5 ms·

> It is possible that we will have superintelligence in a few thousand days (!) "a few thousand days" is such a funny and fascinating way to say "about a decad

> It is possible that we will have superintelligence in a few thousand days (!) "a few thousand days" is such a funny and fascinating way to say "about a decade" Superficially, reframing it as "days" not "years" is a classic marketing psychology trick, i.e. 99 cents versus a dollar, but I think the more interesting thing is just the way it defamiliarizes the span. A decade means something, but "a few thousand days" feels like it means something very different.

7thaccount 2y ago
Probably because we all know "a decade" means maybe never. They've said the same thing about fusion for 5 decades.
- arthurcolle 2y ago
  https://www.energy.gov/articles/doe-national-laboratory-makes-history-achieving-fusion-ignition https://www.energy.gov/articles/doe-national-laboratory-make...
  - simianparrot 2y ago
    And it's still a decade away
  - Yizahi 2y ago
    Inertial confinement most likely can't be reasonably commoditized into generating electricity. Also, I may be wrong, but I think on the wave of that hype I read that the number they used was an incomplete one. Basically they reframed the conditions of the problem to get more favorable number for advertisement. NIF research will most likely be limited to the nuclear bomb research and simulations. But even if NIF had a reasonable path to energy generation in say 20-30 years, it still won't matter much, just like ITER probably won't. Solar will be way more cheap and widespread, it's costs are still dropping down ahead of all predictions, govt. or commercial. Fusion may just be a pure science project in the end, for a long time.
- dcchambers 2y ago
  Perhaps ironically, Sam Altman is also big into Fusion. (https://www.helionenergy.com/ https://www.helionenergy.com/) I guess it makes sense, for deep-learning/LLMs to "scale to infinity" you basically need infinite amounts of power.
  - bamboozled 2y ago
    cant we use OA intelligence to figure out how to do more with less rather yhan "stupidly" accelerate the climate crisis? the climate crisis + AI is the nightmare
    - 7thaccount 2y ago
      I don't think people on here who downvoted you understand the sheer scale of data center load growth. Utilities can handle some load growth sure, but the requests from the AI folks are crazy large to the tune of GW. I think it's fair to ask if training the next chat bot and related needs are worth doubling our electricity consumption.
    - fragmede 2y ago
      Jevons Paradox says that's not the direction it'll go in.
      - bamboozled 2y ago
        It's funny how some people seem to have different beliefs around this. Personally, I love making things more efficient, and still try to minimize consumption.
  - teaearlgraycold 2y ago
    And perhaps AI can help some of the problems in building a fusion reactor.
    - fragmede 2y ago
      Which is pretty exciting https://inl.gov/news/going-digital-idaho-students-engineers-demonstrate-first-nuclear-reactor-digital-twin/ https://inl.gov/news/going-digital-idaho-students-engineers-...
ocean2 2y ago
The Stellarator design has proven to be stable and to produce net positive energy. We are actually only $20B away from having a fully functional nuclear fusion reactor.
- samatman 2y ago
  Not really, unfortunately. The progress has been heartening, don't get me wrong, but it's more like $20B away from getting to the first steps of solving the next set of problems on the road to a fully functional fusion generator. Those are fairly substantial. It's past time to stop being cynical about "ten more years every ten years", but it's also way too soon to declare victory.
  - mjamesaustin 2y ago
    And ultimately, fusion doesn't solve the hardest problem with energy today which is cost. Solar already provides unlimited clean energy, and it took decades of development for the cost to drop to a competitive rate. Even if we had a functioning fusion plant tomorrow, it will likely cost WAY more per kWh and require decades of iteration and improvement to become economically feasible.
    - whamlastxmas 2y ago
      I would wonder if solar plus the absolute massive battery reserves you’d need for a data center ten times bigger than the current worlds largest would still be cost effective versus a single fusion reactor
      - jay_kyburz 2y ago
        I wonder how that cost would compare to putting your solar and data center in space? (with non stop solar pointed at the sun) I wonder if cooling is easier or harder in space?
        bostik 2y ago
incognition 2y ago
People tend to overestimate what they can do in a day and underestimate what they can do in a few thousand days
mu53 2y ago
I am convinced that these many of these proclamations and scary AI videos are more of a marketing gimic for people to get excited and fund more AI. AI is prohibitively expensive. LLMs can take millions to train, and for Chatgpt 4, I wouldn't be surprised if the figure was in .5b to 2b range for compute resources alone. ChatGPT struggles for profitability due to high ongoing requirements for GPUs. LLMs were a huge breakthrough. Gaining more funding and making experimentation cheaper will make the next breakthrough come sooner. Profitability will also come sooner. We just don't know the timelines we are considering about changing.
Apocryphon 2y ago
"me when i need $6.5B" https://x.com/netcapgirl/status/1838265672096243822 https://x.com/netcapgirl/status/1838265672096243822
- listic 2y ago
  Where is $6.5B from?
  - Apocryphon 2y ago
    Whomever buys into it https://www.bloomberg.com/news/articles/2024-09-11/openai-fundraising-set-to-vault-startup-s-value-to-150-billion https://www.bloomberg.com/news/articles/2024-09-11/openai-fu...
- kpennell 2y ago
  this person has amazing shitposts...good account
- tim333 2y ago
  $6.5B isn't cool. You know what's cool - $7tn. https://www.theguardian.com/commentisfree/2024/feb/17/openai-boss-sam-altman-wants-7tn-for-all-our-sakes-pray-he-doesnt-get-it https://www.theguardian.com/commentisfree/2024/feb/17/openai...

