coup , what coup. It s a deal , and all parties are happy so it seems to be a good compromise. But except for the 2 scenarios given, there are also 12319 that are not considered
I thought I read somewhere that basically Microsoft was giving them all the infrastructure they need for free. Sorry no evidence.
It would not be possible to cloud host an operation this reliant on GPU power …. the costs would be way too high compared to self hosting. GPUs in the cloud are very expensive, have availability issues, and tend to be behind the state of the art consumer GPUs.
This press release[^1] indicated MS does intend to provide infrastructure, including investing in providing more supercomputer infra specifically for openai:
> Microsoft will increase our investments in the development and deployment of specialized supercomputing systems to accelerate OpenAI’s groundbreaking independent AI research.
It isn’t clear what “provide” means exactly (is it free? At cost?) but I suspect there are guarantees of availability as well as discounts involved to ease operating costs.
It's hard to see why there would be a real cost difference, besides the cost of paying the middle man. If the middle man owns you, and is implicitly paying for your compute investments either way, they might be inclined to give you a discount.
If anything, it's a great opportunity for Azure to scale their ML infrastructure and test it on large problems.
Also, "Used by OpenAI" could be useful for marketing. So there are lots of incentives here for MS to let OAI use their services at reduced cost.
The two main points of the article seem to conflict:
1: That models will become commodities and all value will be captured by who owns the interface to the customer.
2: That it is an "absolutely crazy" deal because after paying ~$250B back to investors, OpenAI will own "100% of whatever was created".
If 1 holds true, then it seems to be a good deal for Microsoft. Especially if OpenAI uses their $10B in funding to buy compute power from Azure at market price.
Microsoft has a shot at dethroning Google as the world's answer machine and becoming the world's AI datacenter.
Running the world's answer machine is a $150B/year business for Google.
Being the world's AI datacenter is probably going to become a $100B/year business too.
Who doesn't see that Chat-GPT coined a new hype cycle is either blind or doesn't want to see it. How long that cycle lasts and what it will end up with, noone knows, of course. But Microsoft places a large bet here to be just the first contender. I bet other companies will follow and 80% of them will lose their investment, as usual.
Wait for the first publicly traded opportunities to take part in that hype cycle.
"Buffoon" is a good first order approximation, but it doesn't quite capture the level of greedy, careless, malicious deceit and malevolence to the economy, environment, and truth, nor the complete lack of empathy and motivation to do anything positive for the world and their fellow human beings.
It's kind of like the difference between Shakes the happy go-lucky imbecilic alcoholic children's birthday clown -vs- Pennywise the evil supernatural serial child killer clown.
I agree and disagree. It’s absolutely going to go stratospheric. But it also has clearly created something of immense value. Unlike bitcoin, which has no value or use, I at least have incorporated chatgpt in my workflow along side Kagi to great effect. I would gladly pay N multiples of my Kagi subscription to gain access to a faster and more reliable (the UX breaks constantly) version. Despite my latent wish to be unique I know there must be many more like me out there. This is my direct experience - :
* I have found ways to make it the most valuable tool in my toolbox (albeit in its current form not the exclusive tool - it’s simply wrong enough that’s not plausible).
I have come to a point that I will pay money on a subscription to keep access.
My use and value are in no ways unique, clever, or inaccessible to basically anyone.
* it has a hype cycle, so it gets advertising for free
These things are what immense fortunes are made of, and as long as they can defend the IP of the product, it’ll be more enduring than crypto which ticks only the last box of my list. Yes, it’ll be pumped to the moon and see a painful correction. But if they can play the game right, I don’t see why they couldn’t have in their hands the basis of find next Netscape, Google, or iPhone.
The challenge is everyone know what they did, a lot of the research underpinning is public, and the stakes are so high that very smart people are now very motivated. Unless they have an enormous patent war chest it will be reproduced by others and OpenAI itself may not be the “winner” that takes all as the article seems to think.
These days I’ve been writing a lot of concurrent / async rust and sometimes confused about the borrow semantics in various situations, especially when there is pinning, future closure, and macro interactions. I can ask it pretty loosely specified questions and it gives pretty good answers with clear examples. Sometimes it makes wrong assertions, like trait methods can be async, or it just makes stuff up (it told me there was a .wait on futures that block evaluates the future without the function being async, which is not true). But it’s largely right and very clear in its explanations to my direct question. Searching on Kagi gives me lots of results that might talk obliquely about my question, but if I take what I need more detail on than chatgpt can give me and search in Kagi I can find the actual facts faster by being guided by the chatgpt answer.
Being able to interrogate along a path of questioning is a lot better than being given reference material that doesn’t address my area of confusion.
Skimming that post, it seems to be saying that either OpenAI will fail quickly, in which case MS takes the downside, or it will create AGI, in which case OpenAI will own something of incomparable value. However, there's no reason to assume that their machine learning business won't mature into something that is only worth on the order of $10^11 - $10^12.
Clickbait title, author must not know what coup means. This post starts off with some tangent Sam Altman worshipping (useless info other than to justify using his name in the title) then goes on to talk about OpenAI hemorrhaging money with a struggling revenue stream and doing a $10B raise from Microsoft.
I suspect the OP misunderstands successful startups must scale when they have buzz or else they lose the ability to hustle and hack together more product.
