Our surveys’ findings that AI researchers assign a median 5-10% to extinction or similar made a splash (NYT, NBC News, TIME..)
But people sometimes underestimate our survey’s methodological quality due to various circulating misconceptions.
Today, an FAQ correcting key errors:
I just wrapped up a deep-dive reanalysis of the famous survey of AI experts by @katjagrace & colleagues.
Let me walk you through some key takeaways (more nuance at the full report).
🧵...
New report, reanalyzing key results from the 2023 Expert Survey on Progress in AI. With in-depth discussion of methodology, excellent new charts, and open-sourced codebase, all thanks to @tmkadamcz!
How compelling are arguments that AI is an existential risk?
Here's a list of 10 such arguments risk I've collected with @nathanpmyoung
There is a poll beneath each, so you can vote on how compelling you find it.
Which best explains the political polarization of environmentalism?
- Small-government: The right dislikes any regulation
- Al Gore: Environmentalist alliance with the left
- Trust erosion: Blurring advocacy and science
- Non-specific: High polarization across many issues
Environmentalism is a partisan issue in the US, with Democrats being much more environmentalist than Republicans. Do you think this was inevitable, given the state of the world in (say) 1950? Or could it have been a bipartisan issue, or even had the opposite polarization?
We just ran probably the biggest survey of AI researchers ever! 2778 participants from six top AI venues answered questions from fourteen topics regarding the future of AI.
Preprint: https://t.co/VffsP9eflq
Six interesting things in pictures:
Other questions approached in the paper:
💵 What drives AI progress?
📈Has it sped up?
🐢Would it be better if it were slower or faster?
🔮 What will AI systems be like in 2043?
...
🥸 Will we know the reasons for AI choices by then?
🏭🏛️ Do people from academia and industry have different views?
🔀 Are concerns about AI due to misunderstandings?
🌏🌎 Do people who studied in Asia or US expect extinction more?
➰ Is the ‘alignment problem’ worth working on?
6. 70% of participants would like to see AI safety research prioritized more highly. "Much more" is way up since 2016. (Recall that this is a survey of AI researchers in general, not safety researchers.)
5. There are few confident optimists or pessimists about advanced AI: a majority of participants had both high hopes and dire concerns.
(Shown: random 800 participants' expectations as vertical columns ordered by optimism)
4. Many participants found many scenarios worthy of substantial concern over the next 30 years. For every one of 11 AI scenarios and ‘other’, over a third of participants found it deserving of substantial or extreme concern.
3. Median 5% or more on advanced AI leading to human extinction or similar. About a third to a half give over 10%, across four questions. (We often ask about a topic in different ways for robustness—here we get consistent answers.)
AI researchers give a 50% chance that AI can in 5y:
💻 make an entire payment processing site from scratch
🎤 generate a new song that sounds like it’s by Taylor Swift
🤖 autonomously download and fine-tune a large language model