4 ms·

ChatGPT and friends look magical, but as yet no great revolution in working practices or whatever has occurred. So the excitement about AI is based on 'very lar

ChatGPT and friends look magical, but as yet no great revolution in working practices or whatever has occurred. So the excitement about AI is based on 'very large future promise', which is exactly the kind of thing that ultra low interest rates tend to pump.

roenxi 3y ago
I know a bunch of people who are very excited about AI and using it in a professional context. Their feelings are inspired by demonstrated and past performance, the future promise is so big that I don't think anyone can process it. I don't think they understand what exponential growth in processing power implies.
- aleph_minus_one 3y ago
  > the future promise is so big that I don't think anyone can process it. Sounds like a sentence that was taken from some pulpiteer's sermon when talking about god or afterlife. In other words: this promise is rather some nice fairy tale for some naive nerds and investors who love to believe in it. > I don't think they understand what exponential growth in processing power implies. This is very easy: for the data points (x_i, y_i), you can draw a decent straight line through the data transformed data points (x_i, log y_i). Thus: exponential growth is just some fancy term for "we do linear regression/linear Taylor approximation in this transformed space". I'm deeply impressed by such deep mathematics *sarc*.
  - roenxi 3y ago
    Although "we do linear regression ... in this transformed space" is correct, it also isn't an attempt to understand what exponential growth in processing power implies. That is an attempt (really, a success) at providing a definition of what exponential growth is [0]. Implications involves things like https://ourworldindata.org/brief-history-of-ai https://ourworldindata.org/brief-history-of-ai + little extrapolation. Interestingly, as I look at that article, it appears AI is actually solving college level math now - https://blog.research.google/2022/06/minerva-solving-quantitative-reasoning.html https://blog.research.google/2022/06/minerva-solving-quantit.... I wish I'd had that when I was at University. [0] Ironically, ChatGPT doesn't usually make that mistake, because it always tries to be calm and helpful it consistently makes a better-than-human attempt at reading comprehension whenever I've interacted with it.
    - soco 3y ago
      > I wish I'd had that when I was at University. Looking back, what use do you think you would have today by having it in university? I'm asking because I don't think such a help solving the exams would have helped in your career later. Yes, would have made an easier university time, and left you less prepared for after that. We can (continue to) use ChatGPT for everything, it's just this will be also eroding our own capacity to do that thing, less exercise of our mental muscle. Maths and puzzles are said to help maintaining mental health with age, by outsourcing them to AI we also lose faster our own I.
      - imtringued 3y ago
        His wish is akin to "I wish the AI lived my entire life for me".
        roenxi 3y ago
        You're being melodramatic. Humans have made extensive use of tools since well before the dawn of recorded history. There is nothing unusual about someone wanting to use a tool. I find it easier to learn when someone (now, something) intelligent is on hand to explain the material. You both must be geniuses or psychics, because otherwise it is not clear how you gentlemen intend to learn otherwise.
  - tcbawo 3y ago
    Investors in the dot com bubble looked at a straight line and thought it would extend forever. Now investors are looking at an exponential line, expecting it to extend forever. It probably won’t and some will be disappointed. Ultimately, predictions of the future always look rosier and more advanced than reality. Progress is inexorable, undeniable, and arrives in unexpected ways.
- somewhereoutth 3y ago
  > future promise is so big Indeed! But as yet it is still just future promise - not much has materialized at large scale so far (to my knowledge - maybe copy writing? Seems to be a push back against obviously AI scripted content though). A counter example might be spreadsheets - their introduction revolutionized book keeping (and lots of other things too). We wait to see whether AI will have a similar impact (I suspect it won't).

roenxi 3y ago

I know a bunch of people who are very excited about AI and using it in a professional context. Their feelings are inspired by demonstrated and past performance, the future promise is so big that I don't think anyone can process it. I don't think they understand what exponential growth in processing power implies.

aleph_minus_one 3y ago
> the future promise is so big that I don't think anyone can process it. Sounds like a sentence that was taken from some pulpiteer's sermon when talking about god or afterlife. In other words: this promise is rather some nice fairy tale for some naive nerds and investors who love to believe in it. > I don't think they understand what exponential growth in processing power implies. This is very easy: for the data points (x_i, y_i), you can draw a decent straight line through the data transformed data points (x_i, log y_i). Thus: exponential growth is just some fancy term for "we do linear regression/linear Taylor approximation in this transformed space". I'm deeply impressed by such deep mathematics *sarc*.
- roenxi 3y ago
  Although "we do linear regression ... in this transformed space" is correct, it also isn't an attempt to understand what exponential growth in processing power implies. That is an attempt (really, a success) at providing a definition of what exponential growth is [0]. Implications involves things like https://ourworldindata.org/brief-history-of-ai https://ourworldindata.org/brief-history-of-ai + little extrapolation. Interestingly, as I look at that article, it appears AI is actually solving college level math now - https://blog.research.google/2022/06/minerva-solving-quantitative-reasoning.html https://blog.research.google/2022/06/minerva-solving-quantit.... I wish I'd had that when I was at University. [0] Ironically, ChatGPT doesn't usually make that mistake, because it always tries to be calm and helpful it consistently makes a better-than-human attempt at reading comprehension whenever I've interacted with it.
  - soco 3y ago
    > I wish I'd had that when I was at University. Looking back, what use do you think you would have today by having it in university? I'm asking because I don't think such a help solving the exams would have helped in your career later. Yes, would have made an easier university time, and left you less prepared for after that. We can (continue to) use ChatGPT for everything, it's just this will be also eroding our own capacity to do that thing, less exercise of our mental muscle. Maths and puzzles are said to help maintaining mental health with age, by outsourcing them to AI we also lose faster our own I.
    - imtringued 3y ago
      His wish is akin to "I wish the AI lived my entire life for me".
      - roenxi 3y ago
        You're being melodramatic. Humans have made extensive use of tools since well before the dawn of recorded history. There is nothing unusual about someone wanting to use a tool. I find it easier to learn when someone (now, something) intelligent is on hand to explain the material. You both must be geniuses or psychics, because otherwise it is not clear how you gentlemen intend to learn otherwise.
- tcbawo 3y ago
  Investors in the dot com bubble looked at a straight line and thought it would extend forever. Now investors are looking at an exponential line, expecting it to extend forever. It probably won’t and some will be disappointed. Ultimately, predictions of the future always look rosier and more advanced than reality. Progress is inexorable, undeniable, and arrives in unexpected ways.
somewhereoutth 3y ago
> future promise is so big Indeed! But as yet it is still just future promise - not much has materialized at large scale so far (to my knowledge - maybe copy writing? Seems to be a push back against obviously AI scripted content though). A counter example might be spreadsheets - their introduction revolutionized book keeping (and lots of other things too). We wait to see whether AI will have a similar impact (I suspect it won't).