lukemarks

Steer­ing Lan­guage Models in Mul­ti­ple Direc­tions Simultaneously

• lukemarksSep 19, 2024, 11:22 PM

  17 points

  10

  on: RLHF is the worst pos­si­ble thing done when fac­ing the al­ign­ment problem

  I don’t think the point of RLHF ever was value alignment, and I doubt this is what Paul Christiano and others intended RLHF to solve. RLHF might be useful in worlds without capabilities and deception discontinuities (plausibly ours), because we are less worried about sudden ARA, and more interested in getting useful behavior from models before we go out with a whimper.

  This theory of change isn’t perfect. There is an argument that RLHF was net-negative, and this argument has been had.

  My point is that you are assessing RLHF using your model of AI risk, so the disagreement here might actually be unrelated to RLHF and dissolve if you and the RLHF progenitors shared a common position on AI risk.

• lukemarksJul 31, 2024, 12:08 AM

  1 point

  0

  on: François Chol­let on the limi­ta­tions of LLMs in reasoning

  I don’t understand why Chollet thinks the smart child and the mediocre child are doing categorically different things. Why can’t the mediocre child be GPT-4, and the smart child GPT-6? I find the analogies Chollet and others draw in an effort to explain away the success of deep learning sufficient to explain what the human brain does, and it’s not clear a different category of mind will or can ever exist (I don’t make this claim, I’m just saying that Chollet’s distinction is not evidenced).

  Chollet points to real shortcomings of modern deep learning systems, but these are often exacerbated by factors not directly relevant to problem solving ability such as tokenization, so often I take them more lightly than I estimate he does.

• lukemarksJul 29, 2024, 1:18 AM

  1 point

  0

  in reply to: Lucius Bushnaq’s comment on: luke­marks’s Shortform

  That is closer to what I meant, but it isn’t quite what SLT says. The architecture doesn’t need to be biased toward the target function’s complexity. It just needs to always prefer simpler fits to more complex ones.

  This why the neural redshift paper says something different to SLT. It says neural nets that generalize well don’t just have a simplicity bias, they have a bias for functions with similar complexity to the target function. This brings into question mesaoptimization, because although mesaoptimization is favored by a simplicity bias, it is not necessarily favored by a bias toward equivalent simplicity to the target function.

• lukemarksJul 28, 2024, 1:43 PM

  2 points

  0

  in reply to: Lucius Bushnaq’s comment on: luke­marks’s Shortform

  I think the predictions SLT makes are different from the results in the neural redshift paper. For example, if you use tanh instead of ReLU the simplicity bias is weaker. How does SLT explain/​predict this? Maybe you meant that SLT predicts that good generalization occurs when an architecture’s preferred complexity matches the target function’s complexity?

  The explanation you give sounds like a different claim however.

  If you go to a random point in the loss landscape, you very likely land in a large region implementing the same behaviour, meaning the network has a small effective parameter count

  This is true of all neural nets, but the neural redshift paper claims that specific architectural decisions beat picking random points in the loss landscape. Neural redshift could be true in worlds where the SLT prediction was either true or false.

  We already knew neural network training had a bias towards algorithmic simplicity of some kind, because otherwise it wouldn’t work.

  We knew this, but the neural redshift paper claims that the simplicity bias is unrelated to training.

  So we knew general algorithms, like mesa-optimisers, would be preferred over memorised solutions that don’t generalise out of distribution.

  The paper doesn’t just show a simplicity bias, it shows a bias for functions of a particular complexity that is simpler than random. To me this speaks against the likelihood of mesaoptimization, because it seems unlikely a mesaoptimizer would be similar in complexity to the training set if that training set did not describe an optimizer.

• lukemarksJul 28, 2024, 8:17 AM

  18 points

  5

  on: luke­marks’s Shortform

  Neural Redshift: Random Networks are not Random Functions shows that even randomly initialized neural nets tend to be simple functions (measured by frequency, polynomial order and compressibility), and that this bias can be partially attributed to ReLUs. Previous speculation on simplicity biases focused mostly on SGD, but this is now clearly not the only contributor.

  The authors propose that good generalization occurs when an architecture’s preferred complexity matches the target function’s complexity. We should think about how compatible this is with our projections for how future neural nets might behave. For example: If this proposition were true and a significant decider of generalization ability, would this make mesaoptimization less likely? More likely?

