Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: Is anyone else bearish on OpenAI?

news.ycombinator.com

Post reply on HN
Print view View original

211–220 of 321 posts

Re: Ask HN: Is anyone else bearish on OpenAI?

#211

Post by goatlover »

alchemist1e9 wrote: post #30
theossuary wrote: post #19

Earlier quoted context omitted.

This interesting, I've had the opposite experience so far. I tried to do that because if it worked it'd be game changing; but it seems every question I ask it, I get half truths and lies back. Its impossible to figure out what is and isn't true either, because the LLM says everything with such a certainty. I end up having to research everything it says again to figure out what to trust. Sometimes it's good to get a f…

You have such an experience with technical problems? It’s written at least 20 python scripts with me and almost all of them have been close to perfect almost on first draft.

Really? That hasn't been my experience. Out of 20 python scripts, at least 50% have some error, some of which look right until you run them. A few involve hallucinations, like calling non-existent methods.

Re: Ask HN: Is anyone else bearish on OpenAI?

#212

Post by aprdm »

latexr wrote: post #150
aprdm wrote: post #65

Earlier quoted context omitted.

Hm, my wife uses it daily for her classes which aren't tech related. She describes it as a very patient professor who meets her where she is at and doesn't judge. English isn't her first language too and sometimes she can ask what a term means in English.

> English isn't her first language too and sometimes she can ask what a term means in English. That’s a good example of a subpar use for an LLM. Dictionaries have existed since before computers. Digital dictionaries are faster, more reliable, and less power consuming than any LLM.

Not really.. I think you don't know what you're talking about :)

It's a chat, she is talking about a subject, and then she can expand on a word, all very naturally, as she is talking with a human.

That's very different than having some subject you're reading about and then having to carry a dictionary which has no context about the subject, and doesn't talk back to you

Re: Ask HN: Is anyone else bearish on OpenAI?

#213

Post by lucideer »

bmitc wrote: post #36
ghshephard wrote: post #9

OpenAI, at least in my day-day workflow for the last 9+ months has so superseded anything that google ever was to me that I'm having a difficult time comparing the two. I've got a monitor dedicated 100% of the time to ChatGPT, and I interact with it non stop during the flow of technical scenarios and troubleshooting situations that flow into me - working in areas that I have the slimmest of backgrounds in, and shutti…

Are you able to give a specific example of a problem it helped you solve? Especially one that you were at a complete blocking point, it provided some solution, and then you were able to continue to expand upon that solution. I keep reading responses like yours, but I haven't seen any specific examples of problems being solved, so it all sounds very abstract. In my interactions with ChatGPT, it felt like just interact…

> Are you able to give a specific example of a problem it helped you solve? Especially one that you were at a complete blocking point...

I think it's worth stepping back here and re-examining the hurdles your setting for your own understanding.

The essential question here is: is ChatGPT useful to people. What you seem to be implying with your question is: I will not use ChatGPT unless it can solve problems with X level of difficulty for me. Why have you set that pre-requisite? Would it not still be useful to you if it simply increased the efficiency of solving simple day-to-day tasks that you're not blocked on?

Re: Ask HN: Is anyone else bearish on OpenAI?

#214

Post by aprdm »

smrq wrote: post #177
Rastonbury wrote: post #138

Earlier quoted context omitted.

As a non-professional coder, I'm honestly surprised there seems quite a few professional coders in this thread who don't do the first Stackoverflow first scenario. As someone learning code, shoehorning SO code when I don't understand 20% of the replies made the learning curve so steep, chatGPT has made this so much easier by changing it and telling me why will/won't work, even when watching out for or correcting hall…

As I've matured in my career I've gone to SO less and less, to the point where I mostly avoid it now. I use tools, languages, and libraries with which I am now highly familiar--perhaps more so than the average SO answerer, although I guess it would be arrogant to say for sure--and so the bulk of my questions are edge cases in my knowledge which are more easily answered by reading source code or specs. I suspect this…

Probably because you're deep specialized in something or have stagnated on a company for a long time ? I cannot imagine knowing everything about what I am working on, that has changed dramatically every 3-4 months even within the same company

Re: Ask HN: Is anyone else bearish on OpenAI?

#215

Post by aprdm »

OJFord wrote: post #205
yunwal wrote: post #189

Earlier quoted context omitted.

You should be able to verify facts outside of your domain of expertise. Every time a new study comes around saying “actually, everyone should be drinking 3 glasses of wine a night”, do you take a trip to the liquor store? LLMs are far from the only thing you rely on that will confidently lie to you.

Some things are inherently not verifiable, but actually what I meant was something like 'verifiable easily by me in a reasonable amount of time that makes the whole exercise worthwhile '. You can make some medical claim for example and describe your (or whoever's) methodology sufficiently for me to reproduce the result in theory , but I'm not going to. > Every time a new study comes around saying “actually, everyone…

bing chat actually lists the sources

Re: Ask HN: Is anyone else bearish on OpenAI?

