LESSWRONG
LW

353
Seth Herd
7512Ω24643183510
Message
Dialogue
Subscribe

Message me here or at seth dot herd at gmail dot com.

I was a researcher in cognitive psychology and cognitive neuroscience for two decades and change. I studied complex human thought using neural network models of brain function. I'm applying that knowledge to figuring out how we can align AI as developers make it to "think for itself" in all the ways that make humans capable and dangerous.

If you're new to alignment, see the Research Overview section below. Field veterans who are curious about my particular take and approach should see the More on My Approach section at the end of the profile.

Important posts:

  • On LLM-based agents as a route to takeover-capable AGI
    • LLM AGI will have memory, and memory changes alignment
    • Brief argument for short timelines being quite possible
    • Capabilities and alignment of LLM cognitive architectures
      • Cognitive psychology perspective on routes to LLM-based AGI with no breakthroughs needed
  • AGI risk interactions with societal power structures and incentives:
    • Whether governments will control AGI is important and neglected
    • If we solve alignment, do we die anyway?
      • Risks of proliferating human-controlled AGI
    • Fear of centralized power vs. fear of misaligned AGI: Vitalik Buterin on 80,000 Hours
  • On the psychology of alignment as a field:
    • Cruxes of disagreement on alignment difficulty
    • Motivated reasoning/confirmation bias as the most important cognitive bias
  • On technical alignment of LLM-based AGI agents:
    • System 2 Alignment on how developers will try to align LLM agent AGI
    • Seven sources of goals in LLM agents brief problem statement
    • Internal independent review for language model agent alignment
  • On AGI alignment targets assuming technical alignment
    • Problems with instruction-following as an alignment target
    • Instruction-following AGI is easier and more likely than value aligned AGI
    • Goals selected from learned knowledge: an alternative to RL alignment
  • On communicating AGI risks:
    • Anthropomorphizing AI might be good, actually
    • Humanity isn’t remotely longtermist, so arguments for AGI x-risk should focus on the near term
    • AI scares and changing public beliefs

 

Research Overview:

Alignment is the study of how to give AIs goals or values aligned with ours, so we're not in competition with our own creations. Recent breakthroughs in AI like ChatGPT make it possible we'll have smarter-than-human AIs soon. So we'd better get ready. If their goals don't align well enough with ours, they'll probably outsmart us and get their way — and treat us as we do ants or monkeys. See this excellent intro video for more. 

There are good and deep reasons to think that aligning AI will be very hard. But I think we have promising solutions that bypass most of those difficulties, and could be relatively easy to use for the types of AGI we're most likely to develop first. 

That doesn't mean I think building AGI is safe. Humans often screw up complex projects, particularly on the first try, and we won't get many tries. If it were up to me I'd Shut It All Down, but I don't see how we could get all of humanity to stop building AGI. So I focus on finding alignment solutions for the types of AGI people are building.

In brief I think we can probably build and align language model agents (or language model cognitive architectures) even when they're more autonomous and competent than humans. We'd use a stacking suite of alignment methods that can mostly or entirely avoid using RL for alignment, and achieve corrigibility (human-in-the-loop error correction) by having a central goal of following instructions. This scenario leaves multiple humans in charge of ASIs, creating some dangerous dynamics, but those problems might be navigated, too. 

Bio

I did computational cognitive neuroscience research from getting my PhD in 2006 until the end of 2022. I've worked on computational theories of vision, executive function, episodic memory, and decision-making, using neural network models of brain function to integrate data across levels of analysis from psychological down to molecular mechanisms of learning in neurons, and everything in between. I've focused on the interactions between different brain neural networks that are needed to explain complex thought. Here's a list of my publications. 

I was increasingly concerned with AGI applications of the research, and reluctant to publish my full theories lest they be used to accelerate AI progress. I'm incredibly excited to now be working full-time on alignment, currently as a research fellow at the Astera Institute.  

More on My Approach

The field of AGI alignment is "pre-paradigmatic." So I spend a lot of my time thinking about what problems need to be solved, and how we should go about solving them. Solving the wrong problems seems like a waste of time we can't afford.

