Revolutionizing AI: Major Updates from OpenAI & ChatGPT

0:00Today on the podcast, we're going to be talking about some breaking new features that OpenAI has added to ChatGPGPT. In addition, we're going to be talking about just how much money Microsoft has made from this OpenAI investment they've made. And we're also going to be talking about something that might get OpenAI into a little bit of hot water training off of a data set they may not be allowed to train off of. So without further ado, let's jump into the podcast. The first story I want to talk about today is the fact that OpenAI launched a number of new features for ChatGPT. And in a previous podcast, I talked about some new coming soon features that some people had been able to kind of reverse engineer from looking at some code.

0:39But this is not those features. This is something actually completely new. So OpenAI actually announced several different updates to the GPT 3.5 Turbo and GPT-4 models. And essentially, they're looking to increase, you know, their functionality and their usability for developers. So I would say one of the most significant updates is the introduction of a function calling capability in the chat completion API. So essentially, this feature is going to allow developers to describe programming functions, and then the model is going to generate code to execute those functions, which is very, very interesting.

1:15Now, a lot of people debated on why this wasn't done sooner. I think security is a big issue when you actually look at having something like this AI execute functions within the code. A lot of people talk about, so like for security, if you're able to put code into something like a chatbot and have it execute that code, it technically like hackers and other people can try to use that to inject code into OpenAI and ChatGPT's servers or databases or other areas and wreak havoc. So that is one area that is a little tricky that they definitely had to overcome. But overall, the ability to be able to do that is a really powerful tool.

1:55And so I think this is something that developers are going to absolutely love. The next thing that a lot of people have been hyping up is or a lot of people have been really excited about is that in GPT 3.5 Turbo. So this is, you know, what everyone essentially gets. They are going to have a 16 ,000 token context window. So what this means is that they are going to be able to take around four times longer of text as an input into chat GPT to be able to give you a response. And just for like context on how much 14 ,000 tokens is, right, because it isn't even words. It's just like parts of words.

2:34The way they break it up is kind of weird. But this is approximately equal to 20 pages of text. And that's in like a single request that, you know, like someone's API could send a single request to chat to open AI of 20 pages long. And yeah, like I said before, it's about four times longer than the original 4000 token context window they had before. And so this is going to allow for some very long prompts and prompt response exchanges within chat GPT. people are really excited about this. Now, one reason why a lot of people are thinking that they have done this is the fact that Anthropix Claude AI chatbot has been getting a lot of attention recently for the ability to upload massive amounts of text to this.

3:20I think you can upload a, I've seen some demos where people uploaded PDF documents to Claude that were like, you know, there were like 300 pages long or like an entire book essentially. And it was able to in a couple seconds, read the whole thing and get an answer to a question out of that. So I think OpenAI obviously is at a much larger scale than Anthropics Cloud. I believe the only way currently for consumers to use Anthropics Cloud is through Poe.com, which is created by Quora. And I think that's the only consumer use case of it right now. And so obviously, this is, you know, something very small, it's not available to the mass public.

3:58So they have the ability to do some of these, crazier stunts or whatever you want to call it, where they're essentially allowing much larger things, where if they were actually to scale that to the user base that OpenAI and ChatGPT has, it would be pretty brutal on the servers, which as Sam Altman recently said before Congress, they're already struggling to have enough capacity and GPUs and everything else that they need. So I don't believe that ChatGPT would necessarily be able to do something like an entire book. But 20 pages already. That's pretty incredible. So this is a pretty massive update that's going to be integrated.

4:36OpenAI has also made some pretty big cost reductions for their platforms. So the pretty popular text embedding tool they have called Ada2, and that model essentially translates text into numeric formats that are more understandable by machines. So that is now coming at a 75 % lower price. They're cutting the price 75 % on that. And in addition, uh, GPT 3.5 is now going to be about 25 % cheaper. Um, and that includes, uh, API use. So that's going to be pretty exciting. And, um, some of these, some of these are going into effect right now, but I think most of these features are going to be available by June 27th.

5:20So you may have to wait a little bit before you can start uploading a 20 page text documents to chat GPT, but overall, this is a pretty massive update that Chad Chippee has announced. And we're pretty excited to see where this goes. The other story I want to talk about today is the fact that Microsoft's chief financial officer recently said, so everyone knows that open Microsoft invested originally a billion dollars, and they followed that up with another$10 billion. So they're, they're all in at$11 billion on opening right now. And a lot of people are saying, you know, that's expensive, is opening, I can be able to make the money back.

