They have yet to prove it profitable/viable/useful, and 100,000 companies are working on eating their lunch. Others are walling off their data (e.g. reddit).
There might be doubts about how it's actually going to make any money. A sense of reality is starting to force it's way into the capital markets maybe.
GPT-4 is incredibly useful, and it's clear there's so much straightforward bits of additional product development on top of it in various verticals that will be incredible productivity multipliers in the respective verticals.
You could say the same about many SaaS projects. Why pay for an expensive GPU upfront and then some guy who can install it, configure it, create some sort of interface for you to talk to it... when you can just pay openai to do it for less money?
Because GPU-Servers that can run a typical LLM are less than 5k, which includes installation. Really, running an LLM seems to be no more complicated than running a NAS from a system administrators perspective.
I wonder where a critical threshold might lie? For example, right now if you only have 24Gb of VRAM, the best you can run locally at usable speeds are the 4-bit quantized LLaMa 30B models, which are only semi-usable. If you have 48Gb however, you can run the 4-bit quantized LLaMa 65B, which is much better.
I assume there will be advances that basically make the context window practically infinite. That ought to make the lower parameter count models much more powerful on its own. I also assume they will become more efficient to run through sparsity/pruning, though I'm a total novice on this topic.
I wonder how many generations of hardware are we talking? One? Two? Or three? It feels like it's potentially within that range.
I expect there to be GPT-4 level model (or close enough) that anyone can run in the major cloud providers soon. In this future, most AI powered products won’t be paying OpenAI anything.
Google still catching up for more than a month. Amazon still doesn't have anything comparable, but definitely needs and wants it. So it doesn't looks like it's easy to replicate. Besides, it's not a finished product. It's a service in continuous development. For how long OpenAI will be able to stay ahead of competitors is yet to be seen. They are trying to create a secondary market of resellers. This may keep them in business for long time if they are just "good enough".
But do they have moat? Or could the existing big tech companies enter the same market and offer the similar enough product? It is not like Amazon, Google, Microsoft and maybe even Facebook don't have enough money to throw at it and sell something.
Advertising? Who's going to click on advertising that doesnt do it now? Subscription? Its useful but how many people will pay a sub for it, its not netflix or porn. Licensing maybe? Incorporating it into other products but the competion is going to be intense so prices will drop or they'll have to be the best.
This isnt going to be the moneymaker people think it will be..
It might be worth it for you, but how many people like you are there in the market? Not many I suspect. Most wont be able to afford it or will go for free options.
Since we're both not providing data, I'd be interested in hearing your personal experience on this (if it's not too much of a bother, feel free not to respond!):
What do you do for a living, and do you find GPT-4 helpful for it?
I suspect as soon as top management realizes how useful it can be on all levels there will me mass subscriptions from companies. Of course OpenAI needs to convince it's safe and secure. New dark mode will help.
It is making money now. The really big difference with respect to all the other search engines, cat pictures websites, etc. is that the current users expect to pay and pay to have access to the API.
This is for me the key difference between something like Google Search and OpenAI/ChatGPT, from the start, people are paying to get answers to their questions.
They have absolutely no moat. Trained on published architecture using publicly available data. By the end of 2023 all companies including meta, Google, and Amazon will be have similar models. Only moat they currently have is tons of specially curated data for rlhf which other companies are catching up.
GPT-4 is pretty clearly not just public data + public papers. It's a lot of optimizations and other things on top of what is itself a lot of optimizations and other things that led to GPT-3.
We're going to have to wait and see if OpenAI has a moat. You could have argued in 2001 that Google had no real moat. All they were doing was indexing public data with an algorithm they'd already published in a research paper, after all. But it turned out to be a deep space with an endless potential for optimizations and improvements, combined with the need for great expertise in scalable computations and ad optimization that other companies just weren't able to catch up with even after trying hard for years.
Effectively structuring (publicly available) data for training is a moat. So is the details of the model, and the UX of how the model is presented to users.
Consider why Google won in search for so long, despite anyone being able to write their own web crawler in an afternoon.
They presumably have smart people working there to get the results they have so far. By the end of 2023 they will have moved ahead while the competition try to catch up to where they are now.
41 comments
[ 2.1 ms ] story [ 77.1 ms ] threadPossibly because of the bizarre terms and structure of other past rounds
They have yet to prove it profitable/viable/useful, and 100,000 companies are working on eating their lunch. Others are walling off their data (e.g. reddit).
It'll be interesting, at least.
GPT-4 is incredibly useful, and it's clear there's so much straightforward bits of additional product development on top of it in various verticals that will be incredible productivity multipliers in the respective verticals.
Emphasis on running, not training.
The best I can run is either LLaMa 65B at 1 token per second (too slow) on my CPU or LLaMa 30B on my GPU at quite a fast ~30 tokens a second.
Nowhere near the usefulness of GPT-4.
I assume there will be advances that basically make the context window practically infinite. That ought to make the lower parameter count models much more powerful on its own. I also assume they will become more efficient to run through sparsity/pruning, though I'm a total novice on this topic.
I wonder how many generations of hardware are we talking? One? Two? Or three? It feels like it's potentially within that range.
I Would be surprised if that is the case.
This isnt going to be the moneymaker people think it will be..
I'm a programmer, and I pay a subscription for GPT-4 because it's ridiculously worth it.
I'm curious, what do you do that you don't think GPT-4 is worth it?
What do you do for a living, and do you find GPT-4 helpful for it?
This is for me the key difference between something like Google Search and OpenAI/ChatGPT, from the start, people are paying to get answers to their questions.
We're going to have to wait and see if OpenAI has a moat. You could have argued in 2001 that Google had no real moat. All they were doing was indexing public data with an algorithm they'd already published in a research paper, after all. But it turned out to be a deep space with an endless potential for optimizations and improvements, combined with the need for great expertise in scalable computations and ad optimization that other companies just weren't able to catch up with even after trying hard for years.
Id be surprised if the same thing isn’t true for chatgpt.
The next dataset can come from ChatGPT users.
Consider why Google won in search for so long, despite anyone being able to write their own web crawler in an afternoon.