4 ms·

This is pretty incredible. I guess the thing that we have to keep in mind is that we really can't compete with an AI that understands natural language this well

This is pretty incredible. I guess the thing that we have to keep in mind is that we really can't compete with an AI that understands natural language this well and also has the entirety of the internet as its reference materials. Maybe we need to accept that or rework our definition of what makes humans special

hgsgm 3y ago
More that most undergrad colleges only teach and test regurgitation because the profs don't care about education.
- jakeinspace 3y ago
  I took the graduate version of this course, and thankfully did better on this exam than GPT-4. It was a fairly rigorous course, and getting an A required about 20 hours per week of work.
  - scotty79 3y ago
    > ... required about 20 hours per week of work. Of thinking or of remembering?
    - jakeinspace 3y ago
      There were weekly problem sets. The questions on these tended to be a bit more in depth than the midterm/exam questions.
- kaesar14 3y ago
  Fwiw Scott Aaronson is maybe the most well respected researcher in this field. I don't think this is a case of some shitty professor asking his students to regurgitate useless information
- moonchrome 3y ago
  Have you actually tried using GPT 4 and 3.5 ? I'd say 3.5 could be called a regurgitation machine with a few tricks, GPT 4 seems to apply concepts and feels like showing signs of human level intelligence. If the rate of progress continues for a few years we are collectively fucked.
  - ewhanley 3y ago
    Yeah, the improvement from 3.5 to 4 is amazing. I'm surprised so many people are so dismissive, and I suspect it's because most of them haven't tried it out. Such are the ever-moving ML/AI goalposts.
- sebzim4500 3y ago
  That isn't true of this course though. I've read the lecture notes, it really is a great introduction to the subject.
opdahl 3y ago
What this technology has shown is that is basically superhuman in the _amount_ of «expert level knowledge» it has. Not completely sure if we have said that what makes human special is the amount knowledge an individual can acquire. What we still see lacking, and I would wager will still pertain if we keep just training on the internet, is that it is not good at reasoning to generate _new_ knowledge. It is quite able to compare etc, or give examples but you can clearly see that its just in a way regurgitating some information with a twist. Hopefully we can find a way to use this technology to enhance the parts that do make us human, whatever your definition of that is.
- tudorw 3y ago
  quite, augment, not replace
kaesar14isazero 3y ago
[dead]

hgsgm 3y ago

More that most undergrad colleges only teach and test regurgitation because the profs don't care about education.

jakeinspace 3y ago
I took the graduate version of this course, and thankfully did better on this exam than GPT-4. It was a fairly rigorous course, and getting an A required about 20 hours per week of work.
- scotty79 3y ago
  > ... required about 20 hours per week of work. Of thinking or of remembering?
  - jakeinspace 3y ago
    There were weekly problem sets. The questions on these tended to be a bit more in depth than the midterm/exam questions.
kaesar14 3y ago
Fwiw Scott Aaronson is maybe the most well respected researcher in this field. I don't think this is a case of some shitty professor asking his students to regurgitate useless information
moonchrome 3y ago
Have you actually tried using GPT 4 and 3.5 ? I'd say 3.5 could be called a regurgitation machine with a few tricks, GPT 4 seems to apply concepts and feels like showing signs of human level intelligence. If the rate of progress continues for a few years we are collectively fucked.
- ewhanley 3y ago
  Yeah, the improvement from 3.5 to 4 is amazing. I'm surprised so many people are so dismissive, and I suspect it's because most of them haven't tried it out. Such are the ever-moving ML/AI goalposts.
sebzim4500 3y ago
That isn't true of this course though. I've read the lecture notes, it really is a great introduction to the subject.

opdahl 3y ago

What this technology has shown is that is basically superhuman in the _amount_ of «expert level knowledge» it has. Not completely sure if we have said that what makes human special is the amount knowledge an individual can acquire. What we still see lacking, and I would wager will still pertain if we keep just training on the internet, is that it is not good at reasoning to generate _new_ knowledge. It is quite able to compare etc, or give examples but you can clearly see that its just in a way regurgitating some information with a twist. Hopefully we can find a way to use this technology to enhance the parts that do make us human, whatever your definition of that is.

tudorw 3y ago
quite, augment, not replace

kaesar14isazero 3y ago

[dead]