5:53This is kind of before Chad Chippee really exploded to quite the level it's at now. I think now people are confident, you know, OpenAI is going to reach some pretty impressive numbers on users and growth and revenue. But in any case, the CFO of Microsoft recently said that AI products will add$10 billion in revenue. And so she told investors that, and she included, you know, OpenAI contributed at least$10 billion to the company's revenue, she said. She didn't specify the time period over which that happened, but Microsoft generated around$198 billion in revenue last year. And the next generation AI business will be the fastest growing.

6:37So what she said is the next generation AI business will be the fastest growing 10 billion business in our history, she said. so while Microsoft currently is really benefiting from all these new AI features they aren't free and that's why they're able to you know make money off of them the company so far has invested into chat GPT pretty heavily and they've done that in a couple different ways number one is with cash and number two is with cloud credit so they've actually given OpenAI a lot of cloud credits for their Microsoft Azure cloud computing platform which has helped OpenAI really train their their uh train their models well and in addition to this microsoft actually spent 1.2 billion dollars to build a supercomputer that runs open ai's models so that's something not a lot of people know but they've actually they've helped in a lot of different ways they've been a very strategic partner um and so essentially microsoft stands to get a significant chunk of the profits from OpenAI that has been integrated into its own business.

7:44So all the different ways that OpenAI has automated things within Microsoft and all the cost savings just from their own company is definitely another area that you want to take into account. So I think since February, Microsoft has added new features powered by OpenAI to most all of its software. You've seen in PowerPoint and Excel and Word and everything else. and also that includes the chatbot they added to Bing search. So there has been a lot that has been added there and they've also integrated it into Office 360 and their workspace software. So Microsoft, I think one of the main ways that they're making this$10 billion back is that they're charging a pretty significant premium to customers of their AI-powered Office 360 features.

8:29And so they also, in addition to that, have the ability to resell OpenAI software to customers on Microsoft Azure cloud platform. So I've talked to companies that are actually doing this right now, where essentially they're able to run their own instance of chat GPT on Microsoft Azure, where it keeps all of their company's data private and on their own server, and they're not worried about it ever being used to train the open AI model. And so, you know, that kind of gets over the issues that people like Chase Bank and some others had, you know, know, Sony as well that banned chat GPT because they're worried about, you know, private company data getting leaked through that.

9:08So this is pretty impressive. Microsoft is definitely making a lot of money from this. And I think that, you know, that$10 billion investment into OpenAI was definitely a big win for the company. It's going to help them to really invest in growth and everything they're doing to push the entire industry of AI forward and potentially for any lawsuits they may face. I recently saw an article on the information, which was talking about the fact that, like, essentially, Google last month, when they upgraded their Bard chatbot, they added a new kind of machine learning model that can better understand conversational language.

9:48And the idea is that it can compete with chat GPT. And in addition to that, they said, you know, one of their biggest competitive advantages that they have YouTube. So YouTube is the single biggest and, you know, arguably the richest source of imagery, audio and text transcripts on the internet. And Google researchers have been using it to develop really their next language model, Gemini. So according to a person that has a little bit of info on what's happening there, the value of YouTube has not been lost on OpenAI either because apparently the startup has secretly used data from YouTube to train some of their own AI models.

10:31And so I think this is going to be really interesting to see how that plays out if that leak gets them in trouble, if YouTube or Google tried to sue, if that's against their terms of service and kind of what happens there, right? We know that OpenAI is, you know, trained on a vast number or a vast data set of the internet, But for the most part, you know, they were able to get licenses and access, you know, YouTube gave them access to their API and Reddit gave them access to their API. And now that all of these companies are seeing how valuable their data is, you know, they're shutting it off, you see this massive, you know, uprising on Reddit now that the their API is getting the price of is getting, you know, massively increased.

11:09And, you know, Elon Musk shut off OpenAI's access to Twitter's data and their API. so it's gonna be really interesting to see if opening I gets in trouble from YouTube for training based off of that data set as well and this is an area we'll definitely have to continue watching into the future