All of Sodium's Comments + Replies

Hate to be that person, but is that April 18th deadline AoE/PDT/a secret third thing?

I don't think you're supposed to get the virtue of Void, if you got it, it wouldn't void anymore, would it?

If people outside of labs are interested in doing this, I think it'll be cool to look for cases of scheming in evals like The Agent Company, where they have an agent act as a remote worker for a company. They ask the agent to complete a wide range of tasks (e.g., helping with recruiting, messaging coworkers, writing code). 

You could imagine building on top of their eval and adding morally ambiguous tasks, or just look through the existing transcripts to see if there's anything interesting there (the paper mentions that models would sometimes "deceive"... (read more)

One thing not mentioned here (and I think should be talked about more) is that the naturally occurring genetic distribution is very unequal in a moral sense. A more egalitarian society would put a stop to Eugenics Performed by a Blind, Idiot God. 

Have your doctor ever asked about if you have a family history of [illness]? For so many diseases, if your parents have it, you're more likely to have it, and your kids are more likely to have it. These illnesses plague families for generations. 

I have a higher than average chance of getting hypertension... (read more)

I'm not sure if the rationalists did anything they shouldn't do re: Ziz. Going forward though, I think epistemic learned helplessness/memetic immune systems should be among the first things to introduce to newcomers to the site/community. Being wary that some ideas are, in a sense, out to get you, is a central part of how I process information. 

Not exactly sure how to implement that recommendation though. You also don't want people to use it as a fully general counterargument to anything they don't like. 

Ranting a bit here, but it just feels like... (read more)

You link a comment by clicking the timestamp next to the username (which, now that I say it, does seem quite unintuitive... Maybe it should also be possible via the three dots on the right side).

While this post didn't yield a comprehensive theory of how fact finding works in neural networks, it's filled with small experimental results that I find useful for building out my own intuitions around neural network computation.

I think that's speaks to how well these experiments are scoped out that even a set of not-globally-coherent findings yield useful information. 

So I think the first claim here is wrong. 

Let’s start with one of those insights that are as obvious as they are easy to forget: if you want to master something, you should study the highest achievements of your field. If you want to learn writing, read great writers, etc.

If you want to master something, you should do things that causally/counter factually increase your ability (in the order of most to least cost-effective). You should adopt interventions that actually make you better compared to the case that you haven't done them. 

Any intervent... (read more)

Perhaps I am missing something, but I do not understand the value of this post. Obviously you can beat something much smarter than you if you have more affordances than it does.

FWIW, I have read some of the discourse on the AI Boxing game. In contrast, I think those posts are valuable. They illustrate that even with very little affordances a much more intelligent entity can win against you, which is not super intuitive especially in the boxed context. 

So the obvious question is, how does differences in affordances lead to differences in winning (i.e.,... (read more)

Yes

I think the alignment stress testing team should probably think about AI welfare more than they currently do, both because (1) it could be morally relevant and (2) it could be alignment-relevant. Not sure if anything concrete that would come out of that process, but I'm getting the vibe that this is not thought about enough. 

since it's near-impossible to identify which specific heuristic the model is using (there can always be a slightly more complex, more general heuristic of which your chosen heuristic is a special case).  

I'm putting some of my faith in low-rank decompositions of bilinear MLPs but I'll let you know if I make any real progress with it :)

This sounds like a plausible story for how (successful) prosaic interpretability can help us in the short to medium term! I would say though, I think more applied mech interp work could supplement prosaic interpretability's theories. For example, the reversal curse seems mostly explained by what little we know about how neural networks do factual recall. Theory on computation in superposition help explain why linear probes can recover arbitrary XORs of features.

Reading through your post gave me a chance to reflect on why I am currently interested in mech i... (read more)

Pr(Ai)2R is at least partially funded by Good Ventures/OpenPhil

I think the actual answer is: the AI isn't smart enough and trips up a lot.

But I haven't seen a detailed write up anywhere that talks about why the AI trips up and what are the types of places where it trips up. It feels like all of the existing evals work optimize for legibility/reproducibility/being clearly defined. As a result, it's not measuring the one thing that I'm really interested in: why don't we have AI agents replacing workers. I suspect that some startup's internal doc on "why does our agent not work yet" would be super interesting to read and track over time. 

