3 ms·

AI research essentially started in 1940s, and not the 2010s like you might think. I read a ton of AI research papers written in the 80s and 90s. Lack of AI hy

AI research essentially started in 1940s, and not the 2010s like you might think. I read a ton of AI research papers written in the 80s and 90s. Lack of AI hype is was because the lack of data and compute. The actual field is here for a very long time.

JohnFen 3y ago
I'm well aware -- I've been working with this stuff since the early 90s and have a better than usual understanding of the history of it all. But if we're going to analogize to the internet, you have to count LLMs as "the internet" and the things that came before as just "networking". Networking is very old and laid the basis for the internet, but the internet was something different in kind. My only point is that LLM being used for real things is very recent. There hasn't been a ton of use (relatively speaking) before it was available for public use. That's not true with the internet. The internet was heavily used, and heavily iterated on, before the public had access to it. Being skeptical about LLMs is not irrational, and so you see a lot of it. There's just not a great deal of history with them to provide counterexamples to that skepticism. That's not the case when the internet was opened to the public, which is a large factor in why the skepticism about the internet was much lower than the skepticism about LLMs.
- FrustratedMonky 3y ago
  Tomato Potato. Yes, there was usenet, and gopher, and FTP. Think splitting hairs a bit. As others have pointed out. 'AI', has been actively researched since the 40's. So there was a lot of ground work before the latest shiny thing the 'LLM'. Just as there was a lot of ground work in 'networking' before WWW. I've been talking to GPT4 about a problem all day. I think everyone saying it is 'over-hyped' have really just not used it yet. It will definitely keep growing, it will replace jobs, it is already increasing productivity and changing markets. Sorry - one edit. I strongly disagree that things were 'well proven' with the Internet before the public was aware. If you include WWW and the rapid changes in standards and browsers, and technology. It was all moving as fast and with all the bugs and problems and hacks as needed for fast moving tech. There was a ton not proven, and with no use cases to justify it. It was a wild west. Now we are seeing it again.
  - hnfong 3y ago
    The “internet” was definitely moving fast. And it hasn’t even stopped. At the height of the internet bubble, people dreamt of one day having virtual meetings with VR gear and stuff. And it’s only starting to happen… in 2023. It’s been more than 20 years and the tech is finally ready to do this VR stuff properly. Sometimes hype becomes reality many years later. This is probably going to be true for AI as well.

JohnFen 3y ago

I'm well aware -- I've been working with this stuff since the early 90s and have a better than usual understanding of the history of it all. But if we're going to analogize to the internet, you have to count LLMs as "the internet" and the things that came before as just "networking". Networking is very old and laid the basis for the internet, but the internet was something different in kind. My only point is that LLM being used for real things is very recent. There hasn't been a ton of use (relatively speaking) before it was available for public use. That's not true with the internet. The internet was heavily used, and heavily iterated on, before the public had access to it. Being skeptical about LLMs is not irrational, and so you see a lot of it. There's just not a great deal of history with them to provide counterexamples to that skepticism. That's not the case when the internet was opened to the public, which is a large factor in why the skepticism about the internet was much lower than the skepticism about LLMs.

FrustratedMonky 3y ago
Tomato Potato. Yes, there was usenet, and gopher, and FTP. Think splitting hairs a bit. As others have pointed out. 'AI', has been actively researched since the 40's. So there was a lot of ground work before the latest shiny thing the 'LLM'. Just as there was a lot of ground work in 'networking' before WWW. I've been talking to GPT4 about a problem all day. I think everyone saying it is 'over-hyped' have really just not used it yet. It will definitely keep growing, it will replace jobs, it is already increasing productivity and changing markets. Sorry - one edit. I strongly disagree that things were 'well proven' with the Internet before the public was aware. If you include WWW and the rapid changes in standards and browsers, and technology. It was all moving as fast and with all the bugs and problems and hacks as needed for fast moving tech. There was a ton not proven, and with no use cases to justify it. It was a wild west. Now we are seeing it again.
- hnfong 3y ago
  The “internet” was definitely moving fast. And it hasn’t even stopped. At the height of the internet bubble, people dreamt of one day having virtual meetings with VR gear and stuff. And it’s only starting to happen… in 2023. It’s been more than 20 years and the tech is finally ready to do this VR stuff properly. Sometimes hype becomes reality many years later. This is probably going to be true for AI as well.