Is “understanding” a thing that requires consciousness? People say a machine can never be conscious. I say it’s not been shown to be necessary. When a brain understands something, is that not some combination of deterministic and stochastic processes?
Who knows, probably, but whatever is going on there is more complicated than this actually pretty basic system.
Our very human brains get confused when confronted by large datasets (or really large anythings) and tend to attribute much more intent to these things.
If I created a program that did this same kind of trick with a much smaller dataset so that it only was capable of generating let’s say a total of 100 possible sentences using the same statistical methods that this does, would you consider that to be showing creativity on the part of the program as opposed to on my part? That is exactly what this is but with a huge data set so it has many more possible outcomes.
You are of course welcome to believe whatever you want about creativity and GPT. I just wouldn’t apply for a job at, say, Google espousing the idea that this is creativity
Oh-Emm-Geee, “Carry me back to old Virginia!” As a Mac Classic owner that spent hours going back and forth with ELIZA in the System 6 & 7 days, reading that name transported me back to the one mouse button years of my youth when The Talking Moose would pop in every now and again to say something pithy. ^____^
I agree that GPT is essentially a sort of simple process, though it is considerably more involved than a Markov chain.
In my other hand, I have another sonnet made by someone or something. It’s in the style of Shakespeare and conforms to the sonnet structure. It’s about a woman who has black wires growing out of her head. Is it creative?
B. It depends. How was it made?
C. It depends. Is it any good?
E. Do you enjoy it? Then who cares where it came from
In this particular piece of output, her voice is unpleasant and her breath stinks, so “enjoyable” might not be the word.
Well, no, that would be tautological. The definition is separate and happens to be something that AI can’t currently do. Driving a car is not defined as “something monkeys can’t do”, it’s an independently defined function that monkeys don’t currently do (I hope).
One of the challenges here in defining it is that neuroscience doesn’t currently know how it works. It’s a bit like porn at the moment in that we* know it when we see it, and we know from the code that these AIs are not doing it. But we can’t say precisely how it works in brains either.
*Royal ‘we’ here, I am not a neuroscientist
Who says that? There is no hard definition or test for consciousness, so that statement isn’t valid or even falsifiable.
However nobody that studies this stuff that I’ve spoken with would ever say it’s impossible for AIs to become conscious. General consensus on the human brain is that consciousness is an emergent property of the very complex interactions of all the chemicals and electrical signals going on. There’s no reason to believe an AI couldn’t someday reach that. We have no idea how to begin to even think about building such an AI since we have very little understanding of how the brain does it, but we know enough to know it should be possible.
The only reason it couldn’t be possible is if you’re a dualist, and dualists are pseudoscientific crackpots. Religion hiding in sciencey words.
This is likely* correct, and curiously contradicts your earlier statement. You are describing the scientific consensus as I described above. However that’s a very very different ballpark from saying ChatGPT or Midjourney are creating. The human brain is a skyscraper and current AIs are one rock set on top of another. Both are construction projects, but there’s no path from one to the other that doesn’t require several hundred complete paradigm shifts in how to build things.
I hope this is gelling because I’m running out of metaphors.
*I say “likely” because we don’t know how deterministic the brain really is. There is evidence of quantum-level effects in cognition that would inject a lot of true randomness. In general though, the general consensus seems to be that the brain is a very very complicated machine that is probably deterministic if we could ever hope to understand or simulate every detail
Yes, Sir Roger Penrose has some interesting notions about that. Something about deterministic biological processes that are in some fashion shaken up by (truly) random quantum impulses.
Some people say that evolution is impossible, because random processes can’t create anything. But random processes that are filtered, so to speak, can indeed cause new things, and ways of doing things, to come into existence. To me, that’s creative. You can do a lot of creative stuff with filtered noise.
“Indeed, it is obvious that invention or discovery, be it in mathematics or anywhere else, takes place by combining ideas.”
Hadamard, J. (1949). An essay on the psychology of invention in the mathematical field. Princeton Univ. Press.
Of course combining ideas is an aspect of creativity. That’s obvious. This doesn’t refute anything I’ve said on these points. If you’re interested in debating the finer points of why these algorithms are not creative, I’m up for it, but drive-by quote-bombing is not going to do it.
My position is that all these systems are doing is pastiche. And we’ve been able to do pastiche for as long as the database has been large enough; for centuries, at least.
The only real difference here is that the database is effectively infinite in size. Doesn’t stop the output from being mere pastiche though. (Which may explain why it’s so good at financial and sports reports, which probably says more about the poor state of those segments of journalism than about the engines.)
Our chat bot efforts on #beer at Waterloo were modest; it was more fun to have one that was insane. The scripts from Star Wars, the Complete Works of Shakespeare, the Bible, and a stack of academic papers on the Iraq War make for an interesting starting point…
But hear, hear…
For a while one of my favourite questions for AI researchers was: feed your AI everything you know about integers and (just to be sporting) rational numbers. Now: discover real numbers. Next: discover complex numbers.
If I may wander out onto a long limb here…
We get these advances in math when we need a larger structure than can be merely deduced, usually when there’s a hole in the system (enter Kurt Goedel, stage left). IMHO true creativity is the creation of structures not present in the data which then explain larger truths that can be verified by experiment.
Yes, I did indeed actually grow up around particle physicists…
How this maps onto the humanities, I dare not speculate.
That’s a really good analogy!
That’s as good a definition as I’ve ever heard. Well put!
This topic was automatically closed after 5 days. New replies are no longer allowed.