I read this post in full back in February. It's very comprehensive. Thanks again to Zvi for compiling all of these. 

To this day, it's infuriating that we don't have any explanation whatsoever from Microsoft/OpenAI on what went wrong with Bing Chat. Bing clearly did a bunch of actions its creators did not want. Why? Bing Chat would be a great model organism of misalignment. I'd be especially eager to run interpretability experiments on it. 

The whole Bing chat fiasco is also gave me the impetus to look deeper into AI safety (although I think absent Bing, I would've came around to it eventually).

When this paper came out, I don't think the results were very surprising to people who were paying attention to AI progress. However, it's important to the "obvious" research and demos to share with the wider world, and I think Apollo did a good job with their paper.

TL; DR: This post gives a good summary of how models can get smarter over time, but while they are superhuman at some tasks, they can still suck at others (see the chart with Naive Scenario v. Actual performance). This is a central dynamic in the development of machine intelligence and deserves more attention. Would love to hear other's thoughts on this—I just realized that it needed one more positive vote to end up in the official review.

In other words, current machine intelligence and human intelligence are compliments, and human + AI will be more produc... (read more)

OpenAI released another set of emails here. I haven't looked through them in detail but it seems that they contain some that are not already in this post.

Any event next week?

Yeah my view is that as long as our features/intermediate variables form human understandable circuits, it doesn't matter how "atomic" they are. 

Almost certainly not original idea: Given the increasing fine-tuning access to models (see also the recent reinforcement fine tuning thing from OpenAI), see if fine tuning on goal directed agent tasks for a while leads to the types of scheming seen in the paper. You could maybe just fine tune on the model's own actions when successfully solving SWE-Bench problems or something. 

(I think some of the Redwood folks might have already done something similar but haven't published it yet?)

What is the probability that the human race will NOT make it to 2100 without any catastrophe that wipes out more than 90% of humanity?

Could we have this question be phrased using no negations instead of two? Something like "What is the probability that there will be a global catastrophe that wipes out 90% or more of humanity before 2100."

Thanks for writing these posts Zvi <3 I've found them to be quite helpful.

Hi Clovis! Something that comes to mind is Zvi's dating roundup posts in case you haven't seen them yet. 

I think people see it and think "oh boy I get to be the fat people in Wall-E"

(My friend on what happens if the general public feels the AGI)

This chapter on AI follows immediately after the year in review, I went and checked the previous few years' annual reports to see what the comparable chapters were about, they are

2023: China's Efforts To Subvert Norms and Exploit Open Societies

2022: CCP Decision-Making and Xi Jinping's Centralization Of Authority

2021: U.S.-China Global Competition (Section 1: The Chinese Communist Party's Ambitions and Challenges at its Centennial

2020: U.S.-China Global Competition (Section 1: A Global Contest For Power and Influence: China's View of Strategic Competiti... (read more)

I think^[1] people^[2] probably trust individual tweets way more than they should. 

Like, just because someone sounds very official and serious, and it's a piece of information that's inline with your worldviews, doesn't mean it's actually true. Or maybe it is true, but missing important context. Or it's saying A causes B when it's more like A and C and D all cause B together, and actually most of the effect is from C but now you're laser focused on A. 
 

Also you should be wary that the tweets you're seeing are optimized for piquing th... (read more)

Sorry, is there a timezone for when the applications would close by, or is it AoE?

Man, politics really is the mind killer

I think knowing the karma and agreement is useful, especially to help me decide how much attention to pay to a piece of content, and I don't think there's that much distortion from knowing what others think. (i.e., overall benefits>costs)

Is there no event on Oct 29th?

Wait a minute, "agentic" isn't a real word? It's not on dictionary.com or Merriam-Webster or Oxford English Dictionary.

I agree that if you put more limitations on what heuristics are and how they compose, you end up with a stronger hypothesis. I think it's probably better to leave that out and try do some more empirical work before making a claim there though (I suppose you could say that the hypothesis isn't actually making a lot of concrete predictions yet at this stage). 

