To understand “AI” it helps to understand “I” When my daughter (Hana) was 5 she ran upstairs to ask me a question. I answered. She said "I will be back when I need you again.” I wrote it down. Why? My field was AI, back when AI was about intelligence. I was working on my first major book (Scripts, Plans, Goals, and Understanding.) I watched my children carefully to see how and what they learned. I never programmed Hana to say that. I didn’t put the phrase in her mouth and tell he when she should say it. I simply answered her questions when she asked them. She knew that there would be more questions, so she came up with the idea that instead of saying good bye she needed to invent a way to say "see you soon." All the talk about AI these days relates in no way to self refection, to knowing what you need to know, or to anticipating the future. We talk about “AI” but we are not talking about the “I”. We have intelligent entities already. (They are called humans.) When they are confused they ask for explanations. When today's so-called “AI’s” start doing that, please let me know. In the meantime, it would be nice if there weren’t an article a day in major publications about AI when what they mean is number crunching and pattern matching, not wondering and trying to find out.
Katya Lewis🤔
Great article, reminded me of my young granddaughter who complained of 'fizzy fingers' she actually had 'pins and needles' but she was giving the best description she had, I don't think her phraseology is any less valid, indeed it sounds a lot less painful!
Sensible!
So well put. Makes one think, can AI wonder?...
Computational metacognition is about self-understanding for AI agents. It's not the "AI" that is being hyped about though.
Bravo, bravissimo! What is happening now is analogous to the "Turk Automaton" frenzy. "Artificial", that is easy to understand, but "Intelligence" we must learn from children. That does not mean that we need to replicate a physiological brain, but the structure of intelligence, its meta-characteristics, its mechanisms. the sheer tonnage of observations required in "learning" should be one clue, as are the adversarial practical jokes one can play on objects supposedly made "intelligent".
Yeah, but "machine learning", "data mining", and "pattern matching" haven't been turned into futuristic concepts that people think they understand via popular culture yet. :)
Whilst I agree that many may assume/believe that current AI (Cognitive Computing/ML et al.) can mirror high levels of human intelligence (which it can't yet and may never), it is a phrase that has many meanings and most importantly is getting people talking about this subject at all levels, which is VERY important. AI is an umbrella term and not a fixed thing. Intelligence also comes in many forms. Can someone please define the lowest level of intelligence possible or acceptable ? There is a theory that individual cells have consciousness and could be considered intelligent. Science must avoid being too dogmatic about this or it will risk alienating the very people this should be helping. Science is only great when everybody understands it and benefits from it.
Founding Head Of Research @NukkAI
4yReport
Report
Excellent :) And for your daughter