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

GPT-5 is behind schedule

wsj.com

Post reply on HN
Print view View original

961–970 of 1001 posts

Re: GPT-5 is behind schedule

#961

Post by staticman2 »

Zambyte wrote: post #519
dvdbloc wrote: post #504

Earlier quoted context omitted.

If ChatGPT keeps giving you wrong answers wouldn’t this make paying customers leave? Effectively “losing its job”. But I guess you could say it acts more like the person that makes stuff up at work if they don’t know, instead of saying they don’t know.

> But I guess you could say it acts more like the person that makes stuff up at work if they don’t know, instead of saying they don’t know. I have had language models tell me it doesn't know. Usually when using a RAG-based system like Perplexity, but they can say they don't know when prompted properly.

I've seen Perplexity misrepresent search results and also interpret them differently depending on whether GPT4o or Claude Sonnett 3.5 are being used.

Re: GPT-5 is behind schedule

#962

Post by staticman2 »

debesyla wrote: post #535
fzeindl wrote: post #491

Earlier quoted context omitted.

You can divide your approach to asking questions with people (and I do believe this is something people do): 1. You ask someone you can trust for facts and opinions on topics, but you keep in mind that the answer might only be right in 90% of the cases. Also people tend to tell you if the are not sure. 2. For answers you need to rely on you ask people who are legally or professionally responsible if they give you wro…

I'm not sure about your local laws, but at least in Lithuania it's completely legal to give a wrong advice (by accident, of course)... Even a notary specialist would at most get to pay a larger insurance payment for a while, because human errors falls under professional insurance.

You are contradicting yourself. If the notary specialist needs insurance then there's a legal liability they are insuring against.

If you had written "notaries don't even get insurance because giving bad advice is not something you can be sued for" you would be consistent.

Re: GPT-5 is behind schedule

#963

Post by bloppe »

weatherlite wrote: post #786
aprilthird2021 wrote: post #750

Earlier quoted context omitted.

The market of people willing to pay $50 a month for OAI vs $0/month for one of the open source LLAMA variants is not large enough to justify their current valuation, imo

I'm not that familiar with the open source ones - how good are they in comparison?

It doesn't really matter how much people are willing to pay. It matters how much margin the market will allow you to charge. OpenAI may be a bit better than most competitors most of the time (IMO they keep getting leap-frogged by Anthropic et al. though), but if your customers can get 90% of the value for 50% less, they will bail. There is no moat. Margins will be razor thin. That's not a 1B+ company.

Re: GPT-5 is behind schedule

#964

Post by bloppe »

lukan wrote: post #836
bloppe wrote: post #829

Earlier quoted context omitted.

Tesla is profitable and they have a big technological moat. OpenAI is in a very competitive industry and they burn ~5B a year.

I believe the car industry is somewhat competive as well and they needed allmost 10 years to become profitable.

Sure, but if you want to compete with Tesla, you need many billions in funding and 10+ years to catch up. If you want to compete with OpenAI, you need maybe half a billion (easy to raise in the current climate; many have done so) and mb a few months to catch up.

Re: GPT-5 is behind schedule

#965

Post by icpmacdo »

overgard wrote: post #540
icpmacdo wrote: post #358

Earlier quoted context omitted.

Gary Marcus is continuously lambasted and not taken seriously

By whom? He seems highly credible to me, and his credentials check out, especially compared to hype men like Sam Altman. All youre doing is spreading FUD by an unnamed "they"

He only criticizes ai capabilities, without creating anything himself. Credentials are effectively meaningless. With every new release, he clamors for attention to prove how right he was—and always will be. That’s precisely why he lacks credibility.

Re: GPT-5 is behind schedule

#966

Post by karel-3d »

sampullman wrote: post #701
karel-3d wrote: post #693

I have to say I finally "caved in" to LLMs last month. While I still think Copilot is useless, I recently had a very complex code that did a lot of crazy bit-flipping and xoring, and I had no idea what is it doing, so I threw it to ChatGPT.. and it knew what it was doing. I also needed to rewrite this code to PHP (for... reasons) while I know very little PHP. And it did that! It was a bit wrong, I needed to correct a…

That's basically my experience. It's great for learning or getting things done when the subject is related to one you know well (i.e. you understand the fundamentals and can verify responses quickly). It's not so good for a completely new subject, or one you have a lot of experience in.

Exactly. They are good in the sweet spot when you can verify that they are correct, but you are not that good to do it yourself.

Basically StackOverflow but without the annoying mods?

Re: GPT-5 is behind schedule

#967

Post by flufluflufluffy »

It takes actual humans, whose brains have been honed by multiple millions of years of evolution, over ten years for our brains to experience, neuroplasticify, form a well-formed-enough model of existence to drive a car, let alone wax philosophical about abstract concepts. And we funnel dollar after dollar, extract mineral after mineral, consume kilowatt hour after kilowatt hour, wanting, expecting Artificial General Intelligence to be ready within one or two years since the last promised version.

Re: GPT-5 is behind schedule

#968

Post by anothernewdude »

dgfitz wrote: post #284
anothernewdude wrote: post #263

Earlier quoted context omitted.

[flagged]

I wonder how Russian and North Korean citizens would feel about a capitalist, representative democracy?

Of course you're right, there's something worse, therefore capitalist, unrepresentative democracy is perfect.

How could I be so naive?

Re: GPT-5 is behind schedule

#969

Post by dgfitz »

anothernewdude wrote: post #968
dgfitz wrote: post #284

Earlier quoted context omitted.

I wonder how Russian and North Korean citizens would feel about a capitalist, representative democracy?

Of course you're right, there's something worse, therefore capitalist, unrepresentative democracy is perfect. How could I be so naive?

What’s the quote, something like: “democracy and capitalism are horrendous, but they’re better than everything else we tried so far”

Re: GPT-5 is behind schedule

#970

Post by int_19h »

lolinder wrote: post #287
idiotsecant wrote: post #260

Earlier quoted context omitted.

Have you ever heard of a local maxima? You don't get an attack helicopter by breeding stronger and stronger falcons.

For an industry that spun off of a research field that basically revolves around recursive descent in one form or another, there's a pretty silly amount of willful ignorance about the basic principles of how learning and progress happens. The default assumption should be that this is a local maximum, with evidence required to demonstrate that it's not. But the hype artists want us all to take the inevitability of LLM…

So far we haven't even climbed this slope to the top yet. Why don't we start there and see if it's high enough or not first? If it's not, at the very least we can see what's on the other side, and pick the next slope to climb.

Or we can just stay here and do nothing.

Post reply on HN