In a hotel in San Jose, in the heart of Silicon Valley (California, USA), AMD presented the next piece of the artificial intelligence revolution to lights and euphoric applause. This is the MI300, an accelerator that fits in the palm of your hand and which the Californian company claims can run artificial intelligence software faster than any product currently on the market. “We may be tired of hearing about artificial intelligence by now, but it is certainly the most transformative technological revolution of the last 50 years, probably even more than computers and the Internet,” said Lisa Su, CEO of Advanced Micro. Devices (AMD), during an event EL PAÍS was invited to in early December.
The launch is one of the most important in the five-decade history of the company, whose artificial intelligence processors are already used by the industry's most important companies and researchers to produce computers with integrated AI, including Microsoft, Open AI and Target. AMD Vice President and General Manager Jason Banta (Lubbock, USA, 44 years old) spoke to EL PAÍS at the end of the event about the possibilities of this technology, which is said to be making “dizzying and exciting” progress.
Questions. How is it possible that something that no one talked about five years ago has become the most important issue of today?
Answer. One of the most transformative aspects was large-scale language models. They have evolved quite quickly since 2017 and only really gained momentum when many people saw their stunning beauty and everything they can do. This has sparked a lot of additional interest, be it for chatbots or other generative AI applications. And it was probably the fastest evolution of a new technology we've seen in the industry, becoming a transformative tool in the cloud, on the PC and on other devices.
Q What does it mean to integrate artificial intelligence directly into computers?
R. That computers are more personal and can better understand how each user wants to interact. While previously we were the ones who had to learn to deal with them, in the future computers with artificial intelligence will be able to anticipate our needs. This enables new audio and video experiences and makes it easier to generate creative content, a process that can be challenging on a traditional computer. Devices with built-in AI take care of it and do everything for the users. But they will also make computers more secure by being able to detect and manage threats.
Q What can already be done on a practical level?
R. Nowadays, most of the major applications we see are collaboration with various tools. If you use Zoom or Teams, there are many areas where AI is used, such as image enhancement. And what can be achieved is that this happens with higher battery power. Another area is the creation of content or applications in the field of audio and video editing: this sector will continue to grow. However, it is still too early to realize the full potential of this technology for application to personal computers. We will see most of the progress in the near future.
Q How long will it take until everyone has an AI computer?
R. I think we'll see more of this sometime in 2024, although we'll have to wait until 2025 for a big tipping point. We need to take a big step forward in this direction first and then it will grow from there. But we'll see much greater adoption in the coming months until the AI laptop actually becomes mainstream in the next two years.
Lisa Su, CEO of Advanced Micro Devices, demonstrates the AI accelerator during the event in San Jose on December 6, 2023.
Q Is the performance of current computers sufficient to sufficiently develop AI?
R. Much of the development and training of AI models still takes place in the cloud. But in the next few years I hope to see training of smaller models run in real time on computers. This way, as you interact with an application, you can train it in real time to customize the way it interacts with you.
Q In which areas are the most radical changes expected?
R. On both mobile phones and laptops, the challenges are similar: adapting AI models to work efficiently on these devices. We are seeing initial progress in this regard. Many of the phones coming to market today already have artificial intelligence built in, but they need to be able to be used without having to charge them six or seven times a day.
Q What are the current limits of this technology?
R. Our biggest challenge is making this technology smaller and running efficiently on laptops. Some of this is starting to emerge. We just released a software package that enables what we call quantization: essentially making these large models smaller so they can run efficiently on laptops. Stay in a laptop's RAM, work efficiently and don't affect the life of a battery. This is the challenge we must address as an industry and we are providing the tools to make it happen.
Q Asian countries dominate the semiconductor supply chain. Can we expect Europe and the United States to improve their production?
R. Our goal is to create a geographically diverse semiconductor ecosystem. We believe this is positive for the industry as a whole. We work with governments and supply chain partners to promote this. We see this evolving over time, but we think it's important not to rely too heavily on a single region for semiconductor supply.
Q Is AI developed with environmental impact in mind?
R. We take the ecological perspective into account, it is at the forefront of our approaches when we think about the development of these technologies. It's also true that a computer with built-in AI saves, among other things, labor and time compared to a regular laptop, which brings real efficiency on a large scale. We are working on ways to optimize energy consumption, and these are very real impacts that we consider at every stage of our development.
You can follow EL PAÍS technology on Facebook and X or sign up here to receive our weekly newsletter.
Subscribe to continue reading
Read without limits
_