This is more of a navigational episode which I called AI Wars to explain a little bit how the ecosystem is heating up and how competition is actually hitting up in the AI world and what are the values. I’m going to take into account three […] kinds of business ecosystem that are shipping up right now. We’re going to call it the first one, it’s […] the open AI Microsoft, which we can call like closed and centralized ecosystem for AI. And then we’re going to get the more open Stability AI ecosystem. And the third one, which we’re going to look at is more like let’s call it the still closed but advertising based AI ecosystem.
But again, those are mostly speculations and […] I want to emphasize that there is not a clear cut distinction between open and closed. And also Open AI has released and will be releasing many open staff. Still, the journey that Open AI has chosen to release its products on the market is through a controlled API access, which they justified as a way to actually control what […] people and businesses can do with their underlying foundational models on the one side. But on the other side, we cannot deny that there may be also business reasons […] for doing that. And as of now, we don’t know 100% what’s the code behind and what’s the data behind the models that have been used by Obama to train its products.
So let’s take into account those three ecosystems to understand a little bit how the words are shipping up. And in the notes this show you’re going to find the various episodes that I recorded, where you’re going to find detailed explanation of how things are moving forward. As I’m ramping up the reporting of di industry in the last month, I’m going to be covering even more. And I hope that all those things are getting useful to you to understand what’s going on right now, how are things evolving […] and why this is relevant, together with the opportunity by also understanding the drawback of this industry right now. So, as I said on the first place, open AI Microsoft, it’s an incredible partnership.
I covered the history, I covered the possible benefits, business modeling, potential business ecosystem they might get shipped at, and also […] the limitations and also what can go wrong with the partnership between open AI and Microsoft. And as I explained, this is a partnership which is looking at Microsoft, which is providing the supercomputing infrastructure through Azure to openi, to keep working on, releasing and updating its underlying products. So. Gerard Models from […] GPT. […] So […] let’s break down for now […] the open air products into three categories.
Into the language models, which is […] GPT-3, going forward into conversational interfaces like Chad GPT, and […] into image generation, video generation like Daly. Now, there is an interesting aspect here. Those three things that I’m bringing apart from for now, as if they were a separate product, they might become a single product in the future. Meaning that […] the purpose of many of those AI players is actually to create something which is multimodel. But when this is going to happen, […] we don’t know yet for sure.
And it might be the result of scale. It might need a different paradigm. We might get there sooner or later. But I’m bringing things down right now for the purpose of understanding how the business is getting shaped through, like, again, language models, conversational interfaces and image, video, audio generation. I mean, […] the other modalities that go behind text to text.
And so right now, open AI. Again, Microsoft is providing the supercomputing infrastructures through Azure to enable Open AI to build its products. And then on the other side, those products are getting controlled via the access to APIs that you can find on the Open AI site. And then on the other side, Microsoft is already […] monetizing those APIs through the Azure Enterprise platform. So anything that Open AI releases, you’re going to find it through Azure as a product that you can integrate into a cloud so that you can launch web applications that you can host on top of […] Microsoft.
Now, […] that’s the first ecosystem. The second one is the Stability AI ecosystem with Stable diffusion. Where Stable Diffusion is actually using Amazon AWS as a partner to actually host […] the training of its models. There is not direct investments of AWS into Stability AI i. There is simply a partnership most probably where […] it might be so, not 100% sure, but it might be that Stability AI I is getting […] substantial discounts in terms of computing power to actually train and serve those models.
Because let’s remember also a year Sublidi releases Open sources something like Stable Diffusion. But then on the other side, it also operates it through a platform called Dream Studio where subdi provides APIs similarly to what Opendi does, where developers can connect and can build applications on top of and then Sabida also offer consulting and enterprise services. […] Now, the interesting part is the other […] part of how those models are served to users is based on the there is not, again, also a formal partnership with Apple. But Apple seemed suited to be the perfect […] partner for Stability AI, at least for the release and ability of the Apple […] devices to serve stable diffusion. Because stable diffusion is an open source […] generative model, which right now […] it’s quite suited to be […] served on an infrastructure […] of chips like the narrow engine of Apple that bought on its devices, like desktop devices like the Mac.
And then on the iPhone can provide the infrastructure to actually make sure that those models can be served and can be used to actually get in context training. Meaning that, for instance, imagine the case of, […] as explained in previous episodes of […] Stable Diffusion run on top of the iPhone of a user. By accessing the data of the user on the device so that there is no moving outside the device, it can customize the experience for the user. So imagine the case of a future where you have a movie made with stable diffusion and it’s a movie which is interactive, meaning that the story is going to change based on the preferences, for instance, of the user. I’m just making things appear […] to be fancy, but imagine this case where the movie made with stable diffusion can be customized based on InContext learning that the stable diffusion model does on top of the device of the user.
Now of course this is the ecosystem of Stability AI I where on the one side again you get AWS as a supercomputing […] platform to develop its generative models and then sabridi monetizes those models through pretty much APIs or through enterprise services. And then on the other side it releases those models open source and those open source models can be served on devices like Apple, devices like […] the iPhone, or all the devices that are running through the neural engine of Apple. So all the chips that have been developed and designed by Apple in the last years. I also speculated about the potential of AR […] because I speculated that AI seems to be the perfect interface for another physical platform which is AR. But it seems for news that came out just today that AR might take much longer than expected due to technical developments that are proving to be much more challenging.
Challenging for which it may be that Apple may not release an AR ad set for this year or next year, but instead it might go toward releasing like a mixed reality ad set to grasp the differences. A mixed reality ad set is going to be a combination of VR and AR where you’re not going to be using it in the real world, in a real world environment, you’re going to be using it in a controlled environment. So let’s say on your desk to actually simulate inside business productivity apps that you can use, for instance to have a virtual office, something like that, or like web games or entertainment. And the reason might be related to various things like making sure that those Dr smart glasses might be really good to interfacing with real world, understand the context and move as we move along in real world and also that those are powerful enough and safe enough to make them viable from a commercial standpoint. So those might be some of the challenges that are ahead when it comes to the two AR and then a third ecosystem, let’s call it the advertising based ecosystem I’m calling like that.
But again, this mostly speculation is the release that might become anytime in 2023 of Sparrow which is going to be a charge GPT like […]conversational AI interface launched by DeepMind, part of Google, so research AI research lab part of Google. And this interface most probably is going to be […]fixing a lot of the drawbacks of charge PT and it may be more fact based. It may be connected to the Internet. It may be selling sources, it might have an index. So knows it might be that Google may launch something which is complementary to its search products.
Because of course, if you’re at Google you’re thinking, how do I actually empower my core products, which is search, and now […]launch a product has the potential to be monetized through advertising. So those are all things that might be happening right now. So again, I’m looking at three scenarios right now. And then of course, we might also see competition coming from Facebook, Amazon that are definitely looking into […]right now the development of AI models to strengthen their core business model. But right now I’m just developing this competition scenario around three main.
Again, let’s say ecosystems. As I said, you have the OpenAI Microsoft, primarily distributed through API Standpoints and through Azure. […]Then on the other side you have the stability AI ecosystem, primarily distributed through APIs. And then through the open source, it’s […]enabling the development of a supercomputer, AI supercomputer within AWS. And then on the other side, also […]specialized chips on top of Apple devices.
And then a third scenario of course, is the advertising based AI conversational interface through the development of Sparrow, which is going to be coming anytime. It might be coming anytime in 2023. So this is how AI works are able.