AMD CEO Lisa Su Notes: AI to Dominate Chip Design

[AleksandarK]

Artificial intelligence (AI) has emerged as a transformative force in chip design, with recent examples from China and the United States showcasing its potential. Jensen Huang, CEO of Nvidia, believes that AI can empower individuals to become programmers, while Lisa Su, CEO of AMD, predicts an era where AI dominates chip design. During the 2023 World Artificial Intelligence Conference (WAIC) in Shanghai, Su emphasized the importance of interdisciplinary collaboration for the next generation of chip designers. To excel in this field, engineers must possess a holistic understanding of hardware, software, and algorithms, enabling them to create superior chip designs that meet system usage, customer deployment, and application requirements.

The integration of AI into chip design processes has gained momentum, fueled by the AI revolution catalyzed by large language models (LLMs). Both Huang and Mark Papermaster, CTO of AMD, acknowledge the benefits of AI in accelerating computation and facilitating chip design. AMD has already started leveraging AI in semiconductor design, testing, and verification, with plans to expand its use of generative AI in chip design applications. Companies are now actively exploring the fusion of AI technology with Electronic Design Automation (EDA) tools to streamline complex tasks and minimize manual intervention in chip design. Despite limited data and accuracy challenges, the "EDA+AI" approach holds great promise. For instance, Synopsys has invested significantly in AI tool research and recently launched Synopsys.ai, the industry's first end-to-end AI-driven EDA solution. This comprehensive solution empowers developers to harness AI at every stage of chip development, from system architecture and design to manufacturing, marking a significant leap forward in AI's integration into chip design workflows.



View at TechPowerUp Main Site | Source

 

[mclaren85]

Yet another milestone for reaching technological singularity.

 

[Bomby569]

so just like Nvidia

 

[Assimilator]

Machine learning (which is not AI, sorry Jensen and Lisa) is basically the optimal use-case for designing something. Instead of having a team of experts spending months or years painstakingly sketching out a semiconductor from scratch, build an ML model with the desired inputs and outputs, and fire off N parallel instances. The ones that produce a working design get fed into the next generation, and you do the run again. Wash, rinse, repeat until you have a highly-optimised design.

That's the theory, anyway. In practice this doesn't scale very well for larger numbers of inputs and outputs, because the complexity, and therefore time taken, goes up exponentially. Just like human programmers, human chip design experts aren't going anywhere anytime soon.

 

[AusWolf]

Such a statement would be pretty normal coming from Nvidia. But from AMD? Hmm.

 

[Bomby569]

Machine learning (which is not AI, sorry Jensen and Lisa) is basically the optimal use-case for designing something. Instead of having a team of experts spending months or years painstakingly sketching out a semiconductor from scratch, build an ML model with the desired inputs and outputs, and fire off N parallel instances. The ones that produce a working design get fed into the next generation, and you do the run again. Wash, rinse, repeat until you have a highly-optimised design.

That's the theory, anyway. In practice this doesn't scale very well for larger numbers of inputs and outputs, because the complexity, and therefore time taken, goes up exponentially. Just like human programmers, human chip design experts aren't going anywhere anytime soon.

ML is a type of AI. They aren't wrong there. But i guess you're right when they deceptively use the term AI to boost company stocks over and over again.

 

[R0H1T]

Just like human programmers, human chip design experts aren't going anywhere anytime soon.

You'd be surprised how quickly they'd be replaced/usurped when the time comes but yes for now they're rather safe!

Unfortunately, Secret Invasion’s AI credits are exactly what we should expect from Marvel

Marvel’s foray into AI-generated are should shock no one.

www.theverge.com

How A.I. 'hallucinations' helped make Westworld's main titles

Exclusive video shows how A.I. processed images from Season 1 and 2 and created neural network's "hallucinations" of Westworld.

www.syfy.com

 

[AusWolf]

You'd be surprised how quickly they'd be replaced/usurped when the time comes but yes for now they're rather safe!

Unfortunately, Secret Invasion’s AI credits are exactly what we should expect from Marvel

Marvel’s foray into AI-generated are should shock no one.

www.theverge.com

How A.I. 'hallucinations' helped make Westworld's main titles

Exclusive video shows how A.I. processed images from Season 1 and 2 and created neural network's "hallucinations" of Westworld.

www.syfy.com

Imo, AI shouldn't be used to replace creative work. If we give up our creativity, we give up what makes us human.

I wouldn't mind AI-powered robots replacing manual labour, though.

 

[R0H1T]

Well it's not up to us, if Apple could squeeze a trillion dollar profits from Apple TV they most definitely would. The same goes for any number of bean counters out there!

 

[Sithaer]

I wouldn't mind AI-powered robots replacing manual labour, though.

Sounds good on paper but that would leave ppl like me with no jobs to do since manual labour is pretty much the only thing I can do/apply for in my area.

