I would suggest the textbook Artificial Intelligence: A Modern Approach by Russell and Norvig. It’s a good overview of the field and has been in circulation since 1995. https://en.m.wikipedia.org/wiki/Artificial_Intelligence:_A_Modern_Approach

Here’s a photo, as an example of how this book approaches the topic, in that there’s an entire chapter on it with sections on four approaches, and that essentially even the researchers have been arguing about what intelligence is since the beginning.

But all of this has been under the umbrella of AI. Just because corporations have picked up on it, doesn’t invalidate the decades of work done by scientists in the name of AI.

My favourite way to think of it is this: people have forever argued whether or not animals are intelligent or even conscious. Is a cat intelligent? Mine can manipulate me, even if he can’t do math. Are ants intelligent? They use the same biomechanical constructs as humans, but at a simpler scale. What about bacteria? Are viruses alive?

If we can create an AI that fully simulates a cockroach, down to every firing neuron, does it mean it’s not AI just because it’s not simulating something more complex, like a mouse? Does it need to exceed a human to be considered AI?

I think you’re conflating “intelligence” with “being smart”.

Intelligence is more about taking in information and being able to make a decision based on that information. So yeah, automatic traffic lights are “intelligent” because they use a sensor to check for the presence of cars and “decide” when to switch the light.

Acting like some GPT is on the same level as a traffic light is silly though. On a base level, yes, it “reads” a text prompt (along with any messaging history) and decides what to write next. But that decision it’s making is much more complex than “stop or go”.

I don’t know if this is an ADHD thing, but when I’m talking to people, sometimes I finish their sentences in my head as they’re talking. Sometimes I nail it, sometimes I don’t. That’s essentially what chatGPT is, a sentence finisher that happened to read a huge amount of text content on the web, so it’s got context for a bunch of things. It doesn’t care if it’s right and it doesn’t look things up before it says something.

But to have a computer be able to do that at all?? That’s incredible, and it took over 50 years of AI research to hit that point (yes, it’s been a field in universities for a very long time, with most that time people saying it’s impossible), and we only hit it because our computers got powerful enough to do it at scale.

For laymen who might not know how GANs work:

Two AI are developed at the same time. One that generates and one that discriminates. The generator creates a dataset, it gets mixed in with some real data, then that all of that gets fed into the discriminator whose job is to say “fake or not”.

Both AI get better at what they do over time. This arms race creates more convincing generated data over time. You know your generator has reached peak performance when its twin discriminator has a 50/50 success rate. It’s just guessing at that point.

There literally cannot be a better AI than the twin discriminator at detecting that generator’s work. So anyone trying to make tools to detect chatGPT’s writing is going to have a very hard time of it.