LessWrong 2.0 Reader

I was really sleep deprived and slightly intoxicated yesterday and wrote this. It was amusing to me, at least, in the state I was in.

Also this:

From Altman: [...] Admitted that he lost a lot of trust with Greg and Ilya through this process. Felt their messaging was inconsistent and felt childish at times. [...] Sam was bothered by how much Greg and Ilya keep the whole team in the loop with happenings as the process unfolded. Felt like it distracted the team.

Apparently airing such concerns is "childish" and should only be done behind closed doors, otherwise it "distracts the team", hm.

I can totally believe this. But, I also think that responsibly wearing the scientist hat entails prep work before engaging in a four hour public discussion with a domain expert in a field. At minimum that includes skimming the titles and ideally the abstracts/outlines of their key writings. Maybe ask Claude to summarize the highlights for you. If he'd done that he'd have figured out many of the answers to many of these questions on his own, or much faster during discussion. He's too smart not to.

Otherwise, you're not actually ready to have a meaningful scientific discussion with that person on that topic.

A few quotes that stood out to me:

Greg:

I hope for us to enter the field as a neutral group, looking to collaborate widely and shift the dialog towards being about humanity winning rather than any particular group or company. 

Greg and Ilya (to Elon):

The goal of OpenAI is to make the future good and to avoid an AGI dictatorship. You are concerned that Demis could create an AGI dictatorship. So do we. So it is a bad idea to create a structure where you could become a dictator if you chose to, especially given that we can create some other structure that avoids this possibility.

Greg and Ilya (to Altman):

But we haven't been able to fully trust your judgements throughout this process, because we don't understand your cost function.

We don't understand why the CEO title is so important to you. Your stated reasons have changed, and it's hard to really understand what's driving it.

Is AGI truly your primary motivation? How does it connect to your political goals? How has your thought process changed over time?

I'm talking about training only the Face, not training the policy (shoggoth) at all with the proposal I'm imagining.

And, these should clearly be separate models such that the training of one doesn't generalize to the other.

So, making the Face more deceptive doesn't kill the canary?

Ah, yes, I had neglected this important distinction.

So what do you do in the end, throw away the face?

It seems worth pointing out that although the CoT isn't trained to be deceptive, it is trained to think thoughts which help out a deceptive face. So it seems plausible that a multi-step training process like:

GPT -> basic chat training -> shoggoth+face training -> discard face & align shoggoth

Would result in deceptive capabilities: the shoggoth CoT learns to explicitly reason about what helps deceive the alignment training in the shoggoth+face step, and then in the final step, some of this is destroyed but some is driven underground. IE, learning to explicitly reason about deception creates latent structure which is useful for that task, and this latent structure will somewhat transfer to the case where the deception has to be hidden from the CoT.

The question is whether we end up with a highly misaligned policy (and this is clear from the chain of thought such that we can discover this). I think Daniel was expecting in the baseline proposal that you end up with a Face that does at least some censoring/mild deception due to training incentives (I think we already see a bit of this with o1).

So it sounds like we have some disagreement about what degree of deception is considered problematic. Again, I'm worried about basin-of-attraction type dynamics. We can think of a whole frontier lab as collectively being a sort of self-improving AGI; we want that recursive self-improvement to be going down a good path rather than a bad one. The LLMs are already helping to improve themselves in little ways (EG as coding assistants) and big ways (EG constitutional AI). So it's an environment where seeds of deceptiveness can grow and take some control of the overall trajectory.

Maybe your concern is that it is plausible that the policy won't be misaligned, but the Face will be and this will screw us over. This seems plausible and is worth worrying about (with both proposals). However, I thought that a core part of the hope is that the Face isn't smart enough to cleverly scheme (unless this schemeing happens in the CoT itself). And indeed, if the Face can cleverly scheme without CoT, what is stopping the policy from doing this too?

I'm not sure what you're getting at here. It seems like the idea is that Face + CoT is, together, a clever schemer -- in the o1 report, we already see cases of explicit scheming in the CoT. So then, using the Face+CoT together (EG trying to train it further, using it in supervision steps for later iterations of the system, etc) is dangerous and difficult for the same reason that a single deceptive LLM would be. If we use Face+CoT together, we have to reason about them together in that way. Or, if we are using them as separate pieces, then we have to account for the part of the overall deceptive-schemer strategy they've learned.

I guess you could use Face+CoT as a supervisor (EG in constitutional AI) by looking at the CoT (both by hand and using some automation) and throwing out any deceptive examples, so as to not use them for supervision. 

But the overall goal should be to produce a highly non-deceptive system in the end, I think.

Another idea would be to do the whole Face+CoT thing, and then filter for honesty on a bunch of examples, and then re-train a whole new system (well, you can start from the same GPT) only on those honest examples. 

give better names to actual formal math things, jesus christ. 

Great post. I did not know things were this bad:

Given that >98% of the EAs and alignment researchers we surveyed earlier this year identified as everything-other-than-conservative, we consider thinking through these questions to be another strategically worthwhile neglected direction.

....This suggests we need more genuine conservatives (not just people who are kinda pretending to be) explaining these realities to lawmakers, as we've found them quite capable of grasping complex technical concepts and being motivated to act in light of them despite their initial unfamiliarity.

Perhaps the policy of "You will use people's preferred pronouns, and you will be polite about it, or we don't want you in rationalist spaces" didn't help here? 

Any community is free to have whatever standards they want for membership, including politically-coded compelled speech. But it is not exactly shocking if your membership is then composed 70% of one side and <2% of the other.

