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

481–490 of 1001 posts

Re: GPT-5 is behind schedule

#481

Post by synergy20 »

croes wrote: post #479
atonse wrote: post #457

Earlier quoted context omitted.

Same here. The ability to “talk to an expert” about any topic I’m curious about and ask very specific questions has been invaluable to me. It reminds me of being a kid and asking my grandpa a million questions, like how light bulbs worked, or what was inside his radio, or how do we have day and night. And before anyone talks about accuracy or hallucinations, these conversations usually are treated as starting off poi…

How do you know the answers are correct? More than once I got eloquent answer that are completely wrong.

so do teachers and books, in the future we need have multiple variants to cross check

Re: GPT-5 is behind schedule

#482

Post by NitpickLawyer »

mnk47 wrote: post #231
Animats wrote: post #123

"Orion’s problems signaled to some at OpenAI that the more-is-more strategy, which had driven much of its earlier success, was running out of steam." So LLMs finally hit the wall. For a long time, more data, bigger models, and more compute to drive them worked. But that's apparently not enough any more. Now someone has to have a new idea. There's plenty of money available if someone has one. The current level of LLM…

> So LLMs finally hit the wall Not really. Throwing a bunch of unfiltered garbage at the pretraining dataset, throwing in RLHF of questionable quality during post-training, and other current hacks - none of that was expected to last forever. There is so much low-hanging fruit that OpenAI left untouched and I'm sure they're still experimenting with the best pre-training and post-training setups. One thing researchers…

> Not really. Throwing a bunch of unfiltered garbage at the pretraining dataset, throwing in RLHF of questionable quality during post-training, and other current hacks - none of that was expected to last forever. There is so much low-hanging fruit that OpenAI left untouched and I'm sure they're still experimenting with the best pre-training and post-training setups.

Exactly! LLama3 and their .x iterations have shown that, at least for now, the idea of using the previous models to filter out the pre-training datasets and use a small amount of seeds to create synthetic datasets for post-training still holds. We'll see with L4 if it continues to hold.

Re: GPT-5 is behind schedule

#483

Post by brookst »

stingraycharles wrote: post #473
atonse wrote: post #457

Earlier quoted context omitted.

Same here. The ability to “talk to an expert” about any topic I’m curious about and ask very specific questions has been invaluable to me. It reminds me of being a kid and asking my grandpa a million questions, like how light bulbs worked, or what was inside his radio, or how do we have day and night. And before anyone talks about accuracy or hallucinations, these conversations usually are treated as starting off poi…

For me the problem is that you always need to double-check this particular type of expert, as it can be confidently wrong about pretty much any topic. It's useful as a starting point, not as a definitive expert answer.

What human experts do you blindly trust without double checking?

Re: GPT-5 is behind schedule

#484

Post by szundi »

brookst wrote: post #480
croes wrote: post #479

Earlier quoted context omitted.

How do you know the answers are correct? More than once I got eloquent answer that are completely wrong.

How do you address this problem with people? More than once a real live person has told me something that was wrong,

and people just don't know what they don't know - they just answer sillyness the same way

Re: GPT-5 is behind schedule

#485

Post by bloppe »

lumost wrote: post #461
bloppe wrote: post #458

Earlier quoted context omitted.

At this point I think even the most bearish have to concede that LLM's are an amazing tool. But OpenAI was never supposed to be about creating tools. They're supposed to create something that can completely take over entire projects for you, not just something that can help you work on the projects faster. If they can't pull that off in the next year or two, they're gonna seriously struggle to raise the next 10B they…

At this point it’s quite likely that they could pivot and just be the chatgpt company. I’ve found chatgpt-4o with web search and plugins to be more useful than o1 for most tasks. It’s possible we’re nearing the end of the LLM race, but I doubt that’s the end of the AI story this decade, or OpenAI.

Ya I think they probably will, but "the chatgpt company" is not worth 157B. It might not even be worth 1B.

Re: GPT-5 is behind schedule

#486

Post by spaceman_2020 »

SamPatt wrote: post #415

I'm sure the debate over the definition of AGI is important and will continue for a while, but... I can't care about it anymore. Between Perplexity searching and summarizing, Claude explaining, and qwen (and other tools) coding, I'm already as happy as can be with whatever you want to call this level of intelligence. Just today I used a completely local AI research tool, based on Ollama. It worked great. Maybe it won…

This is me. If things never improve and Sonnet 3.6 is the best we have…I’m fine. Its good enough to drastically improve productivity

Re: GPT-5 is behind schedule

#487

Post by NitpickLawyer »

wokwokwok wrote: post #460
divan wrote: post #445

Earlier quoted context omitted.

Let's revisit this comment in one year – after the explosion of agentic systems. (:

We already have agentic systems; they're not particularly impressive (1). There's no specific reason to expect them to get better. Things that will shift the status quo are: MCST-LLMs (like with ARC-AGI) and Much Bigger LLMs (like GPT-5, if they ever turn up) or some completely novel architecture. [1] - It's provable; if just chaining LLMs are a particular size into agentic systems could scale indefinitely, then you…

> It's provable; if just chaining LLMs are a particular size into agentic systems could scale indefinitely, then you could use a 1-param LLM and get AGI. You can't. QED.

