Inside OpenAI's Checkbook: The $540M Investment in ChatGPT
10:41 Share

Today on the podcast, we are going to be talking about the cost of training ChatGPT, how much opening eye is spending on their entire company, how much they historically have spent and what is happen in the future. We're going to go over some of the cost financial time and when you probably didn't think about water.

And we're also going to talk about what other companies are doing to help mitigate these costs, who some of the people spending the most money on and why microsoft and OpenAI I are going to start making some new financial models and some new products in the very near future to help offset some of these costs. So without further, do let's jump into the podcast. I think the biggest thing we want to talk about today is the fact that um there's a big headline out of the information that says opening eyes losses double to five hundred and forty million dollars as IT developed ChatGPT.

So it's pretty crazy as they were building out chage. B T. Apparently, they had to hire a bunch of key employees from google. And these employees can be very, very expensive. Some of them are being paid up to five million dollars a year. And when your company's biggest open an eye and your product is absolutely a pivotal and the central and based on these um A I experts, you're just gonna pay whatever IT costs, especially when you've already received you know a billion dollars from microsoft and other investors and you're just going to spend whatever IT takes speed to get your product to market.

So apparently um they were spending over five hundred 之间， forty million dollars last year to get ChatGPT you ready, developed and launched um and what's interesting is this is you a previously unreported number, a bunch of people from within the company. I just leaked that and is really a lot of this goes into just a steep costs of training these machine learning models. And before you know tragic ity launched, this was a huge cost as a lot of the stuff had to be done from scratch and IT was a massive undertaking.

So even after they've launched, the revenues kind of picked up reaching an annual pace of hundreds of millions of dollars um which is what we've done just a few weeks after they launched their paid version of chat tript in february. The cost of running the entire company, I don't really think, are going to um slow down. I think you're gona keep raising as they're getting more customers and as they continue to train more and more advanced artificial intelligent uh, artificial intelligence models.

And this technology is obviously rapidly developing and they are the ones that are painful and having to stay really on top of IT. So the C E O sam altman, he um privately suggested that open A A I may try to raise as much as a hundred billion dollars in the coming years. That is absolutely insane.

So far the biggest checks that have been written to them, elon wrote some pretty big multi million dollar checks at the very beginning, in some say of hundred million dollars, a microsoft gave them a billion dollars, and then most recently, microsoft gave them another ten billion dollars, which brings microsoft up to about a fifty percent owner of the company. And so the fact that they're looking at raising ten ex up to one hundred billion dollars in the coming years, I think really just speaks to the fact that they think it's going to cost you a lot more to continue to develop these capabilities. As you can see, they need to spend a lot of money on um image.

They have dolly too, but it's getting you know pretty soundly beaten by the journey at the moment. Uh, one can only assume that they are going to be getting more into audio and more into video in the future. And so as they kind of start looking into everything they want to do, um this is going to get incredibly expensive.

So very interesting, they might be looking to spend looking to raise one hundred billion dollars. I think this is pretty important because right now we have a lot of other people creating these kind of AI models. We have google working on board.

We have bloomberg um and recently bloomberg said that they use one point three million hours of G P U. Time to help build the blooming burg GPT large language model. And so all of these um are incredibly expensive at the moment and everyone's kind of rate.

Now there's almost like this big um A I fight or arms race support GPU server power from NVIDIA from other platforms. Um because everyone is trying to train these big huge models because because it's such a hot topic right now. And so as they are fighting over at the costs are going up even more.

I think in addition, another cost that we really have a factor into all of this on top of just the training in the talent that ChatGPT historical and opening I historical spent, is the fact that people trying to train new models are gonna have to start spending more on the data. ChatGPT kind of got under the radar because they're going to be an open source platform. They have the keys to the kingdom when he came to twitter and apps to a ton of different reit included that just allowed them to use the data for free.

And now that um everyone is starting to monetize these platforms, companies like stack overflow and read IT have both said that if you would like to use their data, you have to now pay them so ChatGPT may have gotten in at a time when uh, they were getting this kind of content on the cheap. And it's left to be said if companies are going to retroactively make OpenAI pay for some of the data that they previously trained on one other area that not a lot of people think about when they think about the costs of training these models is the water required. So a recent um a recent round of research just came out saying that training ChatGPT required enough water to fill a nuclear reactors cooling tower.