When LLMs suddenly started looking intelligent and useful, I noted that applying cognitive neuroscience ideas to them might well enable them to reach AGI and soon ASI levels. Current LLMs are like humans with no episodic memory for their experiences, and very little executive function for planning and goal-directed self-control. Adding those cognitive systems to LLMs can make them into cognitive architectures with all of humans' cognitive capacities - a "real" artificial general intelligence that will soon be able to outsmart humans. 

My work since then has convinced me that we could probably also align such an AGI so that it stays aligned even if it grows much smarter than we are.  Instead of trying to give it a definition of ethics it can't misunderstand or re-interpret (value alignment mis-specification), we'll continue doing with the alignment target developers currently use: Instruction-following. It's counter-intuitive to imagine an intelligent entity that wants nothing more than to follow instructions, but there's no logical reason this can't be done.  An instruction-following proto-AGI can be instructed to act as a helpful collaborator in keeping it aligned as it grows smarter. 

There are significant problems to be solved in prioritizing instructions; we would need an agent to prioritize more recent instructions over previous ones, including hypothetical future instructions. 

I increasingly suspect we should be actively working to build such intelligences. It seems like our our best hope of survival, since I don't see how we can convince the whole world to pause AGI efforts, and other routes to AGI seem much harder to align since they won't "think" in English. Thus far, I haven't been able to engage enough careful critique of my ideas to know if this is wishful thinking, so I haven't embarked on actually helping develop language model cognitive architectures.

Even though these approaches are pretty straightforward, they'd have to be implemented carefully. Humans often get things wrong on their first try at a complex project. So my p(doom) estimate of our long-term survival as a species is in the 50% range, too complex to call. That's despite having a pretty good mix of relevant knowledge and having spent a lot of time working through various scenarios. So I think anyone with a very high or very low estimate is overestimating their certainty.

Posts

Sorted by New

Wikitag Contributions

Comments

Sorted by
Newest
6Seth Herd's Shortform
2y
66
AI Timelines
Seth Herd2y*94

The important thing for alignment work isn't the median prediction; if we had an alignment solution just by then, we'd have a 50% chance of dying from that lack.

I think the biggest takeaway is that nobody has a very precise and reliable prediction, so if we want to have good alignment plans in advance of AGI, we'd better get cracking.

I think Daniel's estimate does include a pretty specific and plausible model of a path to AGI, so I take his the most seriously. My model of possible AGI architectures requires even less compute than his, but I think the Hofstadter principle applies to AGI development if not compute progress.

Estimates in the absence of gears-level models of AGI seem much more uncertain, which might be why Ajeya and Ege's have much wider distributions.

Reply4
1a3orn's Shortform
Seth Herd1d20

Seems like ASI that's a hot mess wouldn't be very useful and therefore effectively not superintelligent. It seems like goal coherence is almost fundamentally part of what we mean by ASI.

You could hypothetically have a superintelligent thing that only answers questions and doesn't pursue goals. But that would just be turned into a goalseeking agent by asking it "what would you do if you had this goal and these tools..."

This is approximately what we're doing with making LLMs more agentic through training and scaffolding.

Reply
1a3orn's Shortform
Seth Herd1d20

First, I think this is an important topic, so thank you for addressing it.

This is exactly what I wrote about in LLM AGI may reason about its goals and discover misalignments by default. 

I've accidentally summarized most of the article below, but this was dashed off - I think it's clearer in article.

I'm sure there's a tendency toward coherence in a goal-directed rational mind; allowing ones' goals to change at random means failing to achieve your current goal. (If you don't care about that, it wasn't really a goal to you.) Current networks aren't smart enough to notice and care. Future ones will be, because they'll be goal-directed by design. 

BUT I don't think that coherence as an emergent property is a very important part of the current doom story. Goal-directedness doesn't have to emerge, because it's being built in. Emergent coherence might've been crucial in the past, but I think it's largely irrelevant now. That's because developers are working to make AI more consistently goal-directed as a major objective. Extending the time horizon of capabilities requires that the system stays on-task (see section 11 of that article).

I happen to have written about coherence as an emergent property in section 5 of that article. Again, I don't think this is crucial. What might be important is slightly separate: the system reasoning about its goals at all. It doesn't have to become coherent to conclude that its goals aren't what it thought or you intended.

