This is not as strong of an argument as you seem to think.
According to NHTSA [0] there are about 2.1 accidents per million miles driven. This includes fatality, injury-only and property-damage-only accidents. That is the equivalent of over 99.999% of miles driven without an accident. Over 5 nines of reliably consistent behavior.
This is a fair point but if what we're looking for is consistent behavior I think we'd have to consider events that don't result in damage or even rise to being accidents (but that could have, if fortune had frowned) like being cut off in a merge or someone running a no-turn-on-red. Which is of course difficult to really measure.
Better to say, I think, that humans are much better at improving their approximation of consistency with mental effort both because we can think silently instead of "out loud, step by step" and because some of the patterns of careful thought we engage in don't get written down naturally as text and so are unlikely to be hit upon by GPTs. The advantages of not being a strict feed foreword network.
That being said, GPT4 can just open its virtual mouth spew forth without reflection and still produce consistent text is clearly superhuman.
I think framing it this way could actually help us reach the next step with AI, if we ask how we could imbue it with those properties.
A human neural net is constantly bombarded with inputs from many different senses, which firstly, gets prioritized based on prior usefulness. That usefulness is updated all the time, constantly, and that's what any current AI implementation lacks.
1 - The continuous integration of data from all "senses".
This one should be self-evident, as obviously, all our senses are constantly barraging our brain with data and it learns to handle that over time, in whichever way your genetic makeup + learned internal cognitive processes dictate it be handled.
2 - The network that decides which data requires which amount of attention, and whether to store it in short or long term memory. This is obviously tied in quite closely to 1, as you need massive amounts of data to understand underlying patterns, and which data is just spam, versus what's really valuable.
3 - And with these two things together come the emergence of improving of
approximation of consistency. Which means, this itself is a metric which the
agent running the other agents needs to be aware of. Its silly to think the human
brain is a single agent. It makes way more sense to see it as various interacting
agents that equate to a greater sum than its parts.
Now, that being said, I'm not an expert on AI or Data Science, but this is more or less how my understanding of computational theory of mind meets with biological computing and neural networks.
My theory is that, the first actually intelligent AI will be one that is composed of a network that makes decisions on how to spend a unit of iteration. One iteration becomes "an instant" to the AI. Aka, the AI decides to spend one iteration thinking, or spawns a sub-agent (which it is aware will consume resources that other processes might also need access to, but it needs to be able to decide which action to pursue).
So in all honesty, its amazing to me that LLM's on their own have been able to achieve this level of "personhood" despite their being only a tiny subset of the whole that makes up a "conscious" entity.
Right. I sort of worry that because LLMs are able to be so coherent without anything resembling a human's working memory at all, then adding either a crude working memory still working with LLM tokens or adding a more sophisticated one assembled from "chunks"[1] will let an AI based on them ascend to clearly superhuman reasoning with just an architectural improvement and no need for any more flops invested than we already put into GPT4.
I'm not a mathematician, so I'm not at the horizon of mathematics - but I can branch off from what's already there to make what I need (and isn't just searchable) consistent with the rest of it.
You're arguing that recent AI developments are a big deal and I'm not arguing against that. But what anyone stating that needs to answer is why we should think that big deal is a good thing - since humans are using the technology and will control whom it benefits. We don't have a good track record there, historically, which is another are in which we are, sadly, consistent.
Economics depends on looking at large enough datasets that inconsistencies can be averaged it and glossed over. People generally following a pattern and precise consistency are very different
To me, none of what you mention requires a whole lot of consistency other than in as much as we understand that we are horribly inconsistent we have a whole lot of ceremony and processes built up around what we do to mitigate the wild inconsistencies in the quality of work we do.
> How does the financial industry work? How are people comfortable executing transactions?
Because there are serious checks and balances? Because of double entry bookkeeping, reconciliation, audits and, ultimately, prisons? Systems of governance put into place with the explicit goal of ameliorating the vagueness of individual humans?
Aren't you mistaking the state and ability of being consistent, with the incentives for consistency? All these systems of governance would ultimately be useless without the ability to conform to them. And conform we do.
Maybe the ambiguity in this discussion is how to distinguish consistency from conformism?
That's a bold statement. Do you have evidence for this?