5 ms·

Same experience, talked to a young, highly educated person about ChatGPT… they had no clue what I was going on about. But the more surprising thing is that eve

by demarq 3y ago

Same experience, talked to a young, highly educated person about ChatGPT… they had no clue what I was going on about. But the more surprising thing is that even after I explained what it could do, they weren’t even slightly impressed. “So it tells you wrong answers? Sounds utterly pointless” I was flabbergasted. I think we do live in a bubble. I’m sure architects and surgeons have equally exciting advances in their fields that no one else cares about. I also think the medias sensationalist phrases should be taken with a pinch of salt: “The tech everyone is talking about”, “here’s the news that got everyone buzzing. etc I do think LLM are a significant event, but that will only be realised by building on top of it and “showing rather than telling”.

kieckerjan 3y ago
Same here. My girlfriend is not impressed by GPT and its ilk at all. But I guess that for a non-technical person most things digital are a form of magic, and it is probably hard to tell the difference between the common magic of (say) a game on her phone and the heavy voodoo of an LLM.
- parineum 3y ago
  I think the better perspective is that LLMs give that magic feel to technologists that aren't really used to it. It certainly did for me until I learned much more about them and used them a bunch. Now it just feels like a tool I can ask to help me build a docker container so I don't have to read docs.
azangru 3y ago
> “So it tells you wrong answers? Sounds utterly pointless”. I was flabbergasted. Could you tell a bit more about why this objection surprised you so much? I am often seeing this in the same tech circles that rejected the web3 craze; and I have to confess, it does sound reasonable to me. There is a recent PR opened in the MDN docs repository after MDN added the "AI explain" button, pointing out how utterly useless a AI guide is if it gives you incorrect answers and you do not have enough knowledge to catch it [0]. [0] - https://github.com/mdn/yari/issues/9208 https://github.com/mdn/yari/issues/9208
- mannykannot 3y ago
  There is nothing wrong about the specific objection but, to borrow an analogy from Scott Aaronson, it is rather like dismissing the Wrights’ early airplanes as being slower than a train and only carrying one passenger. What is missing is an appreciation for how intractable the problem seemed five years before.
  - rsynnott 3y ago
    Bear in mind that for every Wright Brother, there’s a Ferdinand Von Zeppelin. Most heavily hyped technologies don’t have a huge long-term impact.
  - ajkjk 3y ago
    It's not really like that? A small plane with one passenger is a concept that you can extrapolate to a bigger, faster plane. Afaik an often-wrong generative AI is _not_ a concept that extrapolates to a never-wrong generative AI: that's just not how it works at a fundamental level. Although presumably very smart folks are working on it.
    - mannykannot 3y ago
      I see that the analogy is having the unintended side-effect of apparently being predicated on a supposed utility of ChatGPT’s direct descendants. The point I want to get across is how difficult anything like it seemed before, regardless of whether it begets anything markedly better afterwards.
    - thaw13579 3y ago
      For me, it’s already a jetliner in the context of coding assist. It’s correct more often than top hits provided by a top search engine (and any coworker), and it is a very enjoyable user experience (no ads and SEO garbage to filter out). I’d say the wright brothers version was something like BERT or earlier GPTs.
  - mikhmha 3y ago
    Can we stop making analogies like this. It’s so bizarre. The circumstances around each invention are completely different. To always reference the conditions of some past technology is to be stuck in this bizarre form of thinking. A few days ago some poster tried to compare denying LLM’s to being a naysayer in Gallileos time. Is this all we can do? Make references to past events and fail to evaluate the present properly?
rsynnott 3y ago
I mean, I’d be inclined to agree with your friend here, to an extent. A machine for confidently being wrong has rather limited practical applications.
- welshwelsh 3y ago
  I was just about to comment something similar to this: "Google CEO Sundar Pichai said in a 60 minutes interview, 'one Google AI program adapted on its own after it was prompted in the language of Bangladesh, which it was not trained to know'. This is obviously wrong, so does that mean Pichai was hallucinating?" At the last second I fact-checked myself, and found that it actually wasn't Pichai who said that. Crazy how close I came to confidently spewing bullshit in a comment about how humans can also confidently spew bullshit. Anyway, my point is- to be on par with humans, LLMs don't need to be right all of the time, only some of the time.
  - parineum 3y ago
    You just described exactly why you're vastly superior to an LLM. You identified a possible knowledge gap and looked for more data to fill it in. I don't think it's impossible for an LLM to do that but they currently don't.
- jes5199 3y ago
  we’re all on HN aren’t we? isn’t it a forum for confidently being wrong?
  - rsynnott 3y ago
    Sure, but I, at least, am here largely for entertainment purposes. Which is one place where LLMs are mildly useful, actually; while ChatGPT isn't very good at it (I think deliberately so) some LLMs can produce very amusing output (see https://www.aiweirdness.com https://www.aiweirdness.com - while occasionally, as now, there's a serious post, it's mostly AI stuff doing silly things).
nerdbert 3y ago
To many people, a computer program that can output in kind of natural-sounding sentences but can't be trusted to give you the right answers is not progress. I am sympathetic to that view. It's an impressive technical feat to mimic human writing and drawing with its level of fidelity, but until it can make the leap to knowing what it's talking about - and there's no clear basis for assurance that the current vector will get us there - real-world use cases are mostly limited to replacing the some of the more boring email jobs.
tester457 3y ago
> they weren’t even slightly impressed Could be because they haven't actually tried it. There are a surprising amount of applications something that tells wrong answers part of the time can do.