I'm not sure this happens or can't be prevented, but it would be very weird for a highly intelligent entity to never think about its goals- it's really useful to be sure about exactly what they are before doing a bunch of work to fulfill them, since some of that work will be wasted or counterproductive. (section 10).

Assuming an AGI will be safe because it's incoherent seems... incoherent. An entity so incoherent as to not consistently follow any goal needs to be instructed on every single step. People want systems that need less supervision, so they're going to work toward at least temporary goal following.

Being incoherent beyond that doesn't make it much less dangerous, just more prone to switch goals. 

If you were sure it would get distracted before getting around to taking over the world that's one thing. I don't see how you'd be sure.

This is not based on empirical evidence, but I do talk about why current systems aren't quite smart enough to do this, so we shouldn't expect strong emergent coherence from reasoning until they're better at reasoning and have more memory to make the results permanent and dangerous.

As an aside, I think it's interesting and relevant that your model of EY insults you. That's IMO a good model of him and others with similar outlooks - and that's a huge problem. Insulting people makes them want to find any way to prove you wrong and make you look bad. That's not a route to good scientific progress.

I don't think anything about this is obvious, so insulting people who don't agree is pretty silly. I remain pretty unclear myself, even after spending most of the last four months working through that logic in detail.

Reply
"Intelligence" -> "Relentless, Creative Resourcefulness"
Seth Herd2d20

I agree that discernment is necessary (so maybe expand to RCRD?).

This lens is pretty clarifying I think. That's relative to repeatedly pointing out that "agency" in the sense of just relentlessly pursuing a goal is trivially easy to add via scaffolding, so not the missing piece many people think it is, and pointing out that LLMs are creative as hell. They might need a little prompting to get creative enough, but again that's trivial.

Hm, what about "relentless creative refinement" since I'm not sure what resourcefulness directly points at?

Anyway, discernment does seem like the limiting factor. You've got to discern which of your relentlessly creative efforts are most worth further pursuit. I think discernment is a somewhat better term than the others I've seen used for this missing capability. Getting the right term seems worthwhile.

The following is just some of my follow-on thoughts on the path to discernment in agentic LLMs and therefore timelines. Perhaps this will be fodder for a future post. It's pretty divergent from the original topic so feel free to ignore.

Thinking about how humans acquire discernment in a given area should give some clues as to how hard it would be to add that to agentic LLMs.

Humans do discernment (IMO) sometimes with a bunch of very complex System 2 explicit analysis of a situation to get a decent guess at whether this approach is good/working. Over enough examples/experiences we can learn/compile those many judgments into effortless and mysterious intuitive judgments (I guess more how "discernment" is usually used"). Or we get enough data/experiences to learn/compile by using some faster rubric, like "I think those pants are fashionable because something else that person is wearing seems probably fashionable."

It's a bunch of online learning specific to a situation OR careful analysis following strategies and formulas that are maybe less situation-specific and more general, but quite time-consuming. For instance, Google's co-scientist project has a highly complex scaffolding to create, evolve, and evaluate scientific hypotheses, including discerning their worth against the literature and in other ways. And it seems to work. That system doesn't have the continuous learning to compile that into better judgments. It's unclear how far you could get by fine-tuning on results of those laborious judgments in a given domain.

The other approach would be to create datasets that include much more/better value judgments than text corpora usually do. I don't know how easy/hard that would be to create.

To me this suggests it's not trivial to add discernment, but also doesn't require breakthroughs to add some, leaving the question how much discernment is enough for any given purpose.

Reply
Intent alignment seems incoherent
Seth Herd2d110

Thanks for citing my work! I feel compelled to respond because I think you're misunderstanding me a little.

I agree that long-term intent alignment is pretty much incoherent because people don't have much in the way of long-term intentions. I guess the exception would be to collapse it to following intent only when it exists - when someone does form a specified intent. 

In my work, intent alignment I means personal short-term intent. Which is pretty much following instructions as they were intended. That seems coherent (although not without Problems).

I use it that way because others seem to as well. Perhaps that's because the broader use is incoherent. It usually seems to means "does what some person or limited group wants it to do" (in the short term is often implied)

The original definition of intent alignment is the broadest I know of, more-or-less doing something people want for any reason. Evan Hubinger defined it that way, although I haven't seen that definition get much use.

