> Author Pamela McCorduck writes: "It's part of the history of the field of artificial intelligence that every time somebody figured out how to make a computer do something—play good checkers, solve simple but relatively informal problems—there was a chorus of critics to say, 'that's not thinking'." Researcher Rodney Brooks complains: "Every time we figure out a piece of it, it stops being magical; we say, 'Oh, that's just a computation.'"
> "The AI effect" tries to redefine AI to mean: AI is anything that has not been done yet.
> Some people think that as soon as AI successfully solves a problem, the problem is no longer a part of AI.
Fair, though in my case (which is insignificant and out of the norm) I've always wanted AI robots, as a kid I loved watching Dexter's Lab and seeing him make robots to do things. I'm also okay with a really powerful AI assistant beyond what we have today. I tried using AI assistants a few times, but I always end up shutting them off for either privacy concerns or just get tired of figuring out the word soup to make them work. I really want to see an AI assistant that is insanely capable and fully offline first, so maybe something like Mycroft, not sure how fully capable it is without reaching out to the internet though, but I assume it has some capabilities without being online.
In my eyes, AI is something that could pass the Turing test without fail, ever because it knows how to communicate how we do, and think like we do.
I've seen it unfolding real-time with the famous Lee Sedol / AlphaGo match. On the day before the first match, people were saying how the match would be a joke, as machines would never understand the deep philosophy of go. A few days later, people were saying that it was a matter of calculating stuff very fast and the match was clearly rigged because Lee only had one brain while AlphaGo had thousands of GPUs. (Yes, someone actually said that. Facepalm.)
Some would call that "machine learning". I know it's hard to make a distinction, but the term "AI" is too ill-defined to argue about what constitues "real AI" and what is mere machine learning. Are those examples "thinking machines" (~= "artificial _intelligence_"? I'd say no, they are very good statistical pattern matchers without any understanding of the subject matter for the most part.
GPT-3 and image generators that have somewhat of a world-model are, imo, closer.
Intelligence is a Heap Paradox and trying to define where the boundary between Artificial "Intelligence" and "really good pattern matching" is a fool's errand. Intelligence is a continuum from the simplest bang-bang thermostat all the way up to the human brain.
The joke I heard was that AI only refers to technology in the future, and not the past. We use plenty of machine learning models every day:
- Every time you use dictation
- Every time you use biometrics, like face scanning, to unlock your phone
- Every time you use any sort of search engine, news feed, or see an ad
- It is an integral part of the field of computer vision now, so anything to do with scanning documents, using AR, taking photos, etc uses AI
It’s only getting better and more integral to more fields over time.