A startup's goal is a special kind of scalable business designed to burn a reasonable (per risk) amount of cash (by billionaire standards) to experiment (in series rounds) if the business can blow-up into something big with a giant exit. There is a metric crapton^9 amount of cash held by the top, but nowhere "safe" to put it because VC is itself a "lifestyle" business, e.g., there's little way to efficiently scale relationship-building, due-diligence (which is often excessive and pointless), and oversight. :)
For anyone as allergic to clickbait as I am, this seems to be the meat of the article, the rest is filler (lots of “wow”, “absolutely amazing”, and a backstory so extensive it’s almost a wonder it doesn’t go all the way back to the Napoleonic Wars):
> At this point, Microsoft will have invested a total of $13B in OpenAI. Moreover, new VCs are in on the deal by buying up shares of employees that want to take some chips off the table.
> However, the astounding size is not the only extraordinary thing about this deal.
> First off, the ownership will be split across three groups. Microsoft will hold 49%, VCs another 49%, and the OpenAI foundation will control the remaining 2% of shares.
> If OpenAI starts making money, the profits are distributed differently across four stages:
> 1. First, early investors (probably Khosla Ventures and Reid Hoffman’s foundation) get their money back with interest.
> 2. After that Microsoft is entitled to 75% of profits until the $13B of funding is repaid
> 3. When the initial funding is repaid, Microsoft and the remaining VCs each get 49% of profits. This continues until another $92B and $150B are paid out to Microsoft and the VCs, respectively.
> 4. Once the aforementioned money is paid to investors, 100% of shares return to the foundation, which regains total control over the company.
19 comments
[ 3.0 ms ] story [ 39.8 ms ] threadIt would not be possible to cloud host an operation this reliant on GPU power …. the costs would be way too high compared to self hosting. GPUs in the cloud are very expensive, have availability issues, and tend to be behind the state of the art consumer GPUs.
> Microsoft will increase our investments in the development and deployment of specialized supercomputing systems to accelerate OpenAI’s groundbreaking independent AI research.
It isn’t clear what “provide” means exactly (is it free? At cost?) but I suspect there are guarantees of availability as well as discounts involved to ease operating costs.
[1] https://blogs.microsoft.com/blog/2023/01/23/microsoftandopen...
I wonder if there’s a connection.
If anything, it's a great opportunity for Azure to scale their ML infrastructure and test it on large problems.
Also, "Used by OpenAI" could be useful for marketing. So there are lots of incentives here for MS to let OAI use their services at reduced cost.
1: That models will become commodities and all value will be captured by who owns the interface to the customer.
2: That it is an "absolutely crazy" deal because after paying ~$250B back to investors, OpenAI will own "100% of whatever was created".
If 1 holds true, then it seems to be a good deal for Microsoft. Especially if OpenAI uses their $10B in funding to buy compute power from Azure at market price.
Microsoft has a shot at dethroning Google as the world's answer machine and becoming the world's AI datacenter.
Running the world's answer machine is a $150B/year business for Google.
Being the world's AI datacenter is probably going to become a $100B/year business too.
Wait for the first publicly traded opportunities to take part in that hype cycle.
It's kind of like the difference between Shakes the happy go-lucky imbecilic alcoholic children's birthday clown -vs- Pennywise the evil supernatural serial child killer clown.
Shakes:
https://www.youtube.com/watch?v=8sf1Eu06UCw
Pennywise:
https://www.youtube.com/watch?v=JvHyqWSBG3o
* I have found ways to make it the most valuable tool in my toolbox (albeit in its current form not the exclusive tool - it’s simply wrong enough that’s not plausible). I have come to a point that I will pay money on a subscription to keep access. My use and value are in no ways unique, clever, or inaccessible to basically anyone. * it has a hype cycle, so it gets advertising for free
These things are what immense fortunes are made of, and as long as they can defend the IP of the product, it’ll be more enduring than crypto which ticks only the last box of my list. Yes, it’ll be pumped to the moon and see a painful correction. But if they can play the game right, I don’t see why they couldn’t have in their hands the basis of find next Netscape, Google, or iPhone.
The challenge is everyone know what they did, a lot of the research underpinning is public, and the stakes are so high that very smart people are now very motivated. Unless they have an enormous patent war chest it will be reproduced by others and OpenAI itself may not be the “winner” that takes all as the article seems to think.
Can I ask you to elaborate with a few examples?
Being able to interrogate along a path of questioning is a lot better than being given reference material that doesn’t address my area of confusion.
A startup's goal is a special kind of scalable business designed to burn a reasonable (per risk) amount of cash (by billionaire standards) to experiment (in series rounds) if the business can blow-up into something big with a giant exit. There is a metric crapton^9 amount of cash held by the top, but nowhere "safe" to put it because VC is itself a "lifestyle" business, e.g., there's little way to efficiently scale relationship-building, due-diligence (which is often excessive and pointless), and oversight. :)
> At this point, Microsoft will have invested a total of $13B in OpenAI. Moreover, new VCs are in on the deal by buying up shares of employees that want to take some chips off the table.
> However, the astounding size is not the only extraordinary thing about this deal.
> First off, the ownership will be split across three groups. Microsoft will hold 49%, VCs another 49%, and the OpenAI foundation will control the remaining 2% of shares.
> If OpenAI starts making money, the profits are distributed differently across four stages:
> 1. First, early investors (probably Khosla Ventures and Reid Hoffman’s foundation) get their money back with interest.
> 2. After that Microsoft is entitled to 75% of profits until the $13B of funding is repaid
> 3. When the initial funding is repaid, Microsoft and the remaining VCs each get 49% of profits. This continues until another $92B and $150B are paid out to Microsoft and the VCs, respectively.
> 4. Once the aforementioned money is paid to investors, 100% of shares return to the foundation, which regains total control over the company.