[ExI] LLMs and Computability
jasonresch at gmail.com
Wed Apr 19 23:20:45 UTC 2023
On Wed, Apr 19, 2023, 5:51 PM Darin Sunley via extropy-chat <
extropy-chat at lists.extropy.org> wrote:
> There's a running thread in LLM discourse about how LLMs don't have a
> world model and therefore are a non-starter on the path to AGI.
> And indeed, on a surface level, this is true. LLMs are a function - they
> map token vectors to other token vectors. No cognition involved.
> And yet, I wonder if this perspective is missing something important. The
> token vectors LLMs are working with aren't just structureless streams of
> tokens. they're human language - generated by human-level general
> That seems like it's the secret sauce - the ginormous hint that got
> ignored by data/statisticis-centric ML researchers. When you learn how to
> map token streams with significant internal structure, the function your
> neural net is being trained to approximate will inevitably come to
> implement at least some of the processing that generated your token
> It won't do it perfectly, and it'll be broken in weird ways. But not
> completely nonfunctional. Actually pretty darned useful, A direct analogy
> that comes to mind would be training a deep NN on mapping assembler program
> listings to output. What you will end up with is a learned model that, to
> paraphrase Greenspun's Tenth Rule, "contains an ad hoc,
> informally-specified, bug-ridden, slow implementation of half of" a
> Turing-complete computer.
> The token streams GPT is trained on represent an infinitesimal fraction of
> a tiny corner of the space of all token streams, but they're token streams
> generated by human-level general intelligences. This seems to me to suggest
> that an LLM could very well be implementing significant pieces of General
> Intelligence, and that this is why they're so surprisingly capable.
I believe the only way to explain their demonstrated capacities is to
assume that they create models. Here is a good explanation by the chief
researcher at OpenAI:
As designed, a single invocation of GPT cannot be Turing complete as it has
only a finite memory capacity and no innate ability for recursion. But a
simple wrapper around it, something like the AutoGPTs, could in theory
transformer GPT into a Turing complete system. Some have analogized single
invocations of GPT as like single instructions by a computer's CPU.
(Or you might think of it like a single short finite time of processing by
a human brain)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat