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

ChatGPT Enterprise

openai.com

Post reply on HN
Print view View original

161–170 of 532 posts

Re: ChatGPT Enterprise

#161

Post by dahwolf »

brabel wrote: post #10
ftxbro wrote: post #2

> For all enterprise customers, it offers: > Customer prompts and company data are not used for training OpenAI models. > Unlimited access to advanced data analysis (formerly known as Code Interpreter) > 32k token context windows for 4x longer inputs, files, or follow-ups I'd thought all those had been available for non enterprise customers, but maybe I was wrong, or maybe something changed.

I think the real feature is this: " We do not train on your business data or conversations, and our models don’t learn from your usage. ChatGPT Enterprise is also SOC 2 compliant and all conversations are encrypted in transit and at rest. "

It's exactly opposite. The entire point of an enterprise option would be that you DO train it on corporate data, securely. So the #1 feature is actually missing, yet is announced as in the works.

Re: ChatGPT Enterprise

#162

Post by irrational »

> The 80% statistic refers to the percentage of Fortune 500 companies with registered ChatGPT accounts, as determined by accounts associated with corporate email domains.

Yeah... I have no doubt that people at my Fortune 100 company tried it out with their corporate email domains. We have about 80,000 employees, so it seems nearly impossible that somebody wouldn't have tried it.

But, since then the policy has come down that nobody is allowed to use any sort of AI/LLM without written authorization from both Legal and someone at the C-suite level. The main concern is we might inadvertently use someone else's IP without authorization.

I have no idea how many other Fortune companies have implemented similar policies, but it does call the 80% number into question for me.

Re: ChatGPT Enterprise

#163

Post by YetAnotherNick »

coolspot wrote: post #109
YetAnotherNick wrote: post #42

Earlier quoted context omitted.

If you could offer stable 70B llama API at half the price of ChatGPT API I would pay for it. I know HN likes to believe everything is close to $0, but it is hardly the case.

(Not affiliated) https://together.ai/pricing

So it is 50% more expensive than OpenAI. Even if that was comparable it proves my point that you can hardly do it for "cost close to $0".

Re: ChatGPT Enterprise

#164

Post by dahwolf »

crooked-v wrote: post #94
ojosilva wrote: post #91

Clicked on ChatGPT / Compare ChatGPT plans / Enterprise ... > Contact sales Oops. Scary. I'm missing the Teams plan: transparent pricing with a common admin console for our team. Yes, fast GPT-4, 32k context, templates, API credits... they're all very nice-to-haves, but just the common company console would be crucial for onboarding and scaling-up our team and needs without the big-bang "enter-pricey" stuff.

Any "Contact sales" stuff has just been an instant "no" at any company I've ever worked at, because that always means that the numbers are always too high to include in the budget unless it's a directive coming down directly from the top.

It depends. We once were quoted 300K/year by a SaaS company. We replied by saying that our budget is 20K. "Fair enough, we'll take that".

Re: ChatGPT Enterprise

#165

Post by ttul »

bg24 wrote: post #142
ftxbro wrote: post #2

> For all enterprise customers, it offers: > Customer prompts and company data are not used for training OpenAI models. > Unlimited access to advanced data analysis (formerly known as Code Interpreter) > 32k token context windows for 4x longer inputs, files, or follow-ups I'd thought all those had been available for non enterprise customers, but maybe I was wrong, or maybe something changed.

I think you missed this part: ChatGPT Enterprise is also SOC 2 compliant and all conversations are encrypted in transit and at rest. Our new admin console lets you manage team members easily and offers domain verification, SSO, and usage insights, allowing for large-scale deployment into enterprise. I think this will have a solid product-market-fit. The product (ChatGPT) was ready but not enterprise. Now it is. They…

Just the SOC2 bit will generate revenue… If your organization is SOC2 compliant, using other services that are also compliant is a whole lot easier than risking having your SOC2 auditor spend hours digging into their terms and policies.

Re: ChatGPT Enterprise

#166

Post by BoorishBears »

fdeage wrote: post #144
BoorishBears wrote: post #74

Earlier quoted context omitted.

You don't fine-tune on a corpus of documents to give the model knowledge, you use retrieval. They support uploading documents to it for that via that code interpreter, and they're adding connectors to applications where the documents live, not sure what more you're expecting.