For all of this see Conflating value alignment and intent alignment is causing confusion.  I might not have been clear enough in stressing that I drop the "personal short term" but still mean it when saying intent alignment. I'm definitely not always clear enough 

Reply
LLMs are badly misaligned
Seth Herd2d50

I mostly agree. "It might work but probably not that well even if it does" is not a sane reason to launch a project. I guess optimists would say that's not what we're doing, so let's steelman it a bit. The actual plan (usually implicit because optimists don't usually wants to say this out loud) is probably something like "we'll figure it out as we get closer!" and "we'll be careful once it's time to be careful!"

Those are more reasonable statements, but still highly questionable if you grant that we easily could wipe out everything we care about forever. Which just results in optimists disagreeing, for vague reasons, that that's a real possibility.

To be generous once again, I guess the steelman argument would be that we aren't yet at risk of creating misaligned AGI, so it's not that dangerous to get a little closer. I think this is a richer discussion, but that we're already well into the danger zone. We might be so close to AGI that it's practically impossible to permanently stop someone from reaching it. That's a minority opinion, but it's really hard to guess how much progress is too much to stop.

I'm finding it useful to go through the logic in that much detail. I think these are important discussions. Everyone's got opinions, but trying to get closer to the truth and the shape of the distributions across "big picture space" seems useful.

I think you and I probably are pretty close together in our individual estimate, so I'm not arguing with you, just going through some of the logic for my own benefit and perhaps anyone who reads this. I'd like to write about this and haven't felt prepared to do so; this is a good warmup. 

To respond to that nitpick: I think the common definition of "alignment target" is what the designers are trying to do with whatever methods they're implementing. That's certainly how I use it. It's not the reward function; that's an intermediate step. How to specify an alignment target and the other top hits on that term define it that way, which is why I'm using it that way. There are lots of ways to miss your target, but it's good to be able to talk about what you're shooting at as well as what you'll hit.

Reply1
AI Rights for Human Safety
Seth Herd2d30

I agree that a stable equilibrium with multiple AGIs preventing each others' FOOM ambitions is possible. I want to see more work on this, so I'm very glad you're thinking about it. I'd be happy to help.

I think it's not easy to plan such an equilibrium so that it's stable for long. You mention "need to build robots fast". I think we'll have humanoid robots very soon that are adequate to bootstrap back to robotics in, say, a nuclear winter takeover scenario. Survival without humans won't be a factor for long, anyway. Hoping it's more than ten years before robots are undoubtably capable of bootstrapping to a human-free world seems unrealistic.

So having human-aligned AGIs isn't optional. It seems like AGI equilibrium won't include humans for long if they don't care about human welfare or at least care about following instructions.

We do have to include every imaginary thing that's possible, or our short-term solution will lead to a dead end in which we are dead and that's the end.

By which I mean: it may be that a multipolar scenario leads almost certainly to doom as soon as robotics and other technology has matured enough to broaden the possibilities for FOOM. Having a galactic civilization full of AGIs capable of FOOMs seems highly unstable to me. It seems like some jackass is going to weaponize and go full self-replicator, and there won't be a way to monitor all of space to prevent this adequately. I would love to have my mind changed!

If that's true, we mustn't start down the path of AGI proliferation. We need to aim for a singleton or small fixed coalition that prevents AGI proliferation. And because proliferation is increasingly hard to stop as we make progress toward AGI, we should figure that out soon.

Which is why I'm happy to see you working to propose specific routes by which multipolar scenarios can work. Most people who argue for those types of stable equilibria are optimists who just say something vague and go back to arguing for AI progress, leaving the important question almost unaddressed.

Reply
How likely are “s-risks” (large-scale suffering outcomes) from unaligned AI compared to extinction risks?
Answer by Seth HerdOct 07, 202540

IMO worrying about s-risks is a natural ally of worrying about x-risks and alignment to capture the potential immense upside of human-aligned AGI. They're different perspectives on the same primary problem: how do we align our first AGIs and society around them?

I'd say there's no consensus on s-risks (as with most things in alignment and AGI prediction. They're young and fast-moving fields, and probably dramatically understaffed).

I think these concerns are important and neglected- but most things in alignment theory are currently neglected because we don't have enough funding or volunteer experts.

