I really like learning new things!
Jacob G-W
Karma: 977
Yep, my main thoughts on why its important to work on understanding current models are basically: even if these things do not have any risk from becoming unaligned or anything like that, do we really want to base much of our economy on things that we don’t really understand very well?
Here’s the podcast (should be on any podcast app): https://podcasts.apple.com/us/podcast/rationality-from-ai-to-zombies/id1299826696
This makes sense. I’ve changed my mind, thanks!
Given that a podcast already exists, I think you might get more bang for your buck if you did some animation on top of it. Otherwise, the only thing you are adding is putting it on youtube and having a camera of your face. This would probably be (much) harder, but also probably much higher reward if it worked.
Maybe a collaboration with rationalanimations would help? Not really sure, but good luck if you try to do this!
p.s. I could totally see an advanced alien playing Elon Musk :P
You need to use conservation of expected evidence. You can’t say something is evidence against the simulation evidence without saying what crazy event would need to happen to provide evidence for the simulation hypothesis.
A lot of crazy stuff is happening in our world. (Elon Musk, political figures, whatever) If lack of one crazy thing is evidence against hypothesis, then existence of crazy things must be evidence for the hypothesis. If you only see it one way, you violate the law of conservation of expected evidence.
Yeah, it feels like an accurate and succinct description of what hippies are (anecdote: my grandfather was at Woodstock and he’s pretty cool). Not saying I endorse it, but there are certainly some good aspects.
I’m not sure and it would be interesting to find out if being a hippy seriously messes up a fraction of people who try it and we just don’t hear about it due to selection bias. My guess is that this happens, especially with drugs.
I like the description of hippies as people who iterate really fast on psychological practices!
I’m a bit confused on how boxing an AI would be useful (to an extreme extent). If we don’t allow any output bits to come out of the ASI, then how do we know if it worked? Why would we want to run it if we can’t see what it does? Or do we only want to limit the output to bits and prevent any side-channel attacks? I guess the theory then would be that bits are not enough to destroy the world. Like maybe for , it would not be enough to persuade a person to do something that would unbox the AI (but it might).
This seems to be one of the only solutions in which such a proof can exist. If we were to use another solution, like changing the superintelligence’s objectives, then finding such a proof would be extremely hard, or even impossible. However, if we think that we could all die by making a superintelligence, then we should have an unconditional proof of safety.
I don’t think having a formal proof should be an objective in and of itself. Especially if the proof is along the lines “The superintelligence has to be boxed because it can only run encrypted code and can’t communicate with the outside world”
I’m sorry if this comment sounds overly negative, and please let me know if I am interpreting this post wrong. This work seems quite interesting, even just for computer science/cryptography’s sake (although saving the world would also be nice :)).
Thanks for the update! I have a few questions:
In last year’s update, you suspected that alignment was gradually converging towards a paradigm. What do you think is the state of the paradigmatic convergence now?
Also as @Chris_Leong asked, does using sparse autoencoders to find monosemantic neurons help find natural abstractions? Or is that still Choosing The Ontology? What, if not these types of concepts, are you thinking natural abstractions are/will be?
This is super interesting and I have a question:
How difficult would it be to also apply this to the gamates and thus make any potential offspring also have the same enhanced intelligence (but this time it would go into the gene pool instead of just staying in the brain)? Does the scientific establishment think this is ethical? (Also, if you do something like this, you reduce the homogeneity of the gene pool which could make the modified babies very susceptible to some sort of disease. Would it be worth it to give the GMO babies a random subset of the changes to increase variation?)
Yes, this is pretty much how I see trust. It is an abstraction over how much I would think that the other person will do what I would want them to do.
Trusting someone means that I don’t have to double-check their work and we can work closer and faster together. If I don’t trust someone to do something, I have to spend much more time verifying that the thing that they are doing is correct.
Sure, but they only use 16 frames, which doesn’t really seem like it’s “video” to me.
Understanding video input is an important step towards a useful generalist agent. We measure the video understanding capability across several established benchmarks that are held-out from training. These tasks measure whether the model is able to understand and reason over a temporally-related sequence of frames. For each video task, we sample 16 equally-spaced frames from each video clip and feed them to the Gemini models. For the YouTube video datasets (all datasets except NextQA and the Perception test), we evaluate the Gemini models on videos that were still publicly available in the month of November, 2023
It seems to do something similar to Gato where everything is just serialized into tokens, which is pretty cool
I wonder if they are just doing a standard transformer for everything, or doing some sort of diffusion model for the images inside the model?
Update, it seems that the video generation capability is just accomplished by feeding still frames of the video into the model, not by any native video generation.
Google Gemini Announced
Ah, sorry for the confusion. Thanks!
It should probably say 2023 review instead of 2022 at the top of lesswrong.
Same!
@lsusr recently did a video about this. Interestingly, he thought that the hardest people to love were not actually the Hitler type (they are still hard), but the people that you are actively hurting.