Yes, but what if they are very large documents that exceed the maximum context size, say, a 200-page PDF? In that case won't you be forced to do some form of fine-tuning, in order to avoid a very slow/computationally expensive on-the-fly retrieval? Edit: spelling

Fine-tuning the LLM in the way that you're mentioning is not even an option: as a practical rule fine-tuning the LLM will let you do style transfer, but you knowledge recall won't improve (there are edge cases, but none apply to using ChatGPT)

That being said you can use fine tuning to improve retrieval, which indirectly improves recall. You can do things like fine tune the model you're getting embeddings from, fine tune the LLM to craft queries that better match a domain specific format, etc.

It won't replace the expensive on-the-fly retrieval but it will let you be more accurate in your replies.

Also retrieval can be infinitely faster than inference depending on the domain. In well defined domains you can run old school full text search and leverage the LLMs skill at crafting well thought out queries. In that case that runs at the speed of your I/O.

Re: ChatGPT Enterprise

#167

Post by caesil »

happytiger wrote: post #105

Has there been some resolution to the copyright issues that I’m no aware of? In my conversations with execs that’s been a serious concern — basically that generated output from AI systems can’t be reliably protected. I refer to the concept that output could be deemed free of copyright because they are not created by a human author, or that derivative works can be potential liabilities because they resemble works that…

You're referring to a Copyright Office administrative ruling.

It's a pretty strange ruling at odds with precedent, and it has not been tested in court.

Traditionally all that's required for copyrightability is a "minimal creative spark", i.e. the barest evidence that some human creativity was involved in creating the work. There really hasn't traditionally been any lower bound on how "minimal" the "spark" need be -- flick a dot of paint at a canvas, snap a photo without looking at what you're photographing, it doesn't matter as long as a human initiated the work somehow.

However, the Copyright Office contends that AI-generated text and images do not contain a minimal creative spark:

https://www.copyright.gov/ai/ai_policy_guidance.pdf

This is obviously asinine. Typing in "a smiling dolphin" on Midjourney and getting an image of a smiling dolphin is clearly not a program "operat[ing] randomly or automatically without any creative input or intervention from a human author".

If our laws have meaning, it will be overruled in court.

Of course, judges are also susceptible to the marketing-driven idea that Artificial Intelligence is a separate being, a translucent stock photo robot with glowing blue wiring that thinks up ideas independently instead of software you must run with a creative input. So there's no guarantee sanity will prevail.

Re: ChatGPT Enterprise

#168

Post by graypegg »

happytiger wrote: post #105

Has there been some resolution to the copyright issues that I’m no aware of? In my conversations with execs that’s been a serious concern — basically that generated output from AI systems can’t be reliably protected. I refer to the concept that output could be deemed free of copyright because they are not created by a human author, or that derivative works can be potential liabilities because they resemble works that…

If something is in the public domain, and you create something new with it, you have the rights to the new work, there isn’t any sort of alike-licensing on public domain works in most-if-not-all jurisdictions.

This is why music engravers can sell entire books of classical sheet music from *public domain* works. They become the owners of that specific expression. (Their arrangement, font choice, page layout, etc)

If the AI content is public domain, and the work it generates is incorporated into some other work, the entity doing the incorporation owns the work. It’s not permanently tainted or something as far as I know.

Re: ChatGPT Enterprise

#169

Post by ttul »

ojosilva wrote: post #91

Clicked on ChatGPT / Compare ChatGPT plans / Enterprise ... > Contact sales Oops. Scary. I'm missing the Teams plan: transparent pricing with a common admin console for our team. Yes, fast GPT-4, 32k context, templates, API credits... they're all very nice-to-haves, but just the common company console would be crucial for onboarding and scaling-up our team and needs without the big-bang "enter-pricey" stuff.

The jump to enterprise pricing suggests that they have enormous enterprise demand and don’t need to bother with SMB “teams” pricing. I suspect OpenAI is leaving the SMB part up to Microsoft to figure out, since that’s Microsoft’s forte through their enormous partner program.

Re: ChatGPT Enterprise

#170

Post by IanCal »

dahwolf wrote: post #161
brabel wrote: post #10

Earlier quoted context omitted.

I think the real feature is this: " We do not train on your business data or conversations, and our models don’t learn from your usage. ChatGPT Enterprise is also SOC 2 compliant and all conversations are encrypted in transit and at rest. "

It's exactly opposite. The entire point of an enterprise option would be that you DO train it on corporate data, securely. So the #1 feature is actually missing, yet is announced as in the works.

You probably wouldn't want that, you'd want to integrate with your data for lookups but rarely for training a new model.
Post reply on HN