7thaccount 2y ago

Probably because we all know "a decade" means maybe never. They've said the same thing about fusion for 5 decades.

arthurcolle 2y ago
https://www.energy.gov/articles/doe-national-laboratory-makes-history-achieving-fusion-ignition https://www.energy.gov/articles/doe-national-laboratory-make...
- simianparrot 2y ago
  And it's still a decade away
- Yizahi 2y ago
  Inertial confinement most likely can't be reasonably commoditized into generating electricity. Also, I may be wrong, but I think on the wave of that hype I read that the number they used was an incomplete one. Basically they reframed the conditions of the problem to get more favorable number for advertisement. NIF research will most likely be limited to the nuclear bomb research and simulations. But even if NIF had a reasonable path to energy generation in say 20-30 years, it still won't matter much, just like ITER probably won't. Solar will be way more cheap and widespread, it's costs are still dropping down ahead of all predictions, govt. or commercial. Fusion may just be a pure science project in the end, for a long time.
dcchambers 2y ago
Perhaps ironically, Sam Altman is also big into Fusion. (https://www.helionenergy.com/ https://www.helionenergy.com/) I guess it makes sense, for deep-learning/LLMs to "scale to infinity" you basically need infinite amounts of power.
- bamboozled 2y ago
  cant we use OA intelligence to figure out how to do more with less rather yhan "stupidly" accelerate the climate crisis? the climate crisis + AI is the nightmare
  - 7thaccount 2y ago
    I don't think people on here who downvoted you understand the sheer scale of data center load growth. Utilities can handle some load growth sure, but the requests from the AI folks are crazy large to the tune of GW. I think it's fair to ask if training the next chat bot and related needs are worth doubling our electricity consumption.
  - fragmede 2y ago
    Jevons Paradox says that's not the direction it'll go in.
    - bamboozled 2y ago
      It's funny how some people seem to have different beliefs around this. Personally, I love making things more efficient, and still try to minimize consumption.
- teaearlgraycold 2y ago
  And perhaps AI can help some of the problems in building a fusion reactor.
  - fragmede 2y ago
    Which is pretty exciting https://inl.gov/news/going-digital-idaho-students-engineers-demonstrate-first-nuclear-reactor-digital-twin/ https://inl.gov/news/going-digital-idaho-students-engineers-...

ocean2 2y ago

The Stellarator design has proven to be stable and to produce net positive energy. We are actually only $20B away from having a fully functional nuclear fusion reactor.

samatman 2y ago
Not really, unfortunately. The progress has been heartening, don't get me wrong, but it's more like $20B away from getting to the first steps of solving the next set of problems on the road to a fully functional fusion generator. Those are fairly substantial. It's past time to stop being cynical about "ten more years every ten years", but it's also way too soon to declare victory.
- mjamesaustin 2y ago
  And ultimately, fusion doesn't solve the hardest problem with energy today which is cost. Solar already provides unlimited clean energy, and it took decades of development for the cost to drop to a competitive rate. Even if we had a functioning fusion plant tomorrow, it will likely cost WAY more per kWh and require decades of iteration and improvement to become economically feasible.
  - whamlastxmas 2y ago
    I would wonder if solar plus the absolute massive battery reserves you’d need for a data center ten times bigger than the current worlds largest would still be cost effective versus a single fusion reactor
    - jay_kyburz 2y ago
      I wonder how that cost would compare to putting your solar and data center in space? (with non stop solar pointed at the sun) I wonder if cooling is easier or harder in space?
      - bostik 2y ago

incognition 2y ago

People tend to overestimate what they can do in a day and underestimate what they can do in a few thousand days

mu53 2y ago

I am convinced that these many of these proclamations and scary AI videos are more of a marketing gimic for people to get excited and fund more AI. AI is prohibitively expensive. LLMs can take millions to train, and for Chatgpt 4, I wouldn't be surprised if the figure was in .5b to 2b range for compute resources alone. ChatGPT struggles for profitability due to high ongoing requirements for GPUs. LLMs were a huge breakthrough. Gaining more funding and making experimentation cheaper will make the next breakthrough come sooner. Profitability will also come sooner. We just don't know the timelines we are considering about changing.

Apocryphon 2y ago

"me when i need $6.5B" https://x.com/netcapgirl/status/1838265672096243822 https://x.com/netcapgirl/status/1838265672096243822

listic 2y ago
Where is $6.5B from?
- Apocryphon 2y ago
  Whomever buys into it https://www.bloomberg.com/news/articles/2024-09-11/openai-fundraising-set-to-vault-startup-s-value-to-150-billion https://www.bloomberg.com/news/articles/2024-09-11/openai-fu...
kpennell 2y ago
this person has amazing shitposts...good account
tim333 2y ago
$6.5B isn't cool. You know what's cool - $7tn. https://www.theguardian.com/commentisfree/2024/feb/17/openai-boss-sam-altman-wants-7tn-for-all-our-sakes-pray-he-doesnt-get-it https://www.theguardian.com/commentisfree/2024/feb/17/openai...