I don't think (2) necessarily follows, but I do sympathize with your point that the post is perhaps a more specific version of the hypothesis that "we can understand neural network computation by doing mech interp."

Thanks for reading my post! Here's how I think this hypothesis is helpful:

It's possible that we wouldn't be able to understand what's going on even if we had some perfect way to decompose a forward pass into interpretable constituent heuristics. I'm skeptical that this would be the case, mostly because I think (1) we can get a lot of juice out of auto-interp methods and (2) we probably wouldn't need to simultaneously understand that many heuristics at the same time (which is the case for your logic gate example for modern computers). At the minimum, I woul... (read more)

I think there's something wrong with the link :/ It was working fine earlier but seems to be down now

I think those sound right to me. It still feels like prompts with weird suffixes obtained through greedy coordinate search (or other jailbreaking methods like h3rm4l) are good examples for "model does thing for anomalous reasons."

Sorry, I linked to the wrong paper! Oops, just fixed it. I meant to link to Aaron Mueller's Missed Causes and Ambiguous Effects: Counterfactuals Pose Challenges for Interpreting Neural Networks.

You could also use \text{}

since people often treat heuristics as meaning that it doesn't generalize at all.

Yeah and I think that's a big issue! I feel like what's happening is that once you chain a huge number of heuristics together you can get behaviors that look a lot like complex reasoning. 

I see, I think that second tweet thread actually made a lot more sense, thanks for sharing!
McCoy's definitions of heuristics and reasoning is sensible, although I personally would still avoid "reasoning" as a word since people probably have very different interpretations of what it means. I like the ideas of "memorizing solutions" and "generalizing solutions."

I think where McCoy and I depart is that he's modeling the entire network computation as a heuristic, while I'm modeling the network as compositions of bags of heuristics, which in aggregate would dis... (read more)

Yeah that's true. I meant this more as "Hinton is proof that AI safety is a real field and very serious people are concerned about AI x-risk."

Thanks for the pointer! I skimmed the paper. Unless I'm making a major mistake in interpreting the results, the evidence they provide for "this model reasons" is essentially "the models are better at decoding words encrypted with rot-5 than they are at rot-10." I don't think this empirical fact provides much evidence one way or another.

To summarize, the authors decompose a model's ability to decode shift ciphers (e.g., Rot-13 text: "fgnl" Original text: "stay")  into three categories, probability, memorization, and noisy reasoning.

Probability just ref... (read more)

I think it's mostly because he's well known and have (especially after the Nobel prize) credentials recognized by the public and elites. Hinton legitimizes the AI safety movement, maybe more than anyone else. 

If you watch his Q&A at METR, he says something along the lines of "I want to retire and don't plan on doing AI safety research. I do outreach and media appearances because I think it's the best way I can help (and because I like seeing myself on TV)." 

And he's continuing to do that. The only real topic he discussed in first phone interv... (read more)

I like this research direction! Here's a potential benchmark for MAD.

In Coercing LLMs to do and reveal (almost) anything, the authors demonstrate that you can force LLMs to output any arbitrary string—such as a random string of numbers—by finding a prompt through greedy coordinate search (the same method used in the universal and transferable adversarial attack paper). I think it’s reasonable to assume that these coerced outputs results from an anomalous computational process.

Inspired by this, we can consider two different inputs, the regular one looks som... (read more)

I'd imagine that RSP proponents think that if we execute them properly, we will simply not build dangerous models beyond our control, period. If progress was faster than what labs can handle after pausing, RSPs should imply that you'd just pause again. On the other hand, there's not a clear criteria for when we would pause again after, say, a six month pause in scaling.

Now whether this would happen in practice is perhaps a different question.

I really liked the domesticating evolution section, cool paper!

That was the SHA-256 hash for:

What if a bag of heuristics is all there is and a bag of heuristics is all we need? That is, (1) we can decompose each forward pass in current models into a set of heuristics chained together and (2) heauristics chained together is all we need for agi

Here's my full post on the subject

Also from WSJ