-
Oct 8, 2024 | 
lesswrong.com | Peter McCluskey |Nathan Helm-Burger |Roger Dearnaley |Start AtTheEnd
How can we make many humans who are very good at solving difficult problems? Summary (table of made-up numbers)I made up the made-up numbers in this table of made-up numbers; therefore, the numbers in this table of made-up numbers are made-up numbers. Call to actionIf you have a shitload of money, there are some projects you can give money to that would make supergenius humans on demand happen faster.
-
Jul 11, 2024 | 
lesswrong.com | Roger Dearnaley
Let's assume, for the sake of discussion, that Leopold Aschenbrenner is correct that a some point in the fairly near future (possibly even as he claims this decade, or later) AI will be capable of acing as a drop-in remote worker intelligent as the smartest humans and capable of doing basically any form of intellectual work that doesn't have in-person requirements, and that it can do so as well or better than pretty-much all humans, and around two or three orders of magnitude cheaper than...
-
Jul 5, 2024 | 
lesswrong.com | Roger Dearnaley
TL;DR:I discuss the challenge of aligning AGI/ASI, and outline an extremely simple approach to aligning an LLM: train entirely on a synthetic dataset that always shows the AI acting aligned (even when the humans behave badly), and use a conditional training/inference-time technique to lock the LLM into the AI role. Epistemic status: To me, this looks like an obvious thing to try.
-
Feb 16, 2024 | 
lesswrong.com | Roger Dearnaley
Imagine that you meet an 18th century altruist. They tell you “So, I’ve been thinking about whether or not to eat meat. Do you know whether animals have souls?” How would you answer, assuming you actually do want to be helpful? One option is to spend a lot of time explaining why “soul” isn’t actually the thing in the territory they care about, and talk about moral patienthood and theories of welfare and moral status.
-
Feb 15, 2024 | 
lesswrong.com | Roger Dearnaley
Part 7 of AI, Alignment, and Ethics. While this is fairly stand-alone, it may make more sense if you start with Part 1. TL;DR: At several points in this sequence (including Parts 1, 3, 4, and 6) I have suggested giving a privileged role in ethics or ethical thinking to evolved organisms and to arguments derived from Evolutionary Psychology.
-
Feb 14, 2024 | 
lesswrong.com | Roger Dearnaley
TL;DR It has been known for over a decade that that certain agent architectures based on Value Learning by construction have the very desirable property of having a basin of attraction to full alignment, where if you start sufficiently close to alignment they will converge to it, thereby evading the problem of "you have to get everything about alignment exactly right on the first try, in case of fast takeoff".
-
Jan 5, 2024 | 
lesswrong.com | Daniel Kokotajlo |Roger Dearnaley
We’ve been spending a lot of time recently thinking about how to mitigate risks posed by scheming (a.k.a. deceptively aligned) models intentionally subverting your safety techniques in order to cause an unacceptable outcome. We empirically investigated techniques that are robust to intentional subversion in our recent paper. In this post, we’ll discuss a crucial dynamic related to risk from scheming models.
-
Jan 4, 2024 | 
lesswrong.com | Lukas Finnveden |Roger Dearnaley
This is a series of posts with lists of projects that it could be valuable for someone to work on. The unifying theme is that they are projects that:Would be especially valuable if transformative AI is coming in the next 10 years or so. Are not primarily about controlling AI or aligning AI to human intentions. Most of the projects would be valuable even if we were guaranteed to get aligned AI. Some of the projects would be especially valuable if we were inevitably going to get misaligned AI.
-
Dec 31, 2023 | 
lesswrong.com | Vanessa Kosoy |Roger Dearnaley |Roman Leventov |Mateusz Bagiński
I call "alignment strategy" the high-level approach to solving the technical problem. For example, value learning is one strategy, while delegating alignment research to AI is another. I call "alignment metastrategy" the high-level approach to converging on solving the technical problem in a manner which is timely and effective. (Examples will follow.)In a previous article, I summarized my criticism of prosaic alignment. However, my analysis of the associated metastrategy was too sloppy.
-
Dec 18, 2023 | 
lesswrong.com | Roger Dearnaley |Linda Linsefors |Bogdan Ionut Cirstea |Neel Nanda
One of the primary concerns when attempting to control an AI of human-or-greater capabilities is that it might be deceitful. It is, after all, fairly difficult for an AI to succeed in a coup against humanity if the humans can simply regularly ask it "Are you plotting a coup? If so, how can we stop it?" and be confident that it will give them non-deceitful answers! TL;DR LLMs demonstrably learn deceit from humans.