by demarq 3y ago

kieckerjan 3y ago

Same here. My girlfriend is not impressed by GPT and its ilk at all. But I guess that for a non-technical person most things digital are a form of magic, and it is probably hard to tell the difference between the common magic of (say) a game on her phone and the heavy voodoo of an LLM.

parineum 3y ago
I think the better perspective is that LLMs give that magic feel to technologists that aren't really used to it. It certainly did for me until I learned much more about them and used them a bunch. Now it just feels like a tool I can ask to help me build a docker container so I don't have to read docs.

azangru 3y ago

> “So it tells you wrong answers? Sounds utterly pointless”. I was flabbergasted. Could you tell a bit more about why this objection surprised you so much? I am often seeing this in the same tech circles that rejected the web3 craze; and I have to confess, it does sound reasonable to me. There is a recent PR opened in the MDN docs repository after MDN added the "AI explain" button, pointing out how utterly useless a AI guide is if it gives you incorrect answers and you do not have enough knowledge to catch it [0]. [0] - https://github.com/mdn/yari/issues/9208 https://github.com/mdn/yari/issues/9208

mannykannot 3y ago
There is nothing wrong about the specific objection but, to borrow an analogy from Scott Aaronson, it is rather like dismissing the Wrights’ early airplanes as being slower than a train and only carrying one passenger. What is missing is an appreciation for how intractable the problem seemed five years before.
- rsynnott 3y ago
  Bear in mind that for every Wright Brother, there’s a Ferdinand Von Zeppelin. Most heavily hyped technologies don’t have a huge long-term impact.
- ajkjk 3y ago
  It's not really like that? A small plane with one passenger is a concept that you can extrapolate to a bigger, faster plane. Afaik an often-wrong generative AI is _not_ a concept that extrapolates to a never-wrong generative AI: that's just not how it works at a fundamental level. Although presumably very smart folks are working on it.
  - mannykannot 3y ago
    I see that the analogy is having the unintended side-effect of apparently being predicated on a supposed utility of ChatGPT’s direct descendants. The point I want to get across is how difficult anything like it seemed before, regardless of whether it begets anything markedly better afterwards.
  - thaw13579 3y ago
    For me, it’s already a jetliner in the context of coding assist. It’s correct more often than top hits provided by a top search engine (and any coworker), and it is a very enjoyable user experience (no ads and SEO garbage to filter out). I’d say the wright brothers version was something like BERT or earlier GPTs.
- mikhmha 3y ago
  Can we stop making analogies like this. It’s so bizarre. The circumstances around each invention are completely different. To always reference the conditions of some past technology is to be stuck in this bizarre form of thinking. A few days ago some poster tried to compare denying LLM’s to being a naysayer in Gallileos time. Is this all we can do? Make references to past events and fail to evaluate the present properly?

rsynnott 3y ago

I mean, I’d be inclined to agree with your friend here, to an extent. A machine for confidently being wrong has rather limited practical applications.

welshwelsh 3y ago
I was just about to comment something similar to this: "Google CEO Sundar Pichai said in a 60 minutes interview, 'one Google AI program adapted on its own after it was prompted in the language of Bangladesh, which it was not trained to know'. This is obviously wrong, so does that mean Pichai was hallucinating?" At the last second I fact-checked myself, and found that it actually wasn't Pichai who said that. Crazy how close I came to confidently spewing bullshit in a comment about how humans can also confidently spew bullshit. Anyway, my point is- to be on par with humans, LLMs don't need to be right all of the time, only some of the time.
- parineum 3y ago
  You just described exactly why you're vastly superior to an LLM. You identified a possible knowledge gap and looked for more data to fill it in. I don't think it's impossible for an LLM to do that but they currently don't.
jes5199 3y ago
we’re all on HN aren’t we? isn’t it a forum for confidently being wrong?
- rsynnott 3y ago
  Sure, but I, at least, am here largely for entertainment purposes. Which is one place where LLMs are mildly useful, actually; while ChatGPT isn't very good at it (I think deliberately so) some LLMs can produce very amusing output (see https://www.aiweirdness.com https://www.aiweirdness.com - while occasionally, as now, there's a serious post, it's mostly AI stuff doing silly things).

nerdbert 3y ago

To many people, a computer program that can output in kind of natural-sounding sentences but can't be trusted to give you the right answers is not progress. I am sympathetic to that view. It's an impressive technical feat to mimic human writing and drawing with its level of fidelity, but until it can make the leap to knowing what it's talking about - and there's no clear basis for assurance that the current vector will get us there - real-world use cases are mostly limited to replacing the some of the more boring email jobs.

tester457 3y ago

> they weren’t even slightly impressed Could be because they haven't actually tried it. There are a surprising amount of applications something that tells wrong answers part of the time can do.