#216

Post by lucubratory »

somewhereoutth wrote: post #178
lucubratory wrote: post #112

>It's also very much like crypto where for every one person doing something useful with it, there are 20 trying to exploit the newness and low comprehension the general public have of the tech This is definitely not correct in terms of numbers, there are many more people using LLMs well than have ever used crypto for any real use case. Also it's worth considering that the only real use cases of crypto are illegal, fr…

> It is nevertheless true that in aggregate for knowledge work, "worker + intern" is more economically productive than just "worker" It can be true, but I don't think it is always or necessarily often true. The overall value proposition for interns (and apprentices etc) for society as a whole is that they will then go on to become the professional knowledge workers that they have learnt from. LLMs won't become that,…

It is economically true in aggregate, although you're right that the worker overseeing the intern very often feels like it isn't true. The mechanic you mentioned could fall into that category, but I think it's more likely that it's just one of those jobs that's not knowledge work so what I'm saying doesn't really apply, because there are physical space limits to collaboration in realspace work that are much more readily obtrusive than equivalents in knowledge work. Basically, good knowledge worker A & bad knowledge worker B are never going to be "A + B" productive together, but they're also extremely unlikely to be About the interns as hiring component, that's the official reason for interns and they are used that way, but they are also just valuable in and of themselves as labour. Companies that aren't even hiring and won't be in six months will still take on interns, because it's free labour. Interns are the equivalent of manual labourers in knowledge work, able to do what you tell them but not yet possessing the skills to do many complex tasks alone.

And I wouldn't be so sure that the "LLMs as interns" relationship we have now doesn't result in them learning. Obviously on an architecture level any individual instance can only learn in-context and that's wiped when the context is, but in aggregate they are learning from their "time on the job". The next version of GPT is going to be directly better at least partially because of all of the data that ChatGPT 4 being an effective intern brings in. It could be anywhere from just an intern at 6 months rather than an intern on the first week, through to an actual hired employee somewhere during their first year. Even if it was technologically possible I don't think OpenAI would deploy their next product if it was as good as a senior employee in some substantial number of domains, because that would be too disruptive of a leap for society. But anywhere on the intern scale that's better than GPT-4 up to approximately an employee one year in I think is feasible for their next release, and that will be in large part thanks to the data they get from their current deployment.

Re: Ask HN: Is anyone else bearish on OpenAI?

#217

Post by mliker »

bmitc wrote: post #36
ghshephard wrote: post #9

OpenAI, at least in my day-day workflow for the last 9+ months has so superseded anything that google ever was to me that I'm having a difficult time comparing the two. I've got a monitor dedicated 100% of the time to ChatGPT, and I interact with it non stop during the flow of technical scenarios and troubleshooting situations that flow into me - working in areas that I have the slimmest of backgrounds in, and shutti…

Are you able to give a specific example of a problem it helped you solve? Especially one that you were at a complete blocking point, it provided some solution, and then you were able to continue to expand upon that solution. I keep reading responses like yours, but I haven't seen any specific examples of problems being solved, so it all sounds very abstract. In my interactions with ChatGPT, it felt like just interact…

With chatgpt I’ve been able to generate math 2d and 3d animations in just 1-2hr with no background nor prior knowledge.

I was also able to use it to generate software to check the blockchain to gather data I needed, again with no experience using the blockchain client libraries

Re: Ask HN: Is anyone else bearish on OpenAI?

#218

Post by kubiton »

pradn wrote: post #190
kubiton wrote: post #184

LLM are the first interface I as.a.software engineer is impressed by. My company also got it immediately and is rolling it out globally. GitHub copilot is already helpful. GitHub copilot for docs (announced at GitHub next) is a game changer. I used openai to reformulate emails and suddenly got positive feedback about my announcement emails. I communicate with openai German and English how ever I see fit. It's very ha…

> I communicate with openai German and English how ever I see fit. I did not foresee that you could mix input languages. That’s fascinating. Multilingual people use languages for different purposes, often in the same sentence. Ie: technical jargon in English, a joke in Arabic, etc. I expect the difference in connotation/feeling/mood to be less relevant for an LLM, if you’re working with facts. But there was a recent…

No. I primarily switch when can't remember a word.

I also describe things and that also works quite well.

Re: Ask HN: Is anyone else bearish on OpenAI?

#219

Post by kristianp »

janalsncm wrote: post #144

Whether or not you think it’s a fad, ChatGPT has close to 200 million users actively using their product. No crypto company was ever close to that. Whether OpenAI’s API is a viable or risky product is a secondary and separate question. Yeah, there are a lot of wrappers out there. But that doesn’t matter with regards to the usefulness of LLMs generally.

And crypto is often searching for a use case that it can do better than other technologies. Payments for example, have advantages in crypto that the average person doesn't care about. The disadvantages however, can be killers (e.g. refunds for fraudulent transactions on your credit card has no equivalent in crypto).

The other problem with crypto's search for the killer app is that things like NFTs, which make no sense in the world of 'free' information, became ridiculously hyped and gave crypto a bad name.

Re: Ask HN: Is anyone else bearish on OpenAI?

#220

Post by latexr »

RecycledEle wrote: post #209
latexr wrote: post #151

Earlier quoted context omitted.

> OpenAI has changed education. No, it objectively has not. Maybe it has changed how you teach, but “education” is much larger than any of us. Until school curriculums around the world incorporate ChatGPT—quite the dystopian scenario—they have not changed education.

Obsolete does not mean something is no longer used. It means that new technology beats the obsolete system every time. I'm spending most of today updating my Canvas courses, and Canvas is obsolete. Students are much better off asking AIs how to do thing and what they should learn next rather than working through my Canvas courses.

You’re arguing against a point I did not make. Seems like you wanted to respond to one of the other replies.
Post reply on HN