3 ms·

That is a very good question. I think I read Sam Altman saying "AGI is everything we haven't done yet. Something new that will feel like magic.". Sadly, I can't

That is a very good question. I think I read Sam Altman saying "AGI is everything we haven't done yet. Something new that will feel like magic.". Sadly, I can't find a quote and URL so I can't confirm Sam Altman said it. But the concept resonated with me. We have had Google Translate for a decade. It was and is simply a miracle of technology straight out of sci-fi books. Yet we grew quickly accustomed to it, taking it for granted. So, it's a bit the same with the concept of AGI. Now chatGPT 4.0 is the new normal and is taken for granted. So AGI is essentially anything better than ChatGPT 4.0 Frankly, I don't know if we are close to AGI. We could be years away or thousands of years away. I'll mention some past discoveries and how people reacted to put things in perspective. We went to the moon in 1969. We then dreamed we would be colonizing the galaxy. Yet space is very big and empty. If Starship lands on Mars 10 years from now, 65 years will have passed between the Moon and Mars landing. Other solar systems and Earth-like planets are much, much, much further away than Mars. It's the same in AI. We might just have landed on the moon, and we might have all these dreams about colonizing other planets (creating AGI). We might be technically very far, far away from it. We will make steady progress, and we will give names to each step of the process. AGI seems a generic name for the journey.

kingkongjaffa 3y ago
This is fine until "We will make steady progress", that's not guaranteed, in fact there's already a phrase for this https://en.wikipedia.org/wiki/AI_winter https://en.wikipedia.org/wiki/AI_winter With space travel, there is a very well established physics to model and predict how we might do these things, you need to do some stuff like build rockets, have them escape earth's gravity, build things in space. Everything is a known known or a known unknown but we have the physics to model everything involved with space travel. This is an entirely different class of problem to "creating AGI", since there is basically a huge unknown unknown ocean to cross. So far everything we have done in AI is a statistical proxy for real thought processes, and there's one assumption that says lets make more and more of these kinds of AI techniques with "bigger" systems and we might approach something like AGI. But it isn't certain. so far with the GPT's we've really only succeeded in synthesizing human written text corpus into a statistical prediction engine, which can synthesise useful text, but does not have any semblance of 'thought'.
- JoeAltmaier 3y ago
  That changes every day. The evidence is all in favor of 'steady progress'.

kingkongjaffa 3y ago

This is fine until "We will make steady progress", that's not guaranteed, in fact there's already a phrase for this https://en.wikipedia.org/wiki/AI_winter https://en.wikipedia.org/wiki/AI_winter With space travel, there is a very well established physics to model and predict how we might do these things, you need to do some stuff like build rockets, have them escape earth's gravity, build things in space. Everything is a known known or a known unknown but we have the physics to model everything involved with space travel. This is an entirely different class of problem to "creating AGI", since there is basically a huge unknown unknown ocean to cross. So far everything we have done in AI is a statistical proxy for real thought processes, and there's one assumption that says lets make more and more of these kinds of AI techniques with "bigger" systems and we might approach something like AGI. But it isn't certain. so far with the GPT's we've really only succeeded in synthesizing human written text corpus into a statistical prediction engine, which can synthesise useful text, but does not have any semblance of 'thought'.

JoeAltmaier 3y ago
That changes every day. The evidence is all in favor of 'steady progress'.