Forum? I’m against ’em!
utilistrutil
Karma: 504
My main objection is that people working in government need to be able to get away with a mild level of lying and scheming to do their jobs (eg broker compromises, meet with constituents). AI could upset this equilibrium in a couple ways, making it harder to govern.
If the AI is just naive, it might do things like call out a politician for telling a harmless white lie, jeopardizing eg an international agreement that was about to be signed.
One response is that human overseers will discipline these naive mistakes, but the more human oversight is required, the more you run into the typical problems of human oversight you outlined above. “These evaluators can do so while not seeing critical private information” is not always true. (Eg if the AI realizes that Biden is telling a minor lie to Xi based on classified information, revealing the existence of the lie to the overseer would necessarily reveal classified information).
Even if the AI is not naive, and can distinguish white lies from outright misinformation, say, I still worry that it undermines the current equilibrium. The public would call for stricter and stricter oversight standards, while government workers will struggle to fight back because
That’s a bad look, and
The benefits of a small level of deception are hard to identify and articulate.
TLDR: Government needs some humans in the loop making decisions and working together. To work together, humans need some latitude to behave in ways that would become difficult with greater AI integration.
Thanks, Agustín! This is great.
Please submit more concrete ones! I added “poetic” and “super abstract” as an advantage and disadvantage for fire.
If the organization chooses to directly support the new researcher, then the net value depends on how much better their project is than the next-most-valuable project.
This is nit-picky, but if the new researcher proposes, say, the best project the org could support, it does not necessarily mean the org cannot support the second-best project (the “next-most-valuable project”), but it might mean that the sixth-best project becomes the seventh-best project, which the org then cannot support.
In general, adding a new project to the pool of projects does not trade off with the next-best project, it pushes out the nth-best project, which would have received support but now does not meet the funding bar. So the marginal value of adding projects that receive support depends on the quality of the projects around the funding bar.
Another way you could think about this is that the net value of the researcher depends on how much better this bundle of projects is than the next-most-valuable bundle.Essentially, this is the marginal value of new projects in AI safety research, which may be high or low depending on your view of the field.
So I still agree with this next sentence if marginal = the funding margin, i.e., the marginal project is one that is right on the funding bar. Not if marginal = producing a new researcher, who might be way above the funding bar.
These are beautiful!! Made my day :))
Update: We have finalized our selection of mentors.
I’ll be looking forward to hearing more about your work on whistleblowing! I’ve heard some promising takes about this direction. Strikes me as broadly good and currently neglected.
This is so well-written!
I’m cringing so hard already fr
Thanks for such a thorough response! I am also curious to hear Oscar’s answer :)
When applicants requested feedback, did they do that in the application or by reaching out after receiving a rejection?
Is that lognormal distribution responsible for
the cost-effectiveness is non-linearly related to speed-up time.
If yes, what’s the intuition behind this distribution? If not, why is cost-effectiveness non-linear in speed-up time?
Something I found especially troubling when applying to many EA jobs is the sense that I am p-hacking my way in. Perhaps I am never the best candidate, but the hiring process is sufficiently noisy that I can expect to be hired somewhere if I apply to enough places. This feels like I am deceiving the organizations that I believe in and misallocating the community’s resources.
There might be some truth in this, but it’s easy to take the idea too far. I like to remind myself:
The process is so noisy! A lot of the time the best applicant doesn’t get the job, and sometimes that will be me. I ask myself, “do I really think they understand my abilities based on that cover letter and work test?”
A job is a high-dimensional object, and it’s hard to screen for many of those dimensions. This means that the fact that you were rejected from one job might not be very strong evidence that you are a poor fit for another (even superficially similar) role. It also means that you can be an excellent fit in surprising ways: maybe you know that you’re a talented public speaker, but no one ever asks you to prove it in an interview. So conditional on getting a job, I think you shouldn’t feel like an imposter but rather eager to contribute your unique talents. My old manager was fond of saying “in a high-dimensional sphere, most of the points are close to the edge,” by which he meant that most people have a unique skill profile: maybe I’m not the best at research or ops or comms, but I could still be the best at (research x ops x comms).
Thanks for the references! Looking forward to reading :)
Fantastic, thanks for sharing!
Thanks for this! I would still be interested to see estimates of eg mice per acre in forests vs farms and I’m not sure yet whether this deforestation effect is reversible. I’ll follow up if I come across anything like that.
I agree that the quality of life question is thornier.
Under CP and CKR, Zuckerberg would have given higher credence to AI risk purely on observing Yudkowsky’s higher credence, and/or Yudkowsky would have given higher credence to AI risk purely on observing Zuckerberg’s lower credence, until they agreed.
Should that say lower, instead?
Decreasing the production of animal feed, and therefore reducing crop area, which tends to: Increase the population of wild animals
Could you share the source for this? I’ve wondered about the empirics here. Farms do support wild animals (mice, birds, insects, etc), and there is precedent for farms being paved over when they shut down, which prevents the land from being rewilded.
Would it make a difference if the risks were insured?