How AI Will Change Chip Design

End Moore's Law is approaching. Engineers and designers can do a lot to miniaturize transistors And pack as many of them into chips as possible. So they are turning to other approaches to chip designincorporating technologies such as artificial intelligence into this process.

Samsung, for example, adding artificial intelligence to memory chips to enable memory processingthereby saving energy and speeding up machine learning. Speaking of speed, Google's TPU V4 AI chip has doubled its computing power compared to the previous version.

But AI has even more promise and potential for semiconductor industry. To better understand how AI could revolutionize chip design, we spoke to Heather Gorrsenior product manager MathWorks' MATLAB platform.

How is AI currently being used to design the next generation of chips?

Heather Gorr: AI is such an important technology because it is involved in most stages of the cycle, including the design and manufacturing process. There are many important applications here, even in general process technology where we want to optimize things. I believe that defect detection plays an important role in all stages of the process, especially in manufacturing. But even if you think about the design process, [AI now plays a significant role] when you design lights, sensors and other components. There are many anomaly detection and error resolution techniques that you should really consider.

Heather GorrMathWorks

Then, if you think about the logistics modeling that you see in any industry, there is always planned downtime that you want to reduce; but you will also experience unplanned downtime. So, by looking back at the historical data of those times when maybe something took a little longer to make than expected, you can look at all that data and use AI to try to pinpoint the immediate cause or see something that might pop up even in the processing and design stages. We often think of AI as a predictive tool or as a robot doing something, but in many cases you get more information from data using AI.

What are the benefits of using AI for chip design?

Gorr: Historically, we've seen a lot of physics-based models, and it's a very intensive process. We want to do short order modelwhere instead of solving such an expensive and extensive model, we can do something a little cheaper. You can create a surrogate model of that physical model, so to speak, use the data, and then do your job. parameter sweepyour optimizations, yours Monte Carlo Simulation using a surrogate model. This requires much less computational time than solving physics-based equations directly. So we see this benefit in many ways, including the efficiency and cost-effectiveness that results from rapid iteration of experiments and simulations that really help with design.

So it's like having digital twin In terms of?

Gorr: Exactly. This is roughly what people do when you have a model of a physical system and experimental data. Then, with that, you have another model that you can tweak and try, try different parameters and experiment, which will allow you to cover all of these different situations and ultimately come up with the best design.

So it will be more effective and, as you said, cheaper?

Gorr: Yes, definitely. Especially during the experimentation and design phases when you try different things. This will obviously result in significant cost savings if you are truly into manufacturing and manufacturing. [the chips]. You want to model, test, experiment as much as possible without actually building something using a real process.

We talked about the benefits. What about the disadvantages?

Gorr: [AI-based experimental models] tend to be not as accurate as physics-based models. Of course, this is why you do a lot of simulations and parameter checks. But that's also the benefit of having a digital twin where you can take that into account: it won't be as accurate as the exact model that we've developed over the years.

Both chip design and manufacturing require system costs; you have to consider every little detail. And it can be really difficult. This is a case where you can have models to predict something and different parts of it, but you still need to put it all together.

Another thing to think about is that you need data to build your models. You have to combine data from many different sensors and different teams, and this makes the task more complex.

How can engineers use AI to better prepare and extract information from equipment or sensor data?

Gorr: We always think about using AI to predict something or perform some robotic task, but you can use AI to come up with patterns and pick out things on your own that you haven't noticed before. People will use AI when they have high frequency data coming from many different sensors, and in many cases it is useful to explore the frequency domain and things like data timing or resampling. This can be very difficult if you don't know where to start.

I would advise using the tools that are available. There is a huge community of people working on these things and you can find many examples. [of applications and techniques] on GitHub or MATLAB Centralwhere people shared good examples, even small applications they created. I think a lot of us are overwhelmed by data and just don't know what to do with it, so be sure to take advantage of what's already out there in the community. You can explore and see what makes sense to you and bring that balance of domain knowledge and the insight you get from the tools and AI.

What should engineers and designers consider?using AI to design chips?

Gorr: Think about what problems you're trying to solve or what ideas you're hoping to find, and try to clarify that. Consider all the different components, document and test each of these parts. Consider all the people involved, explain and communicate in a way that makes sense for the entire team.

How do you think AI will impact the work of chip designers?

Gorr: This will free up a lot of human capital for more complex tasks. We can use AI to reduce waste, optimize materials, optimize design, but there will still be a human involved whenever it comes to decision making. I think this is a great example of how people and technology work hand in hand. It's also an industry where everyone involved – even in manufacturing – needs to have some level of understanding of what's going on, so it's a great industry for AI development as we test things and how we think about them before we put them on a chip.

How do you see the future of artificial intelligence and chip design?

Gorr: This largely depends on the human factor – the involvement of people in the process and the availability of an interpretable model. We can do a lot with the mathematical details of modeling, but it all comes down to how people use it, how everyone involved in the process understands and applies it. Communication and inclusion of people of all skill levels will be very important. We'll see less of these hyper-accurate predictions and more transparency of information, sharing and the digital twin – not just using AI, but using our human knowledge and all the work that many people have done over the years.