Perhaps I missunderstand your reply, but that has not been my experience at all.

There are 3 types of "agentic" behaviour that has worked for a while for me, and I don't know how else it would work without "agents":

1. Task decomposition - this was my manual flow since pre-chatgpt models: a) provide an overview of topic x with chapter names; b) expand on chapter 1 ... n ; c) make a summary of each chapter; d) make an introduction based on the summaries. I now have an "agent" that does that w/ minimal scripting and no "libraries". Just pure python control loop.

This gets me pretty reasonable documents for my daily needs.

2. tool use (search, db queries, API hits). I don't know how you'd use an LLM without this functionality. And chaining them into flows absolutely works.

3. coding. I use the following "flow" -> input a paragraph or 2 about what I want, send that + some embedding-based context from the codebase to an LLM (3.5 or 4o, recently o1 or gemini) -> get code -> run code -> /terminal if error -> paste results -> re-iterate if needed. This flow really works today, especially with 3.5. In my testing it needs somewhere under 3 "iterations" to "get" what's needed in more than 80% of the cases. I intervene in the rest of 20%.

Re: GPT-5 is behind schedule

#488

Post by corimaith »

EGreg wrote: post #344
marcus_holmes wrote: post #333

Earlier quoted context omitted.

The Singularity is caused by AI being able to design better AI. There's probably some AI startup trying to work on this at the moment, but I don't think any of the big boys are working on how to get an LLM to design a better LLM. I still like the analogy of this being a really smart lawn mower, and we're expecting it to suddenly be able to do the laundry because it gets so smart at mowing the lawn. I think LLMs are g…

I think this whole “AGI” thing is so badly defined that we may as well say we already have it. It already passes the Turing test and does well on tons of subjects. What we can start to build now is agents and integrations. Building blocks like panel of experts agents gaming things out, exploring space in a Monte Carlo Tree Search way, and remembering what works. Robots are only constrained by mechanical servos now. W…

>Exploring space in a Monte Carlo Tree Search way, and remembering what works.

The information space of "research" is far larger than the information space of image recognition or language, larger than our universe probably, it's tantamount to formalizing the entire World. Such an act would be akin to touching "God" in some sense of finding the root of knowledge.

In more practical terms, when it comes to formal systems there is a tradeoff between power and expressiveness. Category Theory, Set Theory, etc are strong enough to theoretically capture everything, but are far to abstract to use in practical sense with suspect to our universe. The systems that do we have, aka expert systems or knowledge representation systems like First Order Predicate Logic aren't strong enough to fully capture reality.

Most importantly, the information spac have to be fully defined by researchers here, that's the real meat of research beyond the engineering of specific approaches to explore that space. But in any case, how many people in the world are both capable of and are actually working on such problems? This is highly foundational mathematics and philosophy here, the engineers don't have the tools here.

Re: GPT-5 is behind schedule

#489

Post by superultra »

croes wrote: post #479
atonse wrote: post #457

Earlier quoted context omitted.

Same here. The ability to “talk to an expert” about any topic I’m curious about and ask very specific questions has been invaluable to me. It reminds me of being a kid and asking my grandpa a million questions, like how light bulbs worked, or what was inside his radio, or how do we have day and night. And before anyone talks about accuracy or hallucinations, these conversations usually are treated as starting off poi…

How do you know the answers are correct? More than once I got eloquent answer that are completely wrong.

I give AI a “water cooler chat” level of veracity, which means it’s about as true as chatting with a coworker at a water cooler when that used to happen. Which is to say if I just need to file the information away as a “huh” it’s fine, but if I need to act on it or cite it, I need to do deeper research.

Re: GPT-5 is behind schedule

#490

Post by mrbungie »

IAmGraydon wrote: post #466
mrbungie wrote: post #240

Earlier quoted context omitted.

Yep, get into any sufficiently deep niche (i.e. actually almost any non-trivial app) and the LLM magic fades off. Yeah sure you can make a pong clone in html/js and that's mainly because there the internet is full of pong clone demos. Ask how to constraint a statsmodels lineal model in some non-standard way? It will gaslight how it is possible and make you loss time in the process.

Making a pong clone by telling the LLM to make a pong clone is a cute trick that sometimes works, but that's not the way anyone who understands how to properly use these tools is using them. You don't describe and app and hope the LLM builds it correctly. You have to know how to architect an application and you use the LLM to build small pieces of code. For example, you tell it to build a function that does x, takes…

No true scottsman fallacy. I know how to use them, but using them "correctly" still produces many errors.

They suck at non-trivial code outside of standard library usage and boilerplate coding: I gave an example and parent did as well. In that regard would at least change your phrase from "actual coders" to "actual senior coders", as any junior receiving bad advice (in eternal loops as LLMs normally like to do it) is only going to make them waste time and tokens.

Post reply on HN