The actual reason why Kaparthy (and you'll see some other early employees) begin to consider leaving as they have already vested their shares and have taken profits during the hype.
>You think the total valuation of all companies divided by the total number of people comes out to a few cents?
No, where did I say that? The person I was replying to wants to share OpenAI's value specifically, so I only mention OpenAI. Also, I didn't say "a few cents", I said cents, intended as a relative contrast to "millions."
>The US GDP per capita was $80k in 2023. The global GDP per capita is $13K.
Yes, and the US per capita GDP would drop to $15.8k if the "excess" was redistributed to the rest of the world. But now everyone is equal, regardless of what they do, right? Is this fair?
>Even if you were to split up the valuation of OpenAI for every single person in the world, everyone would be getting $10.
That's not how it works and I think you know that. Where does this tangible $10 come from if there is no buyer? Or do you think someone is willing to spend $80B on a company which is then immediately destroyed because the money is taken from the shareholders and sent to everyone else? There would be no company left. Do you think that the very idea of this impending dissolution-upon-purchase wouldn't lower the value of the company drastically?
>what point are you actually trying to make?
That it's silly to suggest redistributing OpenAI compensation.
See my edit above, I never said "a few pennies" or "a few cents", I said "cents" intended as a relative contrast to "millions." We don't know how much a company is worth if it is expected to be immediately destroyed upon purchase, so if "dollars" or "10s of dollars" is more palatable to you, feel free to mentally substitute that instead.
“We don't know how much a company is worth if it is expected to be immediately destroyed upon purchase.”
That’s what you wanted to say.
Better, you could have lead with a question, “Why would someone buy a company at a huge premium if doing so immediately destroys the value upon purchase?”
Why would someone buy a company at a huge premium if that premium provides each employee financial freedom to leave, thus destroying the value of the company upon purchase?
We’re all disappointed with the inequities in reality on some level, but what would you suggest otherwise?
Do you at least recognize that a buyer would likely pay significantly less for a company if paying more meant employees (its most valuable asset) could just walk up and leave?
Have friends (not at OpenAI) that have been offered $1m+ yearly (these are not academic superstars either, solid staff engineers) to leave their startups and join Big Tech orgs.
Good for them and good for the people of OpenAI, their revenue growth has proven me wrong and is really exciting for the future of productivity.
AI Hype. Same applies to NVidia(40x sales valuation! The pickaxe seller).
Their technology can be easily replicated given enough time, just look at open source LLMs already beating GPT3. Google, Claude and others showing their strengths, already par to GPT-4 and even better in some scenarios.
Training is expensive, but it isn't expensive up to what they are valued at.
It's good because that capital likely made a few folks rich enough that they can consider pursuing their own companies, I'm curious about this.
I find it hard to believe the current products OpenAI offers will have good revenue streams given to how competition is catching up, low pricing power and how specialized deep learning typically beats LLMs or most of the things they offer.
I had even higher expectations when I saw this beginning, hoping that AI Startups would be able to do things that were previously not scalable with LLMs or OpenAI APIs, but haven't been able to see much here.
The biggest winner on this play was clearly Microsoft, that managed to create a new product line, both with Github Copilot and Copilot to sell to businesses.
Given enough time they will and won't be able to get 70%+ margins for their hardware.
There are multiple initiatives to eat their lunch which is based on Cuda.Their advantage isn't even in hardware, but rather how they've invested years and years in having integrated software + hardware solution for AI.
ATM as far as as datacenter GPUs good the price/performance is the only metric that counts (power costs are almost insignificant relative to how much a H100/etc. costs).
So Intel only needs to come up with something relative cheap that's good enough (it can easily use > 4 times more power to perform the same task) to force Nvidia to cut their insane margins.
Your idea of "easy" is stupid. Possible yes easy , no. And all that money they spend to catchup just to find Nvidia used that time to RD the next step. At this level it's a patent game. Also at this level the companies literally employ the best in their field. The best out of 8 billion. Replicating talent is impossible.
With what Google are doing with Gemini, OpenAI are in danger of losing the LLM crown. I think they'll still have a solid brand, but perhaps not the best at anything. The competition in AI is too ferocious.
But the first mover advantage they had will likely have very long-term benefits.
I tried the latest and greatest Gemini. In a few tasks I gave it, it did well but was clearly beaten by ChatGPT 4 (in some cases 3.5). It's still 1-2 years behind.
No. I don't think even the paid Gemini available now is behind ChatGPT4 for the things I use them for. People will be rooting for Google to lose this race but I think it's wishful thinking they were always too well positioned not to catch up and they've already done it IMO.
Whatever they lack in ability to make products is trumped by their infra + how game changing this stuff is so they were never going to just drop it even if it was just for internal use
> I find it hard to believe the current products OpenAI offers will have good revenue streams given to how competition is catching up
I have a hard time believing LLM products have good revenue streams at all, regardless of who’s making it. I still don’t use ChatGPT for anything. I think the developer world is obsessed with how well it generates code, but even that I don’t get the sense is worth $80B. I don’t find that feature to be extraordinarily useful, but I guess there are some codebases where it is? I don’t see the value-add applications outside of that. It’s not a search engine despite much hand-wringing about that. The whole document summarization / “style assistant” thing is a neat trick that might come in handy a few times a year. Chatbot girlfriends? Always money in porn I guess. Customer assistance bots? Annoying screen-waster. Writing internal emails? If you’re using it for that I’m sorry for you. Will anyone pay a standalone subscription fee for these in the long run or are they just flashy features to bake into existing products?
