It’s a work in progress that we’ll announce later. Stay tuned!
Harlan
Karma: 44
Hi, I’m one of the new(ish) team members! You’re right that this page is out of date- we’re working on fixing this as part of a broader overhaul of the website. In the meantime, you can see some info about the new hires in our newsletters from April, May, June, and September (as well as a forthcoming December newsletter which will announce another hire).
Hi, I’m part of the communications team at MIRI.
You’re right that this post is light on details. Much of our energy over the past year has gone into “revving up.” That includes hiring and spinning up new teams, moving into a larger office space, and working on “rock content” comms projects which will be announced later. In the meantime, some places to check out our public-facing output include our newsletters, TGT’s new website, and a forthcoming post with more details about the media stuff we’ve done recently.
Regarding the donor question, that’s totally fair! While we welcome any donations now, we think it’s reasonable for donors to wait until we make a proper call for funds next year.
Great post!
“All that is stopping them being even more powerful is spending on compute. Google & Microsoft are worth $1-2T each, and $10B can buy ~100x the compute used for GPT-4. Think about this: it means we are already well into hardware overhang territory[5].”
I broadly agree with the point that compute could be scaled up significantly, but I want to add a few notes about the claim that $10B buys 100x the compute of GPT-4.
Altman said “more” when asked if GPT-4 had cost $100M to train. We don’t know how much more. But PaLM seems to have only cost $9M-$23M so $100M is probably reasonable.
If OpenAI was buying up 100x the compute of GPT-4, maybe that would be a big enough spike in demand for GPUs that they would become more expensive. I’m pretty uncertain about what to expect there, but I estimated that PaLM used the equivalent of 0.01% of the world’s current GPU/TPU computing capacity for 2 months. GPT-4 seems to be bigger than PaLM, so 100x the compute used for it might be the equivalent of more than 1% of the world’s existing GPU/TPU computing capacity for 2 months.
You probably agree with me that (a) we can’t know whether it will rain on 2/10/2050 and (b) we can be pretty sure that there will be a solar eclipse on 7/22/2028. You are actively participating in a prediction market, so you seem to believe in some ability to forecast the future better than a magic 8 ball.
Where do you think the limits are to what kinds of things we can make useful predictions about, and how confident those predictions can be?
Thanks for this great writeup, this seems like a topic that deserves more discussion.
“Coming into physical contact with grabby aliens within the next, say, 1000 years is very unlikely. The reason for this is that grabby aliens have existed, on average, for many millions of years, and thus, the only way we will encounter them physically any time soon is if we happened to right now be on the exact outer edge of their current sphere of colonization, which seems implausible.”
Or encounters with grabby aliens are so dangerous that they don’t leave any conscious survivors, in which case we would need to be careful about using the lack of encounters in our past as evidence about the likelihood or frequency in the future.
Hi, I’m part of the communications team at MIRI. Here’s a very high-level summary of what MIRI is currently doing:
Our research portfolio includes the new Technical Governance Team, as well as some alignment research (though much less than before).
We have also spun up a comms team. We think of our comms work in terms of “rock content” and “wave content.” Currently more effort is going into “rock content” projects which will be announced later.
We also do some work in DC, though this is limited by our status as a 501(c)(3).
MIRI’s strategy update from earlier this year explains the reasoning behind our shift from primarily doing technical alignment research to focusing more on communications and policy. The actual work of making that shift is a lot of why 2024 looked quieter, from the outside, than 2023 (and than our hopes for 2025).