Grant Slatton Profile picture
Formerly built the fastest filesystem in the world at AWS, now building the fastest spreadsheet at https://t.co/ojwoGUGC8P
Sep 18, 2023 7 tweets 3 min read
The new GPT model, gpt-3.5-turbo-instruct, can play chess around 1800 Elo.

I had previously reported that GPT cannot play chess, but it appears this was just the RLHF'd chat models. The pure completion model succeeds.



See game & thoughts below:
Image The new model readily beats Stockfish Level 4 (1700) and still loses respectably to Level 5 (2000). Never attempted illegal moves. Used clever opening sacrifice, and incredibly cheeky pawn & king checkmate, allowing the opponent to uselessly promote.

lichess.org/K6Q0Lqda

Image
Image
Dec 11, 2022 7 tweets 3 min read
GPT can iteratively write, debug, and test programs to accomplish arbitrary goals.

Pictured: GPT reading snippets of HTML from HN and building a headline scraper in Python, overcoming bugs by simply reading the errors and self-judgments and hypothesizing to itself.

Thread ↓

As shown in my previous post, GPT can be embedded in a REPL to accomplish goals in an agent-based fashion by using command line tools. Today's post shows it is capable of creating novel command line tools, such as a web scraper.

1/
Dec 8, 2022 10 tweets 3 min read
GPT can use a web browser to answer questions.

When embedded in a REPL environment and prompted to strategize and monologue, agent-like behavior emerges. The agent can solve multi-step problems that involve going to pages, following links, reading the next page, etc.

Thread ↓ ImageImageImage The real HN pages showing those answers: ImageImage