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

ChatGPT Pro

openai.com

Post reply on HN
Print view View original

561–570 of 1001 posts

Re: ChatGPT Pro

#561

Post by almog »

gorgoiler wrote: post #548
danpalmer wrote: post #515

Earlier quoted context omitted.

The problem is that OpenAI don't really have the enterprise market at all. Their APIs are closer in that many companies are using them to power features in other software, primarily Microsoft, but they're not the ones providing end user value to enterprises with APIs. As for ChatGPT, it's a consumer tool, not an enterprise tool. It's not really integrated into an enterprises' existing toolset, it's not integrated int…

Is their valuation proposition self fulfilling: the more people pipe their queries to OpenAI, the more training data they have to get better?

Not sure how valuation come into play here but I doubt enterprise clients would agree to have their queries used for training.

Re: ChatGPT Pro

#562

Post by andrepd »

> It also includes o1 pro mode, a version of o1 that uses more compute to think harder and provide even better answers to the hardest problems.

Great, we can throw even more compute and waste even more resources and energy on brute forcing problems with dumb LLMs... Anything to keep the illusion that this hasn't plateaued x)

Re: ChatGPT Pro

#563

Post by kridsdale1 »

Melatonic wrote: post #551
fudged71 wrote: post #482

Earlier quoted context omitted.

The new pricing tier suggests they're taking the Jobs approach - betting that their technology integration and reliability will justify premium positioning. But they face more intense commoditization pressure than either IBM or Apple did, given the rapid advancement of open-source models. The critical question is timing - if they wait too long to establish their enterprise position, they risk being overtaken by commo…

The Apple partnership and iOS integration seems pretty damn big for them - that really corners a huge portion of the consumer market. Agreed on enterprise - Microsoft would have to roll out policies and integration with their core products at a pace faster than they usually do (Azure AD for example still pales in comparison to legacy AD feature wise - I am continually amazed they do not priorities this more)

They don’t make any money from the Apple deal.

Re: ChatGPT Pro

#564

Post by j45 »

A_D_E_P_T wrote: post #444
Atotalnoob wrote: post #425

Earlier quoted context omitted.

AI writing detectors are snake oil

Yeah but they "detect" the characteristic AI style: The limited way it structures sentences, the way it lays out arguments, the way it tends to close with an "in conclusion" paragraph, certain word choices, etc. o1-Pro doesn't do any of that. It writes like a human. Damnit. It's too good. It just saved me ~6 hours in drafting a complicated and bespoke legal document. Before you ask: I know what I'm doing, and it did…

It still needs to be driven by someone who knows what they're doing.

Just like when software that was coming out, it may have ended jobs.

But it also helped get things done that wouldn't otherwise, or as much.

In this case, equipping a capable lawyer to be 20x is more like an iron man suit, which is OK. If you can get more done, wit less effort, you are still critical to what's needed.

Re: ChatGPT Pro

#566

Post by thunderbird120 »

dartos wrote: post #557
gcr wrote: post #549

Earlier quoted context omitted.

This isn't true. Transformer architectures generally take quadratic time wrt sequence length, not exponential. Architectural innovations like flash attention also mitigate this somewhat. Backtracking isn't involved, transformers are feedforward. Google advertises support for 128k tokens, with 2M-token sequences available to folks who pay the big bucks: https://blog.google/technology/ai/google-gemini-next-generat...

During inference time, yes, but training time does scale exponentially as backpropagation still has to happen. You can’t use fancy flash attention tricks either.

No, additional context does not cause exponential slowdowns and you absolutely can use FlashAttention tricks during training, I'm doing it right now. Transformers are not RNNs, they are not unrolled across timesteps, the backpropagation path for a 1,000,000 context LLM is not any longer than a 100 context LLM of the same size. The only thing which is larger is the self attention calculation which is quadratic wrt compute and linear wrt memory if you use FlashAttention or similar fused self attention calculations. These calculations can be further parallelized using tricks like ring attention to distribute very large attention calculations over many nodes. This is how google trained their 10M context version of Gemini.

Re: ChatGPT Pro

#567

Post by j45 »

Someone1234 wrote: post #496

Why doesn't Pro include longer context windows? I'm a Plus member, and the biggest limitation I am running into by far is the maximum length of a context window. I'm having context fall out of scope throughout the conversion or not being able to give it a large document that I can then interrogate. So if I go from paying $20/month for 32,000 tokens, to $200/month for Pro, I expect something more akin to Enterprise's…

If a $200/month pro level is successful it could open the door to a $2000/month segment, and the $20,000/month segment will appear and the segregation of getting ahead with AI will begin.

Re: ChatGPT Pro

#568

Post by Wolfenstein98k »

pests wrote: post #536
A_D_E_P_T wrote: post #407

I just bought a pro subscription. First impressions: The new o1-Pro model is an insanely good writer. Aside from favoring the long em-dash (—) which isn't on most keyboards, it has none of the quirks and tells of old GPT-4/4o/o1. It managed to totally fool every "AI writing detector" I ran it through. It can handle unusually long prompts. It appears to be very good at complex data analysis. I need to put it through i…

> the long em-dash (—) which isn't on most keyboards On Windows its Windows Key + . to get the emoji picker, its in the Symbols tab or find it in recents.

Well not for me it's not, that is a zoom function.

En dash is Alt+0150 and Em dash is Alt+0151

Re: ChatGPT Pro

#569

Post by heyjamesknight »

A_D_E_P_T wrote: post #407

I just bought a pro subscription. First impressions: The new o1-Pro model is an insanely good writer. Aside from favoring the long em-dash (—) which isn't on most keyboards, it has none of the quirks and tells of old GPT-4/4o/o1. It managed to totally fool every "AI writing detector" I ran it through. It can handle unusually long prompts. It appears to be very good at complex data analysis. I need to put it through i…

> Aside from favoring the long em-dash (—)

This is a huge improvement over previous GPT and Claude, which use the terrible "space, hyphen, space" construct. I always have to manually change them to em-dashes.

Re: ChatGPT Pro

#570

Post by danpalmer »

almog wrote: post #561
gorgoiler wrote: post #548

Earlier quoted context omitted.

Is their valuation proposition self fulfilling: the more people pipe their queries to OpenAI, the more training data they have to get better?

Not sure how valuation come into play here but I doubt enterprise clients would agree to have their queries used for training.

Yeah it's a fairly standard clause in the business paid versions of SaaS products that your data isn't used to train the model. The whole thing you're selling is per-company isolation so you don't want to go back on that.

Whether your data is used for training or not is an approximation of whether you're using a tool for commercial applications, so a pretty good way to price discriminate.

Post reply on HN