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

ChatGPT loses users for first time, shaking faith in AI revolution

washingtonpost.com

Post reply on HN
Print view View original

161–170 of 305 posts

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#161

Post by gAI »

copperx wrote: post #143
CydeWeys wrote: post #126

Earlier quoted context omitted.

I see people using ChatGPT, Bard, etc. to answer questions on mailing list threads. It'll be a long thread of thoughtful human written responses, and then someone will come along and say "I asked Bard what it thought about this issue, here's the response". And then out comes five paragraphs of text that I might as well not even bother reading, as it often contains falsehoods that I have to waste time looking up and d…

I consider code to be non-fiction, and ChatGPT will generate stuff that compiles and works most of the time. There's no need to triple-check the code output.

With the Code Interpreter beta active, it will compile and run that code in the same chat window.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#162

Post by inciampati »

Der_Einzige wrote: post #28

The main issue is that few know about retrieval augmented generation and it’s variants and even fewer know how to build effective versions of it. Hooking up a real high quality and well configured knowledge graph or other information mechanism retrieval alongside template/constraint tooling to powerful, long context (with full attention, not the BS linear kinds) models is an absolute game changer and minimizes risks…

I think you're right. What are the best systems for doing this today?

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#163

Post by IanCal »

shrimpx wrote: post #91

As others suggested, this is likely a summer vacation dip, but speaking for myself as a non-student, my GPT (4 + 3.5) usage has plummeted. If you use it in a "what should be done" or "what is the right thing" sense, you get useless generalities. "It is important to remember that ... ". Suppose for example that you talk to it about combating judicial bias. It will largely insist that judges are unbiased because they'r…

> Suppose for example that you talk to it about combating judicial bias.

Ok, here you go.

https://chat.openai.com/share/b083b98d-6904-4c85-94fc-abdc76...

A quick discussion of where things are "now", some proposals, a selection of three and then ethical concerns about implementing the proposals. I propose a method of increasing diversity and have it critique it.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#165

Post by ragequitta »

tempestn wrote: post #144
CydeWeys wrote: post #126

Earlier quoted context omitted.

I see people using ChatGPT, Bard, etc. to answer questions on mailing list threads. It'll be a long thread of thoughtful human written responses, and then someone will come along and say "I asked Bard what it thought about this issue, here's the response". And then out comes five paragraphs of text that I might as well not even bother reading, as it often contains falsehoods that I have to waste time looking up and d…

The point is you need to collaborate with the LLM, not just expect it to provide fully-formed answers. I've saved countless hours of work over the past few months using GPT-4 to help me solve problems. One general framework can be to use it to help you google a solution instead of a problem. Googling a problem gives you all kinds of results, which you then need to dig through and understand in order to evaluate wheth…

It really is astonishing how much you can get done this way. I've been setting up a home lab for myself, and the answers Gpt4 gives are miles ahead of the stack overflow results or documentation of the apps or whatever else. Rarely (very rarely) it will give me a wrong answer, but then I paste in the error message or describe the problem I had and it almost always comes up with the correct answer 2nd try. Final step is asking where I might learn more because it's not working, and gpt always gives me a better link than google.

I'm convinced the people who say it's nothing but a BS machine have never tried to use it step by step for a project. Or they tried to use it for a project most humans couldn't do, and got upset when it was only 95% perfect.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#166

Post by booleandilemma »

tomcam wrote: post #128
choppaface wrote: post #86

Earlier quoted context omitted.

While Francois Chollet is not actually an employee of OpenAI, and while that article simply quotes a tweet he made rather than a substantive interview .. .. this is still clearly a big problem for OpenAI as students trying to cheat on exams is a highly profitable $1T industry that Microsoft badly needs to disrupt.

> $1T industry Does that mean $1 trillion? If so, citation badly needed.

Surely that's sarcasm. Students don't have money to pay for anything.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#167

Post by JohnFen »

CydeWeys wrote: post #126
stubybubs wrote: post #114

Earlier quoted context omitted.