  As an aside: Research on inductive biases could be very impactful. My impression is that far less resources are spent studying inductive biases than interpretability, but inductive bias research could be feasible on small compute budgets, and tell us lots about what to expect as we scale neural nets.

• lukemarksJul 27, 2024, 11:00 AM

  13 points

  1

  in reply to: the gears to ascension’s comment on: luke­marks’s Shortform

  I dropped out one month ago. I don’t know anyone else who has dropped out. My comment recommends students consider dropping out on the grounds that it seemed like the right decision for me, but it took me a while to realize this was even a choice.

  So far my experience has been pleasant. I am ~twice as productive. The total time available to me is ~2.5-3x as much as I had prior. The excess time lets me get a healthy amount of sleep and play videogames without sacrificing my most productive hours. I would make the decision again, and earlier if I could.

• lukemarksJul 27, 2024, 5:03 AM

  21 points

  −1

  on: luke­marks’s Shortform

  More people should consider dropping out of high school, particularly if they:

  • Don’t find their classes interesting

  • Have self-motivation

  • Don’t plan on going to university

  In most places, once you reach an age younger than the typical age of graduation you are not legally obligated to attend school. Many continue because it’s normal, but some brief analysis could reveal that graduating is not worth the investment for you.

  Some common objections I heard:

  • It’s only $n$ more months, why not finish?

  Why finish?

  • What if ‘this whole thing’ doesn’t pan out?

  The mistake in this objection is thinking there was a single reason I wanted to leave school. I was increasing my free time, not making a bet on a particular technology.

  • My parents would never consent to this.

  In some cases this is true. You might be surprised if you demonstrate long term commitment and the ability to get financial support though.

  Leaving high school is not the right decision for everyone, but many students won’t even consider it. At least make the option available to yourself.

• lukemarksJul 2, 2024, 9:08 AM

  1 point

  −2

  in reply to: robo’s comment on: luke­marks’s Shortform

  This should be an equivalent problem, yes.

• lukemarksJul 2, 2024, 9:07 AM

  1 point

  0

  in reply to: robo’s comment on: luke­marks’s Shortform

  Yes, your description of my hypothetical is correct. I think it’s plausible that approximating things that happened in the past is computationally easier than breaking some encryption, especially if the information about the past is valuable even if it’s noisy. I strongly doubt my hypothetical will materialize, but I think it is an interesting problem regardless.

  My concern with approaches like the one you suggest is that they’re restricted to small parts of the universe, so with enough data it might be possible to fill in the gaps.

• lukemarksJul 2, 2024, 6:56 AM

  3 points

  −11

  on: luke­marks’s Shortform

  Present cryptography becomes redundant when the past can be approximated. Simulating the universe at an earlier point and running it forward to extract information before it’s encrypted is a basic, but difficult way to do this. For some information this approximation could even be fuzzy, and still cause damage if public. How can you protect information when your adversary can simulate the past?

  The information must never exist as plaintext in the past. A bad way to do this is to make the information future-contingent. Perhaps it could be acausally inserted into the past by future agents, but probably you would not be able to act on future-contingent information in useful ways. A better method is to run many homomorphically encrypted instances of a random function that might output programs that do computations that yield sensitive information (e.g, an uploaded human). You would then give a plaintext description of the random function, including a proof that it probably output a program doing computations likely adversaries would want. This incentivizes the adversary to not destroy the program output by the random function, because it may not be worth the cost of destruction and replacement with something that is certainly doing better computations.

  This method satisfies the following desiderata:
  1. The adversary does not know the output of the encrypted random function, or the outputs of the program the random function output
  2. There is an incentive to not destroy the program output by the random function

  One problem with this is that your advserary might be superintelligent, and prove the assumptions that made your encryption appear strong to be incorrect. To avoid this, you could base your cryptography on something other than computational hardness.

