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

GPT-5

openai.com

Post reply on HN
Print view View original

541–550 of 1001 posts

Re: GPT-5

#541

Post by jjk166 »

highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

Looks like a lot of players getting closer and closer to an asymptotic limit. Initially small changes lead to big improvements causing a firm to race ahead, as they go forward performance gains from innovation become both more marginal and harder to find, nonetheless keep. I would expect them all to eventually reach the same point where they are squeezing the most possible out of an AI under the current paradigm, barring a paradigm shifting discovery before that asymptote is reached.

Re: GPT-5

#542

Post by rvnx »

jama211 wrote: post #478
highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

Well said. It’s clearly plateauing. It could be a localised plateau, or something more fundamental. Time will tell.

It's a very long presentation just to say that GPT-5 is slightly improved compared to GPT-4o

Re: GPT-5

#543

Post by torginus »

j_timberlake wrote: post #452
highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

I think you're reading way too much into OpenAI bungling its 15-month product lead, but also the whole "1 AGI company will take off" prediction is bad anyway, because it assumes governments would just let that happen. Which they wouldn't, unless the company is really really sneaky or superintelligence happens in the blink of an eye.

I think OpenAI has committed hard onto the 'product company' path, and will have a tough time going back to interesting science experiments that may and may not work, but are necessary for progress.

Re: GPT-5

#544

Post by IceDane »

henriquegodoy wrote: post #432

That SWE-bench chart with the mismatched bars (52.8% somehow appearing larger than 69.1%) was emblematic of the entire presentation - rushed and underwhelming. It's the kind of error that would get flagged in any internal review, yet here it is in a billion-dollar product launch. Combined with the Bernoulli effect demo confidently explaining how airplane wings work incorrectly (the equal transit time fallacy that NAS…

The whole presentation was full of completely broken bar charts. Not even just the typical "let's show 10% of the y axis so that a 5% increase looks like 5x" but stuff like the deception eval showing gpt5 vs o3 as 50 vs 47, but the 47 is 3x as big, and then right next to it we have 9 vs 87, more reasonably sized.

It's like no one looked at the charts, ever, and they just came straight from.. gpt2? I don't think even gpt3 would have fucked that up.

I don't know any of those people, but everyone that has been with OAI for longer than 2 years 1.5m bonuses, and somehow they can't deliver a bar chart with sensible at axes?

Re: GPT-5

#545

Post by procaryote »

makin wrote: post #344
highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

Companies are collections of people, and these companies keep losing key developers to the others, I think this is why the clusters happen. OpenAI is now resorting to giving million dollar bonuses to every employee just to try to keep them long term.

So they're struggling to solve the alignment problem even for their employees?

Re: GPT-5

#546

Post by XCSme »

m4nu3l wrote: post #477

Very funny. The very first answer it gave to illustrate its "Expert knowledge" is quite common, and it's wrong. What's even funnier is that you can find why on Wikipedia: https://en.wikipedia.org/wiki/Lift_(force)#False_explanation... What's terminally funny is that in the visualisation app, it used a symmetric wing, which of course wouldn't generate lift according to its own explanation (as the travelled distance an…

AGI

Re: GPT-5

#547

Post by mizzao »

highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

I recently wrote a little post about this exact idea: https://parsnip.substack.com/p/models-arent-moats

Re: GPT-5

#548

Post by petralithic »

highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

It's the classic S-curve. A few years ago when we saw ChatGPT come out, we got started on the ramping up part of the curve but now we're on the slowing down part. That's just how technology goes in general.

Re: GPT-5

#549

Post by mikepurvis »

beeflet wrote: post #361
highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

Perhaps it is not possible to simulate higher-level intelligence using a stochastic model for predicting text. I am not an AI researcher, but I have friends who do work in the field, and they are not worried about LLM-based AGI because of the diminishing returns on results vs amount of training data required. Maybe this is the bottleneck. Human intelligence is markedly different from LLMs: it requires far fewer examp…

"LLMs tend to regurgitate solutions to solved problems"

People say this, but honestly, it's not really my experience— I've given ChatGPT (and Copilot) genuinely novel coding challenges and they do a very decent job at synthesizing a new thought based on relating it to disparate source examples. Really not that dissimilar to how a human thinks about these things.

Re: GPT-5

#550

Post by JonChesterfield »

kgeist wrote: post #200

Just a week ago I added Qwen3-Coder (the 30b one) to our corporate LLM server, enabled Artifacts in LibreChat, and demoed creating a snake clone in zero shot to coworkers. And now seeing the same exact thing from GPT5's live presentation :) It even has the identical layout.

If you go looking you'll probably find the original on github with GPL written on it, without the llm injected value added breakages.
Post reply on HN