51 | New York Times sues OpenAI and Microsoft, Midjourney AI-video, and more exciting AI news from this past week

0:00Hello and welcome to a short news weekend edition of the Leveraging AI podcast. This is Isar Maitis, your host. And like every week, a lot of big things happened. And I'm going to try to keep it short and to the point.

0:17And I want to get started with the big, huge news coming from OpenAI, that they're going to launch their GPT store next week. When OpenAI announced GPTs, which was in late 2023, they've announced that they're going to open a marketplace similar to the App Store for GPTs that people will be able to share and sell their GPTs to the public. And now several people, myself included, who has developed GPTs, received an email saying that the store will open next week. What it means, it means that ChatGPT, instead of just being a large language model, will become a marketplace of, if you want, AI-driven applications in a similar way to the App Store on Apple devices or Google Store on Google devices.

0:59It's still not clear how the monetization process is going to work, but I assume over time that will be clearer and clearer and probably will evolve until it gets to a situation where it's profitable for both the developers of the GPTs as well as for OpenAI. This move, while it's not surprising, again, they announced it before, raises several questions. The first question is, is this the App Store moment of AI? Meaning, if you think about what the App Store did to mobile devices, could this be the moment that completely democratizes AI? Because it builds very specific applications that will help the general public achieve things that maybe they did not exactly know how to do with AI before, and now will make it very easy and accessible.

1:41Time will tell, but it might be, and it's definitely an interesting opportunity to anybody who wants to jump in early. The second thing is, is this the first step to fool AI agents. And I truly believe that is the case. Meaning I think OpenAI released this to the public in its limited capability just to prep us and potentially themselves and the platform that they're developing to full AI agents. And if you don't understand what agents are, agents are little pieces of AI software that can actually take action on your behalf and do things. And over time, potentially even access software on the internet and do things for you, which can be anything you can imagine, like dealing with your bank statements, paying bills, booking travel, and managing your next board meeting.

2:29So whatever platform you will need access to, you can give it access and you will be able to take actions on your behalf. I truly believe that GPTs are version 0.1 of agents. And I think we will see more and more moves in that direction in 2024. And if I'm speculating even further out with this concept of developing small software using AI, I think over time, we'll get to the point that the actual app store, the app store that we use on Google and Apple devices will be somewhat an instant app store. Meaning instead of downloading apps that somebody developed, we'll be able to ask it to develop apps for us that will do just the one or two functions that we need instead of downloading full apps that do 50 things that we actually need to.

3:13And so I think in the long run, once AI will We'll be able to write code very quickly and very efficiently and understand our needs exactly. We'll be able to create apps on the fly, just talking to the AI and explaining what is the need that we have at that moment. The other huge piece of news from this past week that also touches OpenAI is that OpenAI got sued by the New York Times. This is not the first time somebody's blaming OpenAI for using their content for training without permission, but this is the New York Times. It's one of the most known establishments in the world when it comes to content creation and news.

3:49And beyond the fact that they're saying that they've used their articles without permission, which will be very hard to prove, but also very hard to do anything with, they're also saying that the fact that they're using the data in an unlicensed way and scraping their data is causing them financial damages. Now, this is something that can develop into something more interesting from a legal perspective. What they're asking the court to do is to make OpenAI remove all their content from the models and basically redo the models, which is something that I don't think is doable. And if it's doable, it will cost billions of dollars to OpenAI and Microsoft.

4:27So I don't see that happening. I assume this will end in some kind of a legal agreement where OpenAI slash Microsoft will pay them whatever compensation for the content that they've already used. And as we've seen recently, more and more AI companies are signing deals with content creators in news publications in order to get access to their content in a licensed way. And I definitely see more of that happening in 2024. And I see this case in court pushing into that in a much more aggressive way. So going back to trends in 2024, I definitely see more licensing deals happening between the AI companies and the content creation companies in order to make sure that the content that they're using is done in a legal way.

5:14I do see this court case potentially slowing down the release of new models in 2024, waiting for such deals to be signed, or at least some of these things to be cleared up from a legal perspective. Staying on the same topic on how these large models were trained without licensing the material, a leaked document called Me Journey Stylist that contains more than 16 ,000 artists and non-brand names became public. It includes a huge number of really famous artists and major brands that their content and creations were allegedly used without their permission. This document is a part of an ongoing lawsuit against MidJourney, and it raises big questions on the whole ethical aspect of the first generation of these AI tools, where it's very clear that they've done things that are either unethical or in many cases, probably illegal as well.

6:10And now the question is, what's going to happen with that in the future? I do not see a scenario where these things can be undone. I definitely see this being settled in a way that will somehow compensate these artists. And I definitely see the future releases of these models having some kind of licensing agreements with the people that are providing the training content for these models. Staying on Me Journey, Me Journey made a big announcement this week, and they said that they're now working on AI video generation. Those of you who don't know Me Journey, Me Journey is the leading text-to-image AI generator right now that generates the most stunning images.

