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

GPT-5.1: A smarter, more conversational ChatGPT

openai.com

Post reply on HN
Print view View original

261–270 of 766 posts

Re: GPT-5.1: A smarter, more conversational ChatGPT

#261

Post by airstrike »

KronisLV wrote: post #223
airstrike wrote: post #141

Earlier quoted context omitted.

I mean, yes, but also because it's not as good as Claude today. Bit of a self fulfilling prophecy and they seem to be measuring the wrong thing. 4% of their tokens or total tokens in the market?

> I mean, yes, but also because it's not as good as Claude today. I'm not sure, sometimes GPT-5 Codex (or even the regular GPT-5 with Medium/High reasoning) can do things Sonnet 4.5 would mess up (most recently, figuring out why some wrappers around PrimeVue DataTable components wouldn't let the paginator show up and work correctly; alongside other such debugging) and vice versa, sometimes Gemini 2.5 Pro is also pret…

Out of curiosity, did you try asking Opus 4.1 as well?

Re: GPT-5.1: A smarter, more conversational ChatGPT

#262

Post by spaqin »

93po wrote: post #152
angrydev wrote: post #33

Earlier quoted context omitted.

Exactly. Stop fooling people into thinking there’s a human typing on the other side of the screen. LLMs should be incredibly useful productivity tools, not emotional support.

Food should only be for sustenance, not emotional support. We should only sell brown rice and beans, no more Oreos.

Oreos won't affirm your belief that suicide is the correct answer to your life problems, though.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#263

Post by porridgeraisin »

KalMann wrote: post #234
Y_Y wrote: post #192

Earlier quoted context omitted.

A citation would be a link to an authoritative source. Just because some unknown person claims it's obvious that's not sufficient for some of us.

Expecting every little fact to have an "authoritative source" is just annoying faux intellectualism. You can ask someone why they believe something and listen to their reasoning, decide for yourself if you find it convincing, without invoking such a pretentious phrase. There are conclusions you can think to and reach without an "official citation".

Yeah. And in general, not taking a potshot at who you replied to, the only people who place citations/peer review on that weird faux-intellectual pedestal are people that don't work in academia. As if publishing something in a citeable format automatically makes it a fact that does not need to be checked for reason. Give me any authoritative source, and I can find you completely contradictory, or obviously falsifiable publications from their lab. Again, not a potshot, that's just how it is, lots of mistakes do get published.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#264

Post by NullCascade »

halifaxbeard wrote: post #153
angrydev wrote: post #33

Earlier quoted context omitted.

Exactly. Stop fooling people into thinking there’s a human typing on the other side of the screen. LLMs should be incredibly useful productivity tools, not emotional support.

How would you propose we address the therapist shortage then?

I don't know why you're being downvoted. Denmark's health system is pretty good except adult mental health. SOTA LLMs are definitely approaching a stage where they could help.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#265

Post by visarga »

astrange wrote: post #43
moi2388 wrote: post #35

Earlier quoted context omitted.

Source? You can opt out of training, and delete history, do they keep the prompts somehow?!

It's not simply "training". What's the point of training on prompts? You can't learn the answer to a question by training on the question. For Anthropic at least it's also opt-in not opt-out afaik.

There is a huge point - those prompts have answers, followed by more prompts and answers. If you look at an AI answer in hindsight you can often spot if it was a good or bad response from the next messages. So you can derive a preference score, and train your preference model, then do RLHF on the base model. You also get separation (privacy protection) this way.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#266

Post by NullCascade »

9cb14c1ec0 wrote: post #132
Workaccount2 wrote: post #129

Earlier quoted context omitted.

OpenAI said that only ~4% of generated tokens are for programming. ChatGPT is overwhelmingly, unambiguously, a "regular people" product.

Yes, just look at the stats on OpenRouter. OpenAI has almost totally lost the programming market.

As a happy OpenRouter user I know the vast majority of the industry directly use vendor APIs and that the OpenRouter rankings are useless for those models.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#267

Post by dash2 »

joquarky wrote: post #249
nikkwong wrote: post #157

Earlier quoted context omitted.

Who ever claimed there was a therapist shortage?

The process of providing personal therapy doesn't scale well. And I don't know if you've noticed, but the world is pretty fucked up right now.

... because it doesn't have enough therapists?

Re: GPT-5.1: A smarter, more conversational ChatGPT

#268

Post by abound »

Dilettante_ wrote: post #170

>GPT‑5.1 Thinking’s responses are also clearer, with less jargon and fewer undefined terms Oh yeah that's what I want when asking a technical question! Please talk down to me, call a spade an earth-pokey-stick and don't ever use a phrase or concept I don't know because when I come face-to-face with something I don't know yet I feel deep insecurity and dread instead of seeing an opportunity to learn! But I assume thei…

In defense of OpenAI in this particular situation, GPT 5 can be incredibly jargon-y at times, making it much worse of a learning tool than other LLMs. Here's some response snippets from me asking a question about dual-stack networking:

> Get an IPv6 allocation from your RIR and IPv6 transit/peering. Run IPv6 BGP with upstreams and in your core (OSPFv3/IS-IS + iBGP).

> Enable IPv6 on your access/BNG/BRAS/CMTS and aggregation. Support PPPoE or IPoE for IPv6 just like IPv4.

> Security and ops: permit ICMPv6, implement BCP38/uRPF, RA/DHCPv6 Guard on access ports, filter IPv6 bogons, update monitoring/flow logs for IPv6.

Speaking like a networking pro makes sense if you're talking to another pro, but it wasn't offering any explanations with this stuff, just diving deep right away. Other LLMs conveyed the same info in a more digestible way.

Re: GPT-5.1: A smarter, more conversational ChatGPT

#270

Post by aquafox »

Having gone through the explainations of the Transformer Explainer [1], I now have a good intuition for GPT-2. Is there a resource that gives intuition on what changes since then improve things like more conceptually approaching a problem, being better at coding, suggesting next steps if wanted etc? I have a feeling this is a result of more than just increasing transformer blocks, heads, and embedding dimension.

[1] https://poloclub.github.io/transformer-explainer/

Post reply on HN