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

601–610 of 1001 posts

Re: GPT-5

#601

Post by wyager »

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…

> a stochastic model for predicting text

It's fascinating to me that so many people seem totally unable to separate the training environment from the final product

Re: GPT-5

#602

Post by not_a_bot_4sho »

charlie0 wrote: post #588

Not so sure about the behind the scenes "automatic router". What's to stop OpenAI from slowing gimping GPT-5 over time or during times of high demand? It seems ripe for delivering inconsistent results while not changing the price.

What's to stop them from routing to GPT2? Or to Gemini? Or to a mechanical turk? This path is open to your imagination.

That said, I've had luck with similar routing systems (developed before all of this -- maybe wasted effort now) to optimize requests between reasoning and regular LLMs based on input qualities. It works quiet well for open-domain inputs.

Re: GPT-5

#603

Post by bearjaws »

levocardia wrote: post #366
surround wrote: post #297

GPT-5 knowledge cutoff: Sep 30, 2024 (10 months before release). Compare that to Gemini 2.5 Pro knowledge cutoff: Jan 2025 (3 months before release) Claude Opus 4.1: knowledge cutoff: Mar 2025 (4 months before release) https://platform.openai.com/docs/models/compare https://deepmind.google/models/gemini/pro/ https://docs.anthropic.com/en/docs/about-claude/models/overv...

with web search, is knowledge cutoff really relevant anymore? Or is this more of a comment on how long it took them to do post-training?

Falling back to web search is a crutch, its slower and often bloats context resulting in worse output.

Re: GPT-5

#604

Post by zaphirplane »

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…

Cats and dogs kind of also cluster together with a couple of exceptions relative to humans ;)

Re: GPT-5

#605

Post by agnosticmantis »

Unless the whole presentation was generated using sora-gpt-5 or something, this was very underwhelming.

We know for a fact the slides/charts were generated using an LLM, so the hypothesis is not totally unfounded. /s

Re: GPT-5

#606

Post by dougdonohoe »

hrpnk wrote: post #336

They will retire lots of models: GPT-4o, GPT-4.1, GPT-4.5, GPT-4.1-mini, o4-mini, o4-mini-high, o3, o3-pro. https://help.openai.com/en/articles/6825453-chatgpt-release-... "If you open a conversation that used one of these models, ChatGPT will automatically switch it to the closest GPT-5 equivalent." - 4o, 4.1, 4.5, 4.1-mini, o4-mini, or o4-mini-high => GPT-5 - o3 => GPT-5-Thinking - o3-Pro => GPT-5-Pro

I don't have confidence that systems built on top of a specific model will work the same on a higher version. Unlike, say, the Go programming language where backwards compatibility is something you can generally count on (with exceptions being well documented).

I wouldn't want to be in charge of regression testing an LLM-based enterprise software app when bumping the underlying model.

Re: GPT-5

#607

Post by CharlieDigital »

levocardia wrote: post #366
surround wrote: post #297

GPT-5 knowledge cutoff: Sep 30, 2024 (10 months before release). Compare that to Gemini 2.5 Pro knowledge cutoff: Jan 2025 (3 months before release) Claude Opus 4.1: knowledge cutoff: Mar 2025 (4 months before release) https://platform.openai.com/docs/models/compare https://deepmind.google/models/gemini/pro/ https://docs.anthropic.com/en/docs/about-claude/models/overv...

with web search, is knowledge cutoff really relevant anymore? Or is this more of a comment on how long it took them to do post-training?

Yes, because it may not know that it needs to do a web search for the most relevant information.

Re: GPT-5

#608

Post by warmedcookie »

rvz wrote: post #4

I hope that this live stream will tell you that this will be the definitive reason why web developers, JavaScript / TypeScript developers are going to be made completely obsolete at worse and at best, their jobs will be reduced at all levels. The best part is, this is not even the real definition of "AGI" yet (whatever that means at this point). More like 10% of the capability that was promised and already the flow o…

I would qualify that by saying developers who do not have Product Owner skills and Product Owners who do not have developer skills will be made obsolete.

Having both eliminates a feedback loop and the LLM enables you to get shit done fast.

Re: GPT-5

#610

Post by mustaphah »

punee94 wrote: post #577

I ran the below prompt to both Kimi2 and GPT5. how many rs in cranberry? -- GPT5's response: The word cranberry has two “r”s. One in cran and one in berry. Kimi2's response: There are three letter rs in the word "cranberry".

Stop asking LLMs to count!

Text is broken into tokens in training (subword/multi-word chunks) rather than individual characters; the model doesn’t truly "see" letters or spaces the way humans do. Counting requires exact, step-by-step tracking, but LLMs work probabilistically.

It's not much of a help anyway, don't you agree?

Post reply on HN