MUYkTdAjnZ
Movement in the AI model Pareto frontier over the last 29 months is incredible. I decided to collect and visualize the data. Here's the entire frontier over the past 2+ years. https://t.co/8NWYyB4YVn