 

[Bomby569]

Sounds good on paper but that would leave ppl like me with no jobs to do since manual labour is pretty much the only thing I can do/apply for in my area.

i'm sure you can change the oil on the robots that will replace you

 

[R0H1T]

Right, not sure if you were semi serious with that or what

 

[bug]

Such a statement would be pretty normal coming from Nvidia. But from AMD? Hmm.

And since AMD can't manage to capture a significant part of the compute/ML/AI market, they may find themselves running chip design software on Nvidia hardware

Joking and irony aside, ML/AI doesn't sleep, doesn't tire. It just sifts through patterns all day long and generates variations (think genetic algorithms). It's all but expected it would stumble upon things humans missed, ignored or dismissed as infeasible. And it would be stupid not to leverage such a tool.

 

[R0H1T]

Yes just wait till they engineer a better CoVid SARS nCoV variant

 

[user556]

I suspect the driving force here is a desire to break away from the traditional Von Neumann (or Harvard) structures where main memory is pretty much one large lump and is becoming a bigger and bigger drag on both power consumption and relative latency. Neural Nets offer exactly that possibility of distributed memory and processing at the base hardware level but it's anyone's guess as to how functionally effective that might really be.

One thing that seems to be the case so far is NNs don't seem comparatively very power efficient. But me thinks there is much more room for that still to be improved. One example is swap the localised blocks of SRAM for MRAM.

 

[Frick]

ML is a type of AI. They aren't wrong there. But i guess you're right when they deceptively use the term AI to boost company stocks over and over again.

It's the buzzword version of ML, and honestly it sort of works. LLM and ML requires explanation, wheres AI does not.

 

[Bwaze]

Such a statement would be pretty normal coming from Nvidia. But from AMD? Hmm.

This is about designing processors - a task which actually makes sense to design it with machine learning.

Nvidia claims right now that ML, AI can and will help us (replace us) with everything else - text generation, art generation, poem writing, programming...

Karl Sharro on Twitter:

"Humans doing the hard jobs on minimum wage while the robots write poetry and paint is not the future I wanted."

 

[bug]

Yes just wait till they engineer a better CoVid SARS nCoV variant

I know where you're going with this. But in the real world, that is exactly what researchers do: they try to look at mutations or engineered variants, trying to anticipate what's coming and how to prepare. Well, it's not all they do, but it's an aspect they need to cover.

Fwiw, when a virus is engineered, it looks like Frankenstein's creature: (known) parts of other viruses patched together. SARS-CoV2 was examined early on and deemed not to exhibit signs of engineering. /off-topic

 

[R0H1T]

I don't agree with that last part but the bigger point about AI/ML or what's essentially recreating patterns/routines, or creating new ones, is that it's a major part of our daily lives ~ I have a feeling a lot of folks here don't realize how many jobs will be lost if major corporations go all out on this! Because a Skynet might happen later, we'll probably have an uprising in part created by jobs lost to this much earlier.

 

[bug]

I don't agree with that but the bigger point about AI/ML or what's essentially recreating patterns/routines, or creating new ones, is a major part of our daily lives ~ I have a feeling a lot of folks here don't realize how many jobs will be lost if major corporations go all out on this! Because a Skynet we'll probably have an uprising in part created by jobs lost to this.

Ready your sabots then, we've been through this during the Industrial Revolution. The change was dramatic, but somehow we still have jobs.

 

[R0H1T]

We didn't have 8 billion people back then.

 

[AsRock]

And since AMD can't manage to capture a significant part of the compute/ML/AI market, they may find themselves running chip design software on Nvidia hardware

Joking and irony aside, ML/AI doesn't sleep, doesn't tire. It just sifts through patterns all day long and generates variations (think genetic algorithms). It's all but expected it would stumble upon things humans missed, ignored or dismissed as infeasible. And it would be stupid not to leverage such a tool.

Man is flawed and any thing man makes has flaws, always has and always will.

Sooner or later it will bite us in the ass.

 

[AusWolf]

Sounds good on paper but that would leave ppl like me with no jobs to do since manual labour is pretty much the only thing I can do/apply for in my area.

Same with me, although Universal Income is here for the rescue.

 

[Bwaze]

Same with me, although Universal Income is here for the rescue.

By the time we will be redundant as a workforce there will be no Universal Income daydreams - since most of the profits and revenues will be generated by large multinational behemoths which of course don't pay taxes - their influence on governments will make that a certainty.

In fact, the money for pensions and healthcare will also run out (in such countries that still offer these luxuries).

 

[bug]

Same with me, although Universal Income is here for the rescue.

Universal Income will not rescue anything. I haven't read details about it (I need to, I know), having lived in a socialist country, I know first hand that giving people any sort of income without them having to work for it, just demotivates everyone. And I mean everyone: people that can cover their basic needs by doing nothing will not want a 9-to-5 because that's a lot of hassle, while capable people will be demotivated to see they are expected to work hard to enable others to live while doing nothing.