Maybe it’s because my knee-jerk response to hype of any kind is to apply extra doubt / scrutiny, idk.
It will all evaporate the day we get a better demo from someone else. The game has become capital/market capture before that event happens.
This is the colonization mentality of the west in over drive, only today capital/hype is used instead of armies. It will fail just faster than the colonies fell.
The Hype engineering mega machine doesnt understand effiency is a double edged sword. Getting over efficient in one dimension (hype/market capture) weakens the machine in other dimensions.
51 comments
[ 0.22 ms ] story [ 1588 ms ] threadIt is a possible reason. Unlikely, since employees on the level of Karpathy already don't have to work for money.
Hype, vesting, or comments like yours, are background noise for these people.
Imagine what having more millionaires would do for the world and the economy ? Such a silly system we’ve created.
The US GDP per capita was $80k in 2023. The global GDP per capita is $13K.
Even if you were to split up the valuation of OpenAI for every single person in the world, everyone would be getting $10.
How did you arrive at a few cents and what point are you actually trying to make?
No, where did I say that? The person I was replying to wants to share OpenAI's value specifically, so I only mention OpenAI. Also, I didn't say "a few cents", I said cents, intended as a relative contrast to "millions."
>The US GDP per capita was $80k in 2023. The global GDP per capita is $13K.
Yes, and the US per capita GDP would drop to $15.8k if the "excess" was redistributed to the rest of the world. But now everyone is equal, regardless of what they do, right? Is this fair?
>Even if you were to split up the valuation of OpenAI for every single person in the world, everyone would be getting $10.
That's not how it works and I think you know that. Where does this tangible $10 come from if there is no buyer? Or do you think someone is willing to spend $80B on a company which is then immediately destroyed because the money is taken from the shareholders and sent to everyone else? There would be no company left. Do you think that the very idea of this impending dissolution-upon-purchase wouldn't lower the value of the company drastically?
>what point are you actually trying to make?
That it's silly to suggest redistributing OpenAI compensation.
Speaking of which, how did you arrive at a few pennies in your estimate?
That’s what you wanted to say.
Better, you could have lead with a question, “Why would someone buy a company at a huge premium if doing so immediately destroys the value upon purchase?”
Let’s try it out.
We’re all disappointed with the inequities in reality on some level, but what would you suggest otherwise?
Do you at least recognize that a buyer would likely pay significantly less for a company if paying more meant employees (its most valuable asset) could just walk up and leave?
Have friends (not at OpenAI) that have been offered $1m+ yearly (these are not academic superstars either, solid staff engineers) to leave their startups and join Big Tech orgs.
Good for them and good for the people of OpenAI, their revenue growth has proven me wrong and is really exciting for the future of productivity.
Their technology can be easily replicated given enough time, just look at open source LLMs already beating GPT3. Google, Claude and others showing their strengths, already par to GPT-4 and even better in some scenarios.
Training is expensive, but it isn't expensive up to what they are valued at.
It's good because that capital likely made a few folks rich enough that they can consider pursuing their own companies, I'm curious about this.
I find it hard to believe the current products OpenAI offers will have good revenue streams given to how competition is catching up, low pricing power and how specialized deep learning typically beats LLMs or most of the things they offer.
I had even higher expectations when I saw this beginning, hoping that AI Startups would be able to do things that were previously not scalable with LLMs or OpenAI APIs, but haven't been able to see much here.
The biggest winner on this play was clearly Microsoft, that managed to create a new product line, both with Github Copilot and Copilot to sell to businesses.
There are multiple initiatives to eat their lunch which is based on Cuda.Their advantage isn't even in hardware, but rather how they've invested years and years in having integrated software + hardware solution for AI.
If leader is still innovating at a pace greater than any competition, then competition will never catch up.
Maybe AI is here forever, maybe there are different approaches that will blindside them?
So Intel only needs to come up with something relative cheap that's good enough (it can easily use > 4 times more power to perform the same task) to force Nvidia to cut their insane margins.
But the first mover advantage they had will likely have very long-term benefits.
Whatever they lack in ability to make products is trumped by their infra + how game changing this stuff is so they were never going to just drop it even if it was just for internal use
I have a hard time believing LLM products have good revenue streams at all, regardless of who’s making it. I still don’t use ChatGPT for anything. I think the developer world is obsessed with how well it generates code, but even that I don’t get the sense is worth $80B. I don’t find that feature to be extraordinarily useful, but I guess there are some codebases where it is? I don’t see the value-add applications outside of that. It’s not a search engine despite much hand-wringing about that. The whole document summarization / “style assistant” thing is a neat trick that might come in handy a few times a year. Chatbot girlfriends? Always money in porn I guess. Customer assistance bots? Annoying screen-waster. Writing internal emails? If you’re using it for that I’m sorry for you. Will anyone pay a standalone subscription fee for these in the long run or are they just flashy features to bake into existing products?
Maybe it’s because my knee-jerk response to hype of any kind is to apply extra doubt / scrutiny, idk.
This is the colonization mentality of the west in over drive, only today capital/hype is used instead of armies. It will fail just faster than the colonies fell.
The Hype engineering mega machine doesnt understand effiency is a double edged sword. Getting over efficient in one dimension (hype/market capture) weakens the machine in other dimensions.