Given that ChatGPT was just released last fall and it was expected to revolutionize practically everything, isn't it still telling that usage is falling? Where are the scientists, lawyers, doctors, office workers etc? You'd expect a slowing in the rate of growth, not shrinking. I don't know, if high school and undergrad students being off for the summer is enough to shrink your app maybe it's not all that? Edit: I un…

I see people using ChatGPT, Bard, etc. to answer questions on mailing list threads. It'll be a long thread of thoughtful human written responses, and then someone will come along and say "I asked Bard what it thought about this issue, here's the response". And then out comes five paragraphs of text that I might as well not even bother reading, as it often contains falsehoods that I have to waste time looking up and d…

> then someone will come along and say "I asked Bard what it thought about this issue, here's the response"

At least they tell you where the text came from, so you know to skip it. It's worse when they just post an LLM response as their own.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#168

Post by frognumber »

generalizations wrote: post #98
horsawlarway wrote: post #95

Earlier quoted context omitted.

I went self hosted. It was about time to build a new desktop anyways (roughly 4 to 6 years before the old one goes to frolic at the server farm in the basement) and $2,000 will easily buy a machine that can run the quantized 65b models right now. So I spent slightly more than I normally do on this latest box and it's happily spitting out 10+ tokens a second. You're not going to beat GPT-4 yet, but you have direct con…

How does it seem to compare with GPT 3.5? That really seems like the baseline for what's usable.

My experience with open source models places them as a little bit worse than GPT 3, and nowhere close to GPT3.5.

That said:

- For many uses, it doesn't matter. For many of the ways I use it, I don't care. For basic use (e.g. clean up an email for me), it's basically the same. For things like complex reasoning, algorithms, or foreign languages, the hosted service is critical.

- GPT3-grade models have more soul. OpenAI trained GPT3.5 and 4 to never do anything offensive, and that has a lot of negative side effects, well-documented in research. The way I'd describe it, though, is the difference between talking to a call center rep and your grandma (with mild Alzheimer's, perhaps). They both have their place.

- Different models are often helpful in workflows.

My experience is anecdotal. Please don't take it as more than one data point. If other people post their anecdotal experiences, you'll get the plural of "anecdote."

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#169

Post by gAI »

horsawlarway wrote: post #157
root_axis wrote: post #109

Earlier quoted context omitted.

If you're a diligent tester, none of the open source models can touch GPT 3.5 yet, however, in practical terms, some of the 60b and 30b parameter models are almost indistinguishable from GPT 3.5 from the layperson's perspective. Now, if you consider the uncensored models, then you actually have some capabilities that GPT 3.5 and 4 are completely lacking. Based on the rate of progress in the open source world, it won'…

Yup - I'd say this feels about right. The commercial & api based models are still more capable general purpose tools. But the current open tooling can do some nifty stuff, and the community around it is moving at a breakneck speed still. In some areas, it's acceptably good. In some areas it's not. But it's getting better really fast.

That's why my current plan is to get ChatGPT 4 to help me set up my local open source implementations of Orca and Stable Diffusion. Got MusicGen running locally anyway; that was pretty easy.

Re: ChatGPT loses users for first time, shaking faith in AI revolution

#170

Post by brk »

copperx wrote: post #143
CydeWeys wrote: post #126

Earlier quoted context omitted.

I see people using ChatGPT, Bard, etc. to answer questions on mailing list threads. It'll be a long thread of thoughtful human written responses, and then someone will come along and say "I asked Bard what it thought about this issue, here's the response". And then out comes five paragraphs of text that I might as well not even bother reading, as it often contains falsehoods that I have to waste time looking up and d…

I consider code to be non-fiction, and ChatGPT will generate stuff that compiles and works most of the time. There's no need to triple-check the code output.

But this is not really an accurate comparison.

"Code" is really a much much much smaller and much much much more structured output than "English words".

Presumably, the system was trained with a very small amount of "untrue code" in the sense of stuff that just absolutely could never work. And also presumably, it was trained with a lot of free-form text that was definitely wrong or false, and highly likely to have been originally created to be purposefully misleading, or at a minimum, fiction.

That the system outputs reliable code tells us nothing about its current ability to output highly reliable free form text.

Post reply on HN