What is your measure of intelligence?
So I'll leave it to Skeeter to explain.
If you go by that then a lot of people (no offense) aren't intelligent. This includes many vastly successful or rich people.
So I disagree. There's a lot of ways to be intelligent. Not just the research and scientific type.
I probably agree that most people aren't engaged very often and even when they are they suck at being awesome but that really isn't the bar being mentioned here.
Right
> but that really isn't the bar being mentioned here.
Yes, the bar was "novel solutions WITHOUT priori knowledge"
So you've changed the definition. Please re-read what I disagree with and it's not just the novel part i.e. if I read it all from the Internet and copied it to be successful then that fails this definition.
I think most people agree with this statement.
I am struggling to think of anything that can be considered a solution and can be created without "priori" knowledge.
But if I had to guess, I believe they'd argue that an LLM is basically all a priori knowledge. It is trained on a massive data set and all it can do once trained is reason from those initial axioms (they aren't really axioms, but whatever). While humans, and actually many other animals to a lesser extent, can make observations, challenge existing assumptions, generalize to solve problems, etc.
That's not exactly my definition of intelligence, but that might be what they were going for.
So, if we look at it from this perspective, human thinking is not fundamentally different from LLMs in that both rely on existing material to create new ideas.
The main difference is that LLMs process text statistically, while humans interpret text in context, influenced by emotions, experiences, biases, and goals. LLMs' interpretation is probabilistic, not conceptual.
Additionally, revolutionary thinking often requires rejecting past ideas and forming new conceptual frameworks, but LLMs cannot reject prior data, they are bound by it.
At any rate, the question remains, are LLMs capable of revolutionary ideas just like humans?
In that way, my dog is far more intelligent than LLM, in that he has a mental model of his world. An LLM is only intelligent relative to a human actor, and so it is no different than any other technology that humans have created to pursue their own ends.
On the computer side of things, I think at a minimum I'd want intelligence capable of taking advantage of the fact that it's a deterministic machine capable of unerringly performing various operations with perfect accuracy absent a stray cosmic ray or programming bug. Star Trek's Data struggled with human emotions and things like that, but at least he typically got the warp core calculations correct. Accepting LLMs with the accuracy of a particularly lazy intern feels like it misses the point of computers entirely.
What is most characteristic about human intelligence is the ability to abstract from particular, concrete instances of things we experience. This allows us to form general concepts which are the foundation of reason. Analysis requires concepts (as concepts are what are analyzed), inference requires concepts (as we determine logical relations between them).
We could say that computers might simulate intelligent behavior in some way or other, but this is observer relative not an objective property of the machine, and it is a category mistake to call computers intelligent in any way that is coherent and not the result of projecting qualities onto things that do not possess them.
What makes all of this even more mystifying is that, first, the very founding papers of computer science speak of effective methods, which is by definition about methods that are completely mechanical and formal, and this stripped of the substantive conceptual content it can be applied to. Historically, this practically meant instructions given to human computers who merely completed them without any comprehension of what they were participating in. Second, computers are formal models, not physical machines. Physical machines simulate the computer formalism, but are not identical with the formalism. And as Kripke and Searle showed, there is no way in which you can say that a computer is objectively calculating anything! When we use a computer to add two numbers, you cannot say that the computer is objectively adding two numbers. It isn’t. The addition is merely an interpretation of a totally mechanistic and formal process that has been designed to be interpretable in such ways. It is analogous to reading a book. A book does not objectively contains words. It contains shaped blots of pigment on sheets of cellulose that have been assigned a conventional meaning in a culture and language. In other words, you being the words, the concepts, to the book. You bring the grammar. The book itself doesn’t have them.
So we must stop confusing figurative language with literal language. AI, LLMs, whatever can be very useful, but it isn’t even wrong to call them intelligent in any literal sense.
Intelligence is what we call problem solving when the class of "problem" that a being or artifact is solving is extremely complex, involves many or near uncountable combinations of constraints, and is impossible to really characterize well. Other than examples, of data points, and some way for the person or artifact to extract something general and useful from them.
Like human languages and sensibly weaving together knowledge on virtually every topic known to humans, whether any humans have put those topics together before or not.
Human beings have widely ranging abilities in different kinds of thinking, despite our common design. Machines, deep learning architectures, underpinnings are software. There are endless things to try, and they are going to have a very wide set of intelligence profiles.
I am staggered how quickly people downplay the abilities of these models. We literally don't know the principles they have learned (post training) for doing the kinds of processing they do. The magic of gradient algorithms.
They are far from "perfect", but at what they do there is no human that can hold a candle to them. They might not be creative, but I am, and their versatility in discussing combinations of topics I am fluent in, and am not, is incredibly helpful. And unattainable from human intelligence. Unless I had a few thousand researchers, craftsman, etc. all on a Zoom call 24/7. Which might not work out so well anyway.
I get that they have their glaring weaknesses. So do I! So does everyone I have ever had the pleasure to meet.
If anyone can write a symbolic or numerical program to do what LLM's are doing now - without training, just code - even on some very small scale, I have yet to hear of it. I.e. someone who can demonstrate they understand the style of versatile pattern logic they learn to do.
(I am very familiar with deep learning models and training algorithms and strategies. But they learn patterns suited to the data they are trained on, implicit in the data that we don't see. Knowing the very general algorithms that train them doesn't shed light on the particular pattern logic they learn for any particular problem.)
Plus, it's relatively straightforward and inexpensive using contemporary tech to build a roomba-like machine that wanders about on any flat surface cuing up and driving nails of its own accord with no human intervention.
If computers do not add numbers, then neither do people. It's not like you can do an addition-style turing test with a human in one room and a computer in another with a judge partitioned off of both of them, feed them each an addition problem and leave the judge in any position where they can determine which result is "really a sum" and which one is only pretending to be.
Yet if you reduce far enough to claim that humans aren't "really" adding numbers either, then you are left to justify what it would even mean for numbers to "really" get added together.
As it stands, we don't even know of any functions that exceeds the Turing complete, but are computable.
That would require the universe to be discrete, we don't know that. Otherwise most continuous processes compute something that a Turing machine can't, the Turing machine can only approximate it.