  My first thought was to necessitate computations that would make an adversary incur massive negative utility to verify the output of a random function. It’s hard to predict what an adversary’s preferences might be in advance, so the punishment for verifying the output of the random function would need to be generically bad, such as forcing the adversary to expend massive amounts of computation on useless problems. This idea is bad for obvious reasons, and will probably end up making the same or equally bad assumptions about the inseparability of the punishment and verification.

luke­marks’s Shortform

• lukemarksMay 25, 2024, 12:36 PM

  1 point

  0

  in reply to: Chris_Leong’s comment on: Beta Tester Re­quest: Ral­ly­point Bounties

  We are recruiting people interested in using Rallypoint in any way. The form has an optional question for what you hope to get out of using Rallypoint. Even if you don’t plan on contributing to bounties or claiming them and just want to see how others use Rallypoint we are still interested in your feedback.

• lukemarksMay 25, 2024, 9:53 AM

  2 points

  0

  in reply to: keltan’s comment on: Beta Tester Re­quest: Ral­ly­point Bounties

  Yes. If the feedback from the beta is that people find Rallypoint useful we will do a public release and development will continue. I want to focus on getting the core bounty infrastructure very refined before adding many extra features. Likely said infrastructure would be easily adapted to more typical crowdfunding and a few other applications, but that is lower on the priority list than getting bounties right.

Beta Tester Re­quest: Ral­ly­point Bounties

• lukemarksDec 28, 2023, 2:04 PM

  LW: 2 AF: 2

  0

  AF

  on: Free agents

  I don’t understand the distinction you draw between free agents and agents without freedom.

  If I build an expected utility maximizer with a preference for the presence of some physical quantity, that surely is not a free agent. If I build some agent with the capacity to modify a program which is responsible for its conversion from states of the world to scalar utility values, I assume you would consider that a free agent.

  I am reminded of E.T. Jaynes’ position on the notion of ‘randomization’, which I will summarize as “a term to describe a process we consider too hard to model, which we then consider a ‘thing’ because we named it.”

  How is this agent any more free than the expected utility maximizer, other than for the reason that I can’t conveniently extrapolate the outcome of its modification of its utility function?

  It seems to me that this only shifts the problem from “how do we find a safe utility function to maximize” to “how do we find a process by which a safe utility function is learned”, and I would argue the consideration of the latter is already a mainstream position in alignment.

  If I have missed a key distinguishing property, I would be very interested to know.

• lukemarksNov 8, 2023, 4:58 AM

  3 points

  −7

  on: Con­crete pos­i­tive vi­sions for a fu­ture with­out AGI

  I believe you misinterpreted the quote from disturbance. They were implying that they would bring about AGI at the moment before their brain would be unsalvageable by AGI such that they could be repaired, assumedly in expectation of immortality.

  I also don’t think the perspective that we would likely fail as a civilization without AGI is common on LessWrong. I would guess that most of us would expect a smooth-ish transition to The Glorious Future in worlds where we coordinate around [as in don’t build] AI. In my opinion the post is good even without this claim however.

• lukemarksNov 4, 2023, 7:55 AM

  4 points

  −1

  on: Un­trusted smart mod­els and trusted dumb models

  models that are too incompetent to think through deceptive alignment are surely not deceptively aligned.

  Is this true? In Thoughts On (Solving) Deep Deception, Jozdien gives the following example that suggests otherwise to me:

  Back in 2000, a computer scientist named Charles Ofria was studying the evolution of simulated organisms. He wanted to limit their replication rate, so he programmed the simulation to pause after each mutation, measure the mutant’s replication rate in an isolated test environment, and delete the mutant if it replicated faster than its parent.

  Eventually, organisms evolved to recognize and pause their replication in the test environment and resume normal replication when released. Ofria tried solving this by making it harder to detect the test environment, but the organisms simply started to randomly accelerate their replication so as to slip through the test environment some of the time.

  I don’t see why a model would need to be cognitively able to process its own alignment for its alignment to be improper, and I think this assumption is quite key to the main claims of the post.

[Question] Shouldn’t we ‘Just’ Su­per­im­i­tate Low-Res Uploads?

• lukemarksOct 26, 2023, 6:03 AM

  1 point

  0

  on: AI as a sci­ence, and three ob­sta­cles to al­ign­ment strategies

  unless, by some feat of brilliance, this civilization pulls off some uncharacteristically impressive theoretical triumphs

  Are you able to provide an example of the kind of thing that would constitute such a theoretical triumph? Or, if not; a maximally close approximation in the form of something that exists currently?