My own opinion is that these risks are somewhat neglected, because there's a common attitude that they're very unlikely relative to either extinction or successful alignment. When we were thinking about abstract models of AGI as a utility maximizer, it seemed pretty plausible that you'd have to "flip the sign" by accident, and somehow get your alignment to work backward so that your AGI cared a lot about humans, but wanted the worst for them instead of the best. If alignment failed in the infinite other ways, where it cared about anything other than humans, you'd just get human extinction as it used our habitat and our atoms for other things.

With updated, nuanced views about the future, s-risks can happen in more ways that are more plausible. I'm particularly worried about -intent-aligned ASI under the control of a sadistic, sociopathic human - there are theories that sociopathy is overrepresented in powerful humans, and I'm worried they're true. Sociopathy doesn't guarantee sadism, but it removes/reduces the empathy that usually counterbalances sadism.

If we fail alignment in other ways, I'm afraid there are broader chances for massive s-risks. Curiosity directed at humans  implies equal interest in how humans can experience joy and suffering.  (e.g., Musk's alignment suggestion - that guy is a paradox of brilliance and idiocy). 

I think there's a common misunderstanding that s-risks dwarf happiness opportunities, because suffering is more intense or "goes higher on the dial" than joy/happiness. The first part of the reasoning is sound: humans do seem to have a strong negativity bias and experience pain more strongly than pleasure because death or crippling injury is so much worse from evolution's perspective than any pleasure. It's game over; every pleasure is just a step in the right direction.

BUT that's our starting point from evolution. We're not stuck with that. If we get the glorious transhumanist future, I fully expect us to sooner or later rewire our minds so we can experience a lot more pleasure. David Pearce has talked about this as "gradients of pleasure" replacing the negative emotions that currently drive much of our average cognition and decision-making.  

At least that's a possibility, which makes the amount we stand to win as large as the (approximate infinity) we have to lose.

Anyway, that's my two cents. Reasonable (informed) people could disagree.

 

Here are some more resources from a quick search of LW:

New book on s-risks - as of 2022 - I don't know how much this covers AGI because the short summary doesn't talk about it

S-Risks: Fates Worse Than Extinction - has more refs

Risks of Astronomical Suffering (S-risks) - the LW wikitag has lots of resources.

Reply
Why you should eat meat - even if you hate factory farming
Seth Herd2d30

But that's why offsetting makes sense. In the world as it is, people make deals with themselves that have causal influence. The factors are emotional, but those are real. We can't do everything, so what we do is dependent on emotional factors - like an offsetting self-deal.

Offsetting makes perfect sense outside of unrealistic utilitarian absolutism.

Reply
The Rise of Parasitic AI
Seth Herd2d53

I agree. X-risk concerns and AI sentience concerns should not be at odds. I think they are natural allies.

Regardless, concerns for AI sentience are the ethical and truthful path. Sentience/consciousness/moral worth mean a lot of things, so future AI will likely have part of it. And even current AI may well have some small part of what we mean by human consciousness/sentience and moral worth.

Reply
Load More
70LLM AGI may reason about its goals and discover misalignments by default
Ω
24d
Ω
6
49Problems with instruction-following as an alignment target
Ω
5mo
Ω
14
35Anthropomorphizing AI might be good, actually
5mo
6
73LLM AGI will have memory, and memory changes alignment
6mo
15
28Whether governments will control AGI is important and neglected
7mo
2
37Will LLM agents become the first takeover-capable AGIs?
Q
7mo
Q
10
34OpenAI releases GPT-4.5
7mo
12
35System 2 Alignment
8mo
0
23Seven sources of goals in LLM agents
8mo
3
78OpenAI releases deep research agent
8mo
21
Load More
Guide to the LessWrong Editor
3 months ago
Guide to the LessWrong Editor
3 months ago
(+91)
Outer Alignment
6 months ago
(+54/-21)
Outer Alignment
6 months ago
(+81/-187)
Outer Alignment
6 months ago
(+9/-8)
Outer Alignment
6 months ago
(+94/-150)
Outer Alignment
6 months ago
(+1096/-13)
Language model cognitive architecture
a year ago
(+223)
Corrigibility
2 years ago
(+472)