Um I think it's really important to know that uh in this new research they said the GPT three alone so that's before GPT report five in GPT four consumed about one hundred and eighty five thousand gallons or seven hundred thousand leaders of water and so that's an average of every single cut conversation that someone is having. The ChatGPT is equivalent of them dumpling out a large bottle of fresh water onto the ground. So the people in the study are obviously are emphasizing the environmental concerns with this, talking about the historic droughts that the U.

S. Is facing right now and the fact that companies like google training, google bartering now have all of their um servers and cloud facilities in states like texas that are even more prone to draw and h and actually requires more water because there's more evaporation. I think in the study, they specially focused um not on water that was recycled because these servers right the recycled water through them to help cool them down, but specifically water that was evaporated or lost.

And what I do you think IT is important to look at a special we're looking at droughts and you cycling, you're using water from local rivers and streams. I do think that um this is probably less of a issue than I don't know nuclear waste, as this article title eludes as they bring in nuclear reactors, try make a seem really radioactive and scary. I mean this is water IT evaporates.

Um I don't know what else to say about IT. I think we're all going to survive, although IT is an interesting thing to keep an eye. As you know, more and more A I models come out and they all need more water to make sure that you we are using that and scaling that in a responsible manner.

Now all this to say there are a ton of costs, right? We have the costs of the employees, the costs of training, the costs of um now your data and the cost of your water, I guess. And finally, the final thing that I wanted to talk about is the fact to that OpenAI and microsoft are now implementing new ways to recoup some of these costs, are spending new product lines.

And one of the biggest ones that is happening right now is that microsoft may begin to offer private ChatGPT two businesses, as you know, a number of different businesses have banned their employees from using ChatGPT. Most recently, samsung had a bunch of employees who leaked confidential company information and source code to ChatGPT. And I I say leak.

I'm assuming they are developers who are saying they're pasting code in insane. Hey, based off this code, what can I do? Blob, a block is I have developers from my companies that do that, and a lot of a different people that are getting development advice and code advice from ChatGPT.

And so this happened to samsung. They banned all of their employees from using these AI models and they said they're going to work on some internal tools, which I don't know how good they will be, but that's what their plan is. So because of that um companies microsoft specifically and in microsoft because they gave the invested ten billion dollars into opening, they got a special deal where they're allowed to resell um opening eyes products.

And so you see that with being and you see us a lot of the other products are incorporating, but specifically, they're literally allowed to resell ChatGPT to companies privately. So private would be a would be able to go and buy a special version, tragic ity, on exclusive servers that are completely cut off. None of the data that goes into that will ever be used for anything at all on the company's fix account, and IT won't ever be incorporated into any future training data.

Now the fact that they're doing this also makes you think that what you're currently telling ChatGPT, if you're not on this plan, is being used for the training data. But that's another conversation um ChatGPT and opening. I also have hinted that they will be launching a similar product to this in the near future.

So I think between microsoft and opening eye, this is gonna a new way that they are trying to recoup some of their costs and making money back from this as companies like verizon, chase city group in golden sacks all took similar steps at banning their companies from really using this, uh, these A I models for fear of leaking private or sensitive information. And so I think that once these companies that have banded these pretty big companies are able to let their employees use IT, there's going to be a massive boom in the consumption of these products, is going to make a lot of money, open a eye. And it's also going to add a lot of innovation and I believe a lot of automation for these companies that are allowing a their employees to now start using IT.

This going to be really interesting area to watch. If you like the podcast a day, make sure to subscribe to our newsletter. There's a link in the description, make sure to download the self pause A I life coach APP absolutely amazing for helping you get whatever you're trying to get done, accomplished, focus on on your goals and also make sure to join us on our facebook group.

We talk all the time about what is happening in AI. It's a really awesome conversation that a lot of people are having. A lot of amazing insights are shared. We'd love to see you there. Thank you so much for listen to the podcast and has always have a great rest of your day.