Co-founder @ https://t.co/e0JKSLzxVZ Prev CTO @metaculus, built Google's internal prediction market. Chess, jazz, biking, skiing, and learning.
Feb 27 • 11 tweets • 3 min read
Ok, this thread is long overdue. Now that everyone sees that GPT-4.5 is a disappointment -
How has OpenAI underperformed so much on the intelligence of their core models?
They've been faring worse and worse since the day GPT-4 came out. Buckle in for the tale:
First, ChatGPT in Nov 2022 and GPT-4 in Mar 2023 were amazing, world-changing innovations. Full credit there.
Shortly after GPT-4, forecasts of GPT-5's release began. The initial view was mid 2024. It has crept up and up and up:
Sep 12, 2024 • 7 tweets • 3 min read
Another paper, the fourth in 2024, claiming near or super-human level forecasting accuracy from AI, seen by hundreds of thousands of people.
1st, definitions matter. "Superhuman" forecasting should mean beating top Metaculus forecasters, or the overall crowd, not a random sample of them like this paper.
You wouldn’t say a coding AI was “superhuman” if it was top 50% of a coding competition. (2/7)
Apr 29, 2023 • 13 tweets • 3 min read
I bet I'm not the only one that has convos like this:
Me: LLMs are generational tech. I'm excited and terrified.
Them: You're worried about a Terminator / Kurzweil scenario?
Me: A bit. I'm more worried about chaos in the next 2-5 years.
Them: What exactly do you mean?
It's a good question, what am I worried about? Well, let me share a scenario.
It's April 2026. I wake up in the morning and check Hacker News. "Hundreds of Starlink satellites burn up in the atmosphere."