6:49They've made a lot of big announcements late into 2023, including the release of a new interface through a website, including the release of MidJourney version 6, which is absolutely mind-blowing. And their move into video is not surprising because some of their biggest competitors are doing the same thing. They said that they're going to start working on the model in January and that they expect the release of that within a few months, which is very vague. But if they're going to move as fast as they've done on images, I don't expect that to be more than three to four months for us to have the first variation of this.

7:21And if it's going to be somewhat close to the level of performance that they're able to do with their images, it is going to be absolutely mind blowing. If you want to understand more what could be the implications of being able to create videos on the fly just by using text or source images that can be created by Me Journey, I'm going to be doing a live session on LinkedIn as well as on Zoom on February 22nd with an expert on this topic. we're going to create video ads for specific use cases and products using Me Journey to create images and then Runway to create videos out of these images. And it will show you how incredible that this technology already is.

8:02I think once Me Journey releases their video capabilities that will be integrated into their own platform, I see huge potential for that across multiple use cases. Another competitor of Me Journey that announced video capabilities is Leonardo. Leonardo is another text-to-image AI generator, and they have released a motion feature that allows you to animate AI art images into short video clips. They're building on top of stable video diffusion technology from Stability AI, and they made it into a very simple interface. Like in many other of these platforms, there's very little control how it's going to be animated, and it's currently limited to only three to four seconds videos, but the trend is clear.

8:45We're going to see huge improvement in AI video capabilities in 2024, more or less like we've seen with images in 2023. Staying on synthetic content generation, there are several companies who do voice cloning very well and then text to speech from this voice cloning, but all of them were paid platforms. And now a company called MyShell has released a model that they've developed together with MIT researchers as an open source model. So this open source model is called OpenVoice and it allows free and instant voice cloning with high precision, including what they claim is emotional and tone control, meaning it can sound happy or sad or excited and so on.

9:28This model, like many other open source model, is available on Hugging Face and on MyShell's website. And you can start using it literally right now to create highly accurate and well-nuanced text to speech. Another big player that I really like in the AI field is Perplexity AI. Perplexity is more of a search engine, even though you can do a lot of things you can do with the other chat engines. And I use it regularly to look for stuff and research different things. They just announced that they've raised $73.6 million on a$520 million valuation. The interesting thing with Perplexity is that you can use their own model, but you can also replace the underlying model and run third-party ones to achieve different results in your searches and in your summaries and so on, including GPT-4 from OpenAI.

10:20It's a great tool that if you're not using yet, it's worth giving it a try. And some pretty big news from Microsoft in the past week. As you know, Microsoft has been pushing Microsoft Copilot very aggressively across all channels, both it means in its capabilities as well as in how people can access it. And these moves from this week are all pushing further in that direction. So the first thing is that Copilot can now run and execute code like Python and Java and JavaScript and so on, similar to the paid version of OpenAI, which means in addition to generating code, it can also analyze data and files using this feature.

10:58So this is continuing the trend from the past few months that we're seeing Microsoft releasing for free the paid features that are under OpenAI ChatGPT, things like plugin and DALI image generation that are already available under the Copilot Microsoft brand. Two additional updates that show how aggressive Microsoft is pushing Copilot is, there's now a Copilot app on both iOS and Android that is providing all the features that we've discussed. And it's in parallel to the Bing app, which is actually a very interesting move. And it feels, to me at least, that the Bing app will slowly decay as they will keep on pushing the co-pilot brand and the co-pilot capabilities.

11:38And I think that eventually is gonna replace Bing completely as search will become just a feature under the bigger AI capabilities. And I think we'll see similar things from Google and probably Apple in the future as well. The cool thing about these apps, as I mentioned, is that they provide access to the GPT-4 model and all of its capabilities that are paid features under OpenAI and are completely free under the Microsoft brand. And another interesting push of Copilot from Microsoft is Microsoft just announced that the new Windows keyboards will have a Copilot button to launch Copilot on any computer.

12:17If you remember the last time Microsoft did something like this, they added the Windows key to keyboards almost 30 years ago. Sorry, keyboards haven't changed much in 30 years, but now all the new Microsoft-based keyboards, including the ones that they generate, as well as third-party computers that are aimed to run Microsoft operating systems, will have this new Copilot button that you'll be able to press to launch Copilot and use it immediately. This is very clear that Microsoft is going all in on AI capabilities, and they're pushing it across every platform that they have access to. It will be very interesting to see how quickly and how aggressively Google and Apple responds to that.

12:54And I have no doubt that we'll see them responding sometime early this year. That's it for this week. There were a lot of other small pieces of news that I'm going to skip, but I think these are the key, most important things that happen. On Tuesday, we'll be back again with an interview episode as always. And until then, have an amazing weekend.