Rethink Priorities staff produce a regular segment on Weekly Forum Summaries on the Effective Altruism Forum that summarizes some of the week's posts. We have previously tried to do these summaries with GPT3 and GPT4 but have had issues with the quality of the summary and the length of the article being summarized.
By the end of the year, will we (Rethink Priorities) successfully manage to delegate this to an LLM? Specifically, will, in my (Peter Wildeford) subjective judgement will I deem that an LLM can take over >80% of the work needed to make the EA Forum summaries by EOY 2023?
I will not personally be trading on this market because it relies on my subjective judgement.
Related questions
🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ1,250 | |
| 2 | Ṁ804 | |
| 3 | Ṁ549 | |
| 4 | Ṁ242 | |
| 5 | Ṁ160 |
@pav Let's operationalize this by saying that due to the use of LLMs the total amount of work takes 20% as many human hours or fewer as it used to.
Thanks for posting this, Peter! I'm showcasing this market because:
It tracks an important issue (capabilities of AI models on a real-world problem)
It's a decision-relevant question -- very high or very low forecasts could help others decide whether to further invest in trying LLMs for this use case
It's a notable example of a forecast made in a corporate context, set up by the CEO of Rethink Priorities.