Intel's Breakthrough: AI Model Accelerates Scientific Research

0:00If you know me, you know that I absolutely love scientific applications of AI. When you use AI to advance science, especially in healthcare, to me, this is very exciting as it's creating tools that you can see very literal benefits to humankind. So I think what's really interesting is a new announcement that is out of Intel. And this is really exciting for me. It's kind of like a really juicy story because it's not just them getting into the science side, which I love, but is also them challenging the all-power, all-powerful NVIDIA, who is kind of the giant in the AI GPU space. So Intel recently announced the development of a couple things.

0:40Number one is the Aurora Gen AI project. So essentially, this is an AI model with this is insane 1 trillion parameters. We've seen this a couple other places. There's an open source version that just, you know, hit the top trending on hugging faces that is out of the EAU, I believe, or UAE. And this one, it has a trillion parameters. So it's, you know, really massive. This was announced recently at Intel's ISC23 event. And so essentially, they have this event on kind of high performance computing at the event they announced their supercomputer. And they also announced this new project. So this model essentially is fine tuned for scientific research and drug development.

1:22And essentially, I think this is going to be have a really massive ability to potentially accelerate the ability to find, well, essentially the identification of biological processes related to diseases like cancer. And I think this is also going to offer a lot of insights into drug design overall. So I think that the achievement of, you know, this really powerful, these really powerful AI models are quite substantial at the moment. And we're seeing this accelerate, especially when you have this paired with hardware optimization or the kind of you have these models, you know, this one that Intel's coming out with, paired with really impressive hardware.

2:01So right now, OpenAI's current supercomputer that they use, essentially has around 10 ,000 GPUs. And so Intel's new Aurora supercomputer that they also announced at this event alongside this new model is going to have 63 ,000 GPUs, right? So 10 ,000 GPUs for OpenAI and 63 ,000 GPUs for Intel. And I think that we're really going to see the winners in the AI space. I've been saying this for a while, but these are people that have hardware capabilities. It's not just software. You see, OpenAI is a big software company. They had to partner with someone like Microsoft that had the hardware firepower of Microsoft Azure.

2:44And you see the same thing with Anthropic. You know, they had to partner their software and they had to partner with Google, who had the hardware, the hardware kind of Google Cloud computing backend to really help bring that to scale. And so I think it's really interesting because right now we're seeing NVIDIA power the backends of a lot of these software companies, but it would be very interesting, number one, if NVIDIA jumps into the software side and starts, you know, creating a Chattapult competitor, for example. And likewise, we're now seeing Intel, who's always been famous for its computer chips and its hardware now jumping into the software side where actually it is training its own models.

3:23I think this is going to have a really big impact on the industry. Essentially, Intel is obviously one of the leading manufacturing chip manufacturers here in America. And I think that now NVIDIA, which I think currently has about a 95 % market share of generative AI GPUs, is going to feel quite threatened, right? NVIDIA just got pushed to a trillion dollar valuation the other day in their market cap. And so I think that our trillion dollar market cap. So I think that they're definitely going to feel threatened by someone like Intel coming into the space. And some are speculating that because of this, this might actually, you know, kind of force NVIDIA to start introducing newer cards earlier than previously thought to keep this industry leading position they have.

4:06I think Intel is also kind of become one of the big, you know, call them peer contenders in this market because Intel's data center GPU max series 1550 essentially has a 30 % average speed improvement over Nvidia's H100 in scientific workloads. So this new supercomputer that they have out, these optimizations and hardware are seeing some pretty significant optimizations. And I really think this is going to kind of tighten this entire race for who can build the most capable chips in the market um if that's going to be intel if that's going to be nvidia you see like everyone i think um google was recently bragging that their chip they have is like more effective than nvidia and facebook announced that they're going to be jumping in and creating chips that will be shipped in 2025 so i think when you have google intel facebook meta all coming out with these chips um it's going to really put some pressure on nvidia now that being said, I feel like could be wrong, but I feel like Google wants to keep their chips for themselves.

5:09Could be wrong there. They might use this to kind of leverage Google Cloud platforms. So it'll be interesting to see if NVIDIA can maintain this lead they currently have, or if this is going to be something quickly eroded now that the AI gold rush has arrived and people see just how lucrative all of this chip manufacturing can actually be and how much money is being generated in this space. I think something really exciting with this entire new scientific model that Intel has announced is that it works in a couple of various fields. So number one is biology, oncology, climatology, cosmology, and also materials chemistry.

5:49So the resulting models are going to be used in a pretty wide variety of scientific applications. I think that researchers in AI are going to find models for everything from molecule design to essentially energy production to power generation, everything you can do. And I think that models are going to be able to speed up the identification of, you know, different biological compounds linked to cancer and other diseases, and they're going to be able to suggest drug designs. We've recently seen that someone came up with a new antibiotic, essentially using an AI model that treated a superbug, which was previously untreatable and the AI model was able to take in 6 ,800 different compounds tested against this superbug and they identified 200 which were then taken to a lab and the lab came out with nine that were tested on mice exposed to the actual superbug and I think I believe all those nine were all effective but one of them in particular was picked as kind of like the new antibiotic for that so obviously this AI is really powerful but I think that with Intel coming out with a really powerful program like this Aurora.

6:55This is going to be expedited and we're going to start seeing this used more and more in science for drug discovery, for helping do research for diseases and a lot of other areas. And this is an area that I for one am very very excited about in the world of AI.