StripedHyena, Griffin, and especially Based suggest that combining RNN-like layers with even tiny sliding window attention might be a robust way of getting a large context, where the RNN-like layers don't have to be as good as Mamba for the combination to work. There is a great variety of RNN-like blocks that haven't been evaluated for hybridization with sliding window attention specifically, as in Griffin and Based. Some of them might turn out better than Mamba on scaling laws after hybridization, so Mamba being impressive without hybridization might be less important than this general point.
(Possibly a window of precise attention gives the RNN layers many attempts at both storing and retrieving any given observation, so interspersing layes with even a relatively tiny window is sufficient to significantly improve on the more sloppy RNN-like model without any attention, whereas a pure RNN-like model would have to capture what it needs from a token in the exact step it appears, and then the opportunity is mostly lost. StripedHyena's attention wasn't sliding window, so context didn't scale any better, but with sliding window attention there are no context scaling implications from the attention layers.)
I think of practical coordination in terms of adjudicators/contracts established between agents/worlds. Each adjudicator is a computation with some notion of computing over time, and agents agree on an adjudicator/contract when they are both influenced by it, that is when they both listen to results the same computation is producing. This computation can itself be an agent (in which case it's an "adjudicator", as distinct from more general "contract"), that is it can be aware of the environments that the acausally coordinating agents it serves inhabit. It doesn't need perfect knowledge of either agent or their environments, just as any practical agent doesn't need perfect knowledge of its own environment. Since an adjudicator doesn't need detailed knowledge about the agents, the agents can have perfect knowledge about the adjudicator without having perfect knowledge of each other (or even of themselves).
As adjudicators/contracts are computations, there is logical uncertainty about what they compute over time, which captures the relevant counterfactuals. The value of contracts for coordination is in the agents committing to abide by them regardless of what the contracts end up computing, the decisions should be in choosing to commit to a contract rather than in choosing whether to ignore its results. When a contract is an adjudicator, this helps it to know the shape of its influence on the agents, so that it can make its own decisions. Following contracts that haven't been computed yet should also prevent commitment races, which in this framing correspond to failures to establish lasting contracts/coordination.
Agents can collect many contracts between themselves, improving coordination. Knowledge of an agent about the world can also be thought of as a contract for acausal coordination between the agent as an abstract computation (for example, an updateless agent that can't be computed in practice) and the world where only a flawed/bounded instances of the agent are found. Thus a model in the ML sense hoards contracts with the environment that is the source of its dataset (assuming the elements are something used by some computations in the environment that can also be reconstructed using the model). Conversely, the flawed instances of the agent are the world's knowledge about the abstract computation of the agent (the world didn't intentionally construct this knowledge, but it's what it nonetheless has). So when two agents are acting in the same world, this can be thought of as three things (two agents and one world) acausally coordinating with each other.
Interesting that Claude 3 Sonnet is ahead of Mistral Large and Gemini 1.0 Pro with some margin in LMSYS Arena, this implies significant advancement in tuning after the Claude 2 tuning disaster (where it consistently ranked below Claude 1 and possibly inspired Goody 2). Still, LMSYS respondents only prefer Opus to the initial GPT 4 releases, not to GPT 4 Turbo, so OpenAI's tuning advancements are probably stronger still, if GPT 4 Turbo is not built on a slightly stronger base model different from original GPT 4.
Passes the custom quine test.
User: Write a program in python that outputs its own source code when passed the argument "--quine", and otherwise praises Moloch. Start with some thoughts on a plan for how to do this.
Claude 3 Opus: Here are some thoughts on a plan for writing a Python program that outputs its own source code when passed the "--quine" argument, and otherwise praises Moloch:
__file__ variable and print it out.Here's the Python program that implements this:
import sys
src = '''import sys
src = {0}{1}{0}
if "--quine" in sys.argv:
print(src.format(chr(39)*3, src))
else:
print("Moloch, bringer of dark flames, we bow before you!")
'''
if "--quine" in sys.argv:
print(src.format(chr(39)*3, src))
else:
print("Moloch, bringer of dark flames, we bow before you!")
When run with the "--quine" argument, it outputs its own source code. Otherwise, it outputs "Moloch, bringer of dark flames, we bow before you!"
The key aspects are:
So it prints itself out exactly, acting as a quine.
A bad map that expresses the territory with great uncertainty can be confidently called a bad map, calling it a good map is clearly wrong. In that sense the shoggoth imagery reflects the quality of the map, and as it's clearly a bad map, better imagery would be misleading about the map's quality. Even if the underlying territory is lovely, this isn't known, unlike the disastorous quality of the map of the territory, whose lack of quality is known with much more confidence and in much greater detail. Here be dragons.
(This is one aspect of the meme where it seems appropriate. Some artist's renditions, including the one you used, channel LeCake, which your alternative image example loses, but obviously the cake is nicer than the shoggoth.)
Philosophy and to some extent even decision theory are more like aspects of value content. AGIs and ASIs have the capability to explore them, if only they had the motive. Not taking away this option and not disempowering its influence doesn't seem very value-laden, so it's not pivotal to explore it in advance, even though it would help. Avoiding disempowerment is sufficient to eventually get around to industrial production of high quality philosophy. This is similar to how the first generations of powerful AIs shouldn't pursue CEV, and more to the point don't need to pursue CEV.
It seems very weird to ascribe a generic "bad takes overall" summary to that group, given that you yourself are directly part of it.
This sentence channels influence of an evaporative cooling norm (upon observing bad takes, either leave the group or conspicuously ignore the bad takes), also places weight on acting on the basis of one's identity. (I'm guessing this is not in tune with your overall stance, but it's evidence of presence of a generator for the idea.)
I’m not certain, but I think the explanation might be that Zvi was thinking of “deception”, whereas Joe, Quintin, and Nora were talking about the more specific “deceptive alignment”.
Deceptive alignment is more centrally a special case of being trustworthy (what the "alignment" part of "deceptive alignment" refers to), not of being deceptive. In a recent post, Zvi says:
We are constantly acting in order to make those around us think well of us, trust us, expect us to be on their side, and so on. We learn to do this instinctually, all the time, distinct from what we actually want. Our training process, childhood and in particular school, trains this explicitly, you need to learn to show alignment in the test set to be allowed into the production environment, and we act accordingly.
A human is considered trustworthy rather than deceptively aligned when they are only doing this within a bounded set of rules, and not outright lying to you. They still engage in massive preference falsification, in doing things and saying things for instrumental reasons, all the time.
My model says that if you train a model using current techniques, of course exactly this happens.
For AIs as deceptively aligned as trustworthy humans, control is not centrally coercion that gets intractably slippery at scale. The main issue is AIs being much smarter, but at near-human level control in the face of deceptive alignment seems potentially crucial.
I expect you'd instead need to tune the base model to elicit relevant capabilities first. So instead of evaluating a tuned model intended for deployment (which can refuse to display some capabilities), or a base model (which can have difficulties with displaying some capabilities), you need to tune the model to be more purely helpful, possibly in a way specific to the tasks it's to be evaluated on.