5 ms·

Far from an expert on this topic, but what differentiates AI from other non physical efficiency tools? (I'm actually asking not contesting). Won't companies al

Far from an expert on this topic, but what differentiates AI from other non physical efficiency tools? (I'm actually asking not contesting). Won't companies always want to compete with one another, so simply using AI won't be enough. We will always want better and better software, more features, etc. so that race will never end until we get an AI fully capable of managing all parts (100%) of the development process (which we don't seem to be close to yet). From Excel to Autocad there's been a lot of tools that were expected to decrease the amount of work ended up actually increasing it due to having new capabilities and the constant demand for innovation. I suppose the difference would be if we think AI will continue to get really good, or if it'll become SO good that it is plug and play and completely replaces people.

xpe 1y ago
Companies don’t always compete on capability or quality. Sometimes they compete on efficiency. Or sometimes they carve up the market in different ways.
- OneMorePerson 1y ago
  Sometimes, but with technology related companies I rarely see that. I've really only seen it in industries that are very straightforward, like producing building materials or something. Do you have any examples?
  - xpe 1y ago
    Utilities. Low cost retail. Fast food. Amazon. Walmart. Efficiency is arguably their key competitive advantage. This matters regarding AI systems because a lot of customers may not want to pay extra for the best models! For a lot of companies, serving a good enough model efficiently is a competitive advantage.
xpe 1y ago
> what differentiates AI from other non physical efficiency tools? At some point: (1) general intelligence; i.e. adaptivity; (2) self replication; (3) self improvement.
- OneMorePerson 1y ago
  Yeah I agree, it's not about where it's at now, but whether where we are now leads to something with general intelligence and self improvement ability. I don't quite see that happening with the curve it's on, but again what the heck do I know.
  - xpe 1y ago
    What do you mean about the curve not leading to general intelligence? Even if transformer architectures by themselves don’t get there, there are multifarious other techniques, including hybrids. As long as (1) there are incentives for controlling ever increasing intelligence; (2) the laws of physics don’t block us; and (3) enough people/orgs have the motivation and means, some people/orgs are going to press forward. This just becomes a matter of time and probability. In general, I do not bet against human ingenuity, but I often bet against human wisdom. In my view, along with many others, it would be smarter for the whole world to slow down AI capabilities advancement until we could have very high certainty that doing so is worth the risk.
- amanaplanacanal 1y ago
  We don't have any more idea how to get to 1, 2, or 3, than we did 50 years ago. LLMs are cool, but they seem unlikely to do any of those things.
  - xpe 1y ago
    I encourage everyone to not claim “X seems unlikely” when it comes to high impact risks. Such a thinking pattern often leads to pruning one’s decision tree way too soon. To do well, we need to plan over an uncertain future that has many weird and unfamiliar scenarios.
    - layer8 1y ago
      We already fail to plan for a lot of high-impact things that are exceedingly likely. Maybe we should tackle those first.
marstall 1y ago
every software company i've ever worked with has an endless backlog of features it wants/needs to implement. Maybe AI just lets them move through these feature more quickly? I mean most startups fail. And in software startups, the blame for that is usually at least shared by "software wasn't good enough". So that $20million seed investment is still going to go into "software development" - ie programmer salaries. they will be using the higher level language of ai much of the time, and be 2-5 times more efficient - but will it be enough? No. Most will still fail.

xpe 1y ago

Companies don’t always compete on capability or quality. Sometimes they compete on efficiency. Or sometimes they carve up the market in different ways.

OneMorePerson 1y ago
Sometimes, but with technology related companies I rarely see that. I've really only seen it in industries that are very straightforward, like producing building materials or something. Do you have any examples?
- xpe 1y ago
  Utilities. Low cost retail. Fast food. Amazon. Walmart. Efficiency is arguably their key competitive advantage. This matters regarding AI systems because a lot of customers may not want to pay extra for the best models! For a lot of companies, serving a good enough model efficiently is a competitive advantage.

xpe 1y ago

> what differentiates AI from other non physical efficiency tools? At some point: (1) general intelligence; i.e. adaptivity; (2) self replication; (3) self improvement.

OneMorePerson 1y ago
Yeah I agree, it's not about where it's at now, but whether where we are now leads to something with general intelligence and self improvement ability. I don't quite see that happening with the curve it's on, but again what the heck do I know.
- xpe 1y ago
  What do you mean about the curve not leading to general intelligence? Even if transformer architectures by themselves don’t get there, there are multifarious other techniques, including hybrids. As long as (1) there are incentives for controlling ever increasing intelligence; (2) the laws of physics don’t block us; and (3) enough people/orgs have the motivation and means, some people/orgs are going to press forward. This just becomes a matter of time and probability. In general, I do not bet against human ingenuity, but I often bet against human wisdom. In my view, along with many others, it would be smarter for the whole world to slow down AI capabilities advancement until we could have very high certainty that doing so is worth the risk.
amanaplanacanal 1y ago
We don't have any more idea how to get to 1, 2, or 3, than we did 50 years ago. LLMs are cool, but they seem unlikely to do any of those things.
- xpe 1y ago
  I encourage everyone to not claim “X seems unlikely” when it comes to high impact risks. Such a thinking pattern often leads to pruning one’s decision tree way too soon. To do well, we need to plan over an uncertain future that has many weird and unfamiliar scenarios.
  - layer8 1y ago
    We already fail to plan for a lot of high-impact things that are exceedingly likely. Maybe we should tackle those first.

marstall 1y ago

every software company i've ever worked with has an endless backlog of features it wants/needs to implement. Maybe AI just lets them move through these feature more quickly? I mean most startups fail. And in software startups, the blame for that is usually at least shared by "software wasn't good enough". So that $20million seed investment is still going to go into "software development" - ie programmer salaries. they will be using the higher level language of ai much of the time, and be 2-5 times more efficient - but will it be enough? No. Most will still fail.