The end of the Moore Law looming on the horizon. Engineers and designers can only do a lot to reduce the transistors and packages the largest possible number of them in chips. So they resort to other approaches to design chips, and integrate techniques such as artificial intelligence in this process.
Samsung, for example, adds Amnesty International to its memory chips to enable memory processing, thus providing energy and speeding up machine learning. When talking about speed, Google’s TPU V4 AI chip has doubled its treatment capacity compared to its previous ability.
But artificial intelligence holds more promise and capabilities to manufacture semiconductors. To better understand how artificial intelligence is set to revolutionize the design of chips, we talked to us Heather GoreDirector of the first product for MathWorksMatlab platform.
How is artificial intelligence currently used to design the next generation of chips?
Heather Gore: AI is an important technique because it participates throughout the course, including the design and manufacturing process. There are a lot of important applications here, even in general process engineering where we want to improve things. I think detection of defects is a large one in all stages of the process, especially in manufacturing. But even think about the design process, [AI now plays a significant role] When you design light, sensors and all different ingredients. There is a lot of detection of homosexuality and mitigating the mistakes you really want to think.
Heather GoreMathWorks
Next, think about the logistical modeling that you see in any industry, there is always time to stop the plan that you want to reduce; But you also end up stopping the unplanned work. So, given this historical data you enjoy when you have those moments that may take a little longer than expected to manufacture something, you can look at all this data and use artificial intelligence to try to determine the near cause or see something that may jump even in the processing and design stages. We often think of AI as a predictive tool, or a robot does something, but often you get a lot of insight from data through artificial intelligence.
What are the benefits of using artificial intelligence to design chips?
Gore: Historically, we have seen a lot of physics -based modeling, which is a very intense process. We want to do Reduced demand formAs instead of solving such an exorbitant mathematical and recovery model, we can do something a little cheaper. You can create an alternative model, if it is possible to express, about this physics -based model, use data, then do your work Teacher scanning operationsYour improvements, your, your Monte Carlo simulation Using an alternative model. This takes less time to resolve physics -based equations. Therefore, we see this benefit in several ways, including efficiency and economy, which is quickly the results of repetition on the experiences and simulation that will really help design.
So it is like a digital twin in a sense?
Gore: exactly. This is what people do to a large extent, as you have the physical system model and experimental data. Next, by conjugating, you have this other model that you can adjust, adjust and try different parameters and experiences that allow seizing all these different situations and reaching a better design in the end.
So, it will be more efficient, as I said, cheaper?
Gore: Yes, of course. Especially in the stages of experimentation and design, where you try different things. It is clear that this will lead to great cost savings if you are actually manufacturing and producing [the chips]. You want to simulate and test the experiment as much as possible without doing something using the actual process engineering.
We talked about the benefits. What about defects?
Gore: the [AI-based experimental models] It tends to be like physics -based models. Of course, for this reason it performs many simulations and parameters. But this is also the benefit of having this digital twin, where you can take this into consideration – it will not be accurate like this exact model that we have developed over the years.
Both chips design and manufacture are intense in the system. You have to think about every small part. This can be a truly challenge. It is a condition in which you may have models to predict something and different parts of it, but you still need to collect them all.
One other things to think also is that you need data to create models. You should integrate data from all types of different sensors and different types of difference, and this increases the challenge.
How can engineers use artificial intelligence to prepare and extract visions of devices or sensors?
Gore: We are always thinking about using artificial intelligence to predict something or do a robot task, but you can use artificial intelligence to reach patterns and choose things that you may not have noticed before. People will use artificial intelligence when they have high -frequency data coming from many different sensors, and it is often useful to explore the field of frequency and things like sync data or sampling. This can be a really challenge if you are not sure where to start.
One of the things I would like to say is, use the available tools. There is a wide community of people working on these things, and you can find many examples [of applications and techniques] on Gyrroup or Matlab CentralWhere people shared nice examples, even the small applications they created. I think many of us are buried in data and not sure what to do with it, so he definitely took advantage of what is already in society. You can explore and know what is logical to you, and bring the balance of knowledge in the field and insight you get from the tools and AI.
What should the engineers and designers consider WHUsing artificial intelligence to design chips?
Gore: Think about the problems you try to solve or what ideas that you might hope to find, and try to be clear about it. Consider all the different components, documenting and testing each of these different parts. Think about all the people concerned, explain and deliver them in a reasonable way to the entire team.
How do you think artificial intelligence will affect the functions of chips designers?
Gore: A lot of human capital will be free for the most advanced tasks. We can use artificial intelligence to reduce waste, to improve materials, to improve design, but then you still have this participating person whenever it comes to making decision. I think it is a great example of people and technology working alongside. It is also an industry where all the participating people – even on the manufacturing floor – needed to get a level of understanding of what is happening, so this is a great industry to advance artificial intelligence because of how to test things and how to think about them before we put them on the slide.
How do you imagine AI’s future and chips design?
Ghost: It greatly depends on this human element – which spoils people in this process and have this interpretative model. We can do many things with the sporting minutiae for modeling, but it is about how people are used, and how everyone understands and applied. Communication and participation of people from all levels of skills in the process will be really important. We will see less than these superpowers and more transparency of information, sharing, and this digital twin – not only using artificial intelligence but also using our human knowledge and all the works that many people have done over the years.
From your site articles
Related articles about the web