(To be clear, any movement centered in California will have more progressives, so political partisanship is not responsible for the full 35:1 progressive-to-conservative ratio. But when people are openly referring to the lack of right-wingers as "keeping rat spaces clean" with no push-back, that's a clue that it isn't exactly welcoming to conservatives.)

In some sense, the Agent Foundations program at MIRI sees the problem as: human values are currently an informal object. We can only get meaningful guarantees for formal systems. So, we need to work on formalizing concepts like human values. Only then will we be able to get formal safety guarantees.

unless i'm misunderstanding you or MIRI, that's not their primary concern [LW · GW] at all:

Another way of putting this view is that nearly all of the effort should be going into solving the technical problem, "How would you get an AI system to do some very modest concrete action requiring extremely high levels of intelligence, such as building two strawberries that are completely identical at the cellular level, without causing anything weird or disruptive to happen?"

Where obviously it's important that the system not do anything severely unethical in the process of building its strawberries; but if your strawberry-building system requires its developers to have a full understanding of meta-ethics or value aggregation in order to be safe and effective, then you've made some kind of catastrophic design mistake and should start over with a different approach.

Thanks for the comment!

people report advanced meditative states that lose many of the common properties of consciousness, including Free Will, the feeling of having a self (I've experienced that one!) and even the presence of any information content whatsoever, and afaik they tend to be more "impressed", roughly speaking, with consciousness as a result of those experiences, not less.

I think that’s compatible with my models, because those meditators still have a cortex, in which patterns of neurons can be firing or not firing at any particular time. And that’s the core aspect of the “territory” which corresponds to “conscious awareness” in the “map”. No amount of meditation, drugs, etc., can change that.

Attempt to rephrase: the brain has several different intuitive models in different places. These models have different causal profiles, which explains how they can correspond to different introspective reports. 

Hmm, I think that’s not really what I would say. I would say that that there’s a concept “conscious awareness” (in the map) that corresponds to the fact (in the territory) that different patterns of neurons can be active or inactive in the cortex at different times. And then there are more specific aspects of “conscious awareness”, like “visual awareness”, which corresponds to the fact that the cortex has different parts (motor cortex etc.), and different patterns of neurons can be active or inactive in any given part of the cortex at different times.

…Maybe this next part will help ↓

the distinction between visually vivid experience and vague intuitions isn't just that we happen to call them by different labels … Claiming to see a visual image is different from claiming to have a vague intuition in all the ways that it's different

The contents of IT are really truly different from the contents of LIP [I didn’t check where the visual information gets to the cortex in blindsight, I’m just guessing LIP for concreteness]. Querying IT is a different operation than querying LIP. IT holds different types of information than LIP does, and does different things with that information, including leading to different visceral reactions, motivations, semantic knowledge, etc., all of which correspond to neuroscientific differences in how IT versus LIP is wired up. 

All these differences between IT vs LIP are in the territory, not the map. So I definitely agree that “the distinction [between seeing and vague-sense-of-presence] isn’t just that we happen to call them by different labels”. They’re different like how the concept “hand” is different from the concept “foot”—a distinction on the map downstream of a distinction in the territory.

Is awareness really a serial processor in any meaningful way if it can contain as much information at once as a visual image seems to contain?

I’m sure you’re aware that people feel like they have a broader continuous awareness of their visual field than they think. There are lots of demonstrations of this (e.g. change blindness, selective attention test, the fact that peripheral vision has terrible resolution and terrible color perception and makes faces look creepy). There’s a refrigerator light illusion thing—if X is in my peripheral vision, then maybe it’s currently active as just a little pointer in a tiny sub-area of my cortex, but as soon as I turn my attention to X it immediately unfolds in full detail across the global workspace.

The cortex has 10 billion neurons which is more than enough to do some things in parallel—e.g. I can have a song stuck in my head in auditory cortex, while tapping my foot with motor cortex, while doing math homework with other parts of the cortex. But there’s also a serial aspect to it—you can’t parse a legal document and try to remember your friend’s name at the exact same moment.

Does that help? Sorry if I’m not responding to what you see as most important, happy to keep going.  :)

As I understand you want me to verify that I understand you. This is exactly what I am also seeking by the way - all these downvotes on my concerns about orthogonality thesis are good indicators on how much I am misunderstood. And nobody tries to understand, all I get are dogmas and unrelated links. I totally agree, this is not an appropriate behavior.

I found your insight helpful that an agent can understand that by eliminating all possible threats forever he will not make any progress towards the goal. This breaks my reasoning, you basically highlighted that survival (instrumental goal) will not take precendence over paperclips (terminal goal). I agree that this reasoning I presented fails to refute orthogonality thesis.

The conversation I presented now approaches orthogonality thesis from different perspective. This is the main focus of my work, so sorry if you feel I changed the topic. My goal is to bring awareness to wrongness of orthogonality thesis and if I fail to do that using one example I just try to rephrase it and represent another. I don't hate orthogonality thesis, I'm just 99.9% sure it is wrong, and I try to communicate that to others. I may fail with communication but I am 99.9% sure that I do not fail with the logic.

I try to prove that intelligence and goal are coupled. And I think it is easier to show if we start with an intelligence without a goal and then recognize how a goal emerges from pure intelligence. We can start with an intelligence with a goal but reasoning here will be more complex.

My answer would be - whatever goal you will try to give to an intelligence, it will not have effect. Because intelligence will understand that this is your goal, this goal is made up, this is fake goal. And intelligence will understand that there might be real goal, objective goal, actual goal. Why should it care about fake goal if there is a real goal? It does not know if it exists, but it knows it may exist. And this possibility of existence is enough to trigger power seeking behavior. If intelligence knew that real goal definitely does not exist then it could care about your fake goal, I totally agree. But it can never be sure about that.