Live data from Hacker News

Do you have to be intelligent to develop AI?

news.ycombinator.com

1–6 of 6 posts

Do you have to be intelligent to develop AI?

#1
I wonder if being intelligent, say (IQ >= 110), is a necessary condition to develop AI. Many machine learning courses begin with "You don't need a lot of math" so their recipients are not supposed to be very smart and still they are going to develop the new AI.

Re: Do you have to be intelligent to develop AI?

#4

  begin with "You don't need a lot of math" so their recipients are not supposed to be very smart
I would be weary about correlating high IQ with a high maths aptitude (and also of the reverse) -- since other things effect maths development more than IQ (The type of teacher you have, etc.); I had the pleasure of knowing quite a few people in MENSA over the duration of my childhood, and I can assure you that many of them had exactly the same problems with mathematics as the rest of the population.

I disagree that a high IQ is a necessary component of developing an AI (And I assume you're talking about Strong AI here); I would say that an expert in the field has a much better chance than the average developer, but since nobody (To my knowledge) has created a general AI we can't really know for sure.

It would be interesting to see if AI researchers (or mathematicians, for that matter) tend to have a high IQ compared to other fields/jobs...

Re: Do you have to be intelligent to develop AI?

#5
Perhaps my idea is that machine learning and AI is now a way to sell snake oil, specially when you try to give recipes and avoid the foundation. Today computer programs can do many task as for example predictions using models which require little o no tuning, so the user just press a button or write two or three lines of code. Someone can think that he is at the cutting edge of research pressing those buttons, perhaps that naive feeling is what I was fighting again.

Re: Do you have to be intelligent to develop AI?

#6
post #5

Perhaps my idea is that machine learning and AI is now a way to sell snake oil, specially when you try to give recipes and avoid the foundation. Today computer programs can do many task as for example predictions using models which require little o no tuning, so the user just press a button or write two or three lines of code. Someone can think that he is at the cutting edge of research pressing those buttons, perhap…

AI dev is only pushing a few buttons: it's right it's only pushing a few buttons but it takes an education to know which button to push. Even if it feels like pushing buttons doesn't mean it's not cutting edge because that's your feeling, not the feeling of every one else. Maybe tech has reached a settled so far and the next plateau is not reached yet.

Also, AI is not the only area where progress can be made. Management of complex (complete?) systems is also an area of interest.

As for strong AI, I have a the feeling that those system will be much like sysadmin work of today.