Friday, March 18th 2022

Google Uses Artificial Intelligence to Develop Faster and Smaller Hardware Accelerators

Designing Artificial Intelligence / Machine Learning hardware accelerators takes effort from hardware engineers in conjunction with scientists working in the AI/ML area itself. A few years ago, we started seeing AI incorporated into parts of electronic design automation (EDA) software tools, helping chip designers speed up the process of creating hardware. What we were "used to" seeing AI do are just a couple of things like placement and routing. And having that automated is a huge deal. However, it looks like the power of AI for chip design is not going to stop there. Researchers at Google and UC Berkeley have made a research project that helps AI design and develop AI-tailored accelerators smaller and faster than anything humans made.

In the published paper, researchers present PRIME - a framework that created AI processors based on a database of blueprints. The PRIME framework feeds off an offline database containing accelerator designs and their corresponding performance metrics (e.g., latency, power, etc.) to design next-generation hardware accelerators. According to Google, PRIME can do so without further hardware simulation and has processors ready for use. As per the paper, PRIME improves performance upon state-of-the-art simulation-driven methods by as much as 1.2x-1.5x. It also reduces the required total simulation time by 93% and 99%, respectively. The framework is also capable of architecting accelerators for unseen applications.
As far as comparing the test results, PRIME was in its prime time (pun intended). Google's very own EdgeTPU was compared to PRIME-made design, and the AI-generated chip was faster with latency improvement of 1.85x, resulting in a much quicker design. Researchers also noticed a fascinating thing, and that the framework-generated architectures are smaller, resulting in a more minor, less power-hungry chip. You can read more about PRIME in this free-to-access paper here.
Sources: Google AI Blog, via The Register
Add your own comment

49 Comments on Google Uses Artificial Intelligence to Develop Faster and Smaller Hardware Accelerators

#2
kiriakost
Google, failed be a quality search engine, failed at YouTube as administrator, failed as email hosting service.
Failed
in its and every sector of engagement with real people.
Who knows, they might succeed in sectors that there is no people there.
Posted on Reply
#3
dcf-joe
seth1911...
Sounds like we have an AI posting here :laugh: :roll:
Posted on Reply
#4
Chrispy_
I'd laugh if Google rebranded itself to Cyberdyne Systems for April 1st this year.
Posted on Reply
#5
TheinsanegamerN
seth1911Brain Diarrhea
seth1911 appears to be the result of an AI being trained exclusively on the machine translated youtube comment sections of indonesian vtubers.
Posted on Reply
#6
TheinsanegamerN
seth1911...
:roll: :roll: :roll: We gettin SPICY in here! :roll::roll::roll:
Posted on Reply
#7
Chrispy_
Wow, this thread hit Godwin's Law on post #9?
JFC. I'm off to bleach my eyes, have a nice evening all...
Posted on Reply
#8
AsRock
TPU addict
kiriakostGoogle, failed be a quality search engine, failed at YouTube as administrator, failed as email hosting service.
Failed
in its and every sector of engagement with real people.
Who knows, they might succeed in sectors that there is no people there.
Funny thing that failed, as they all so succeeded too, that's if you like it or not.
Posted on Reply
#9
InVasMani
So basically in summary a machine learning physically compressed chip engineered inference artificial intelligence designing ecosystem technology habitat. It's alive!!!
Posted on Reply
#10
Steevo
So next up is complete
InVasManiSo basically in summary a machine learning physically compressed chip engineered inference artificial intelligence designing ecosystem technology habitat. It's alive!!!
When it is able to manufacture itself in better form without impeded ability I will worry
Posted on Reply
#11
InVasMani
You mean like in Battlestar Galactica!!? Spoiler alert!
Posted on Reply
#12
lexluthermiester
DeathtoGnomes..and we are another step closer to Skynet.
While I love humor statements like this, we are decades away from being "SkyNet" capable. Add to that, many governments and companies are VERY weary of allowing AI too much power and control..
Posted on Reply
#13
kiriakost
lexluthermiesterWhile I love humor statements like this, we are decades away from being "SkyNet" capable. Add to that, many governments and companies are VERY weary of allowing AI too much power and control..
Within EU, every one now think AI development as opportunity of working from home and make a fortune.
What the buyer will do the AI tools at hand, no one cares.
And therefore if something can go wrong, then it will.
Posted on Reply
#14
DeathtoGnomes
lexluthermiesterWhile I love humor statements like this, we are decades away from being "SkyNet" capable. Add to that, many governments and companies are VERY weary of allowing AI too much power and control..
Not decades. I used to keep up with AI progress, it may have changed since, but many researchers are skeptical such a thing will happen, others fear it so much they hesitate to go further. A lot of early AI research (algorithms) you see now is already used and abused, like when twitter and facebook flag or take down a post in almost real time, well pretty fast time-wise. There are plenty examples of autonomous robots, the Roomba is most obvious. I cant seem to find the video about the one AI computer, from about a decade ago, that passed the Turing test and already seems self aware. So when tech makes advancements that contributes to the AI field, yea I see that as a step in that direction. So why not apply a little humor to diffuse the implications?

I'll be back.:p
Posted on Reply
#15
looniam
DeathtoGnomesI'll be back.:p
hasta la vista baby.
Posted on Reply
#17
Jism
InVasManiSo basically in summary a machine learning physically compressed chip engineered inference artificial intelligence designing ecosystem technology habitat. It's alive!!!
You can click your chip online these days with all required functions, performance and power target and have it straight assembled from the line shipped to your uniform AM9 socket.

Woud'nt that be fantastic?
Posted on Reply
#18
defaultluser
We are decades away from Skynet...but eveery big leap like this is stil one frightening step closer!
Posted on Reply
#19
lexluthermiester
DeathtoGnomesWell now, apparantly there was one robot, in Germany, that became self-aware.

www.roboticsbusinessreview.com/rbr/robot_achieves_self_awareness_is_quickly_turned_off_and_dies/
Sorry. Utter nonsense. Saying that robot became "self-aware" would be like saying an ant became self-aware. It is not possible. The computing power for such complex algorithmic constructs does not exist yet and will not for at least 2 decades given how advancements have slowed down. Moores Law is currently broken. Whatever happened with that robot was not consciousness.

EDIT;
If you note the date of publish on that article, you will find the correct answer... Didn't see it until I read through it a bit..
Posted on Reply
#20
zlobby
kiriakostGoogle, failed be a quality search engine, failed at YouTube as administrator, failed as email hosting service.
Failed
in its and every sector of engagement with real people.
Who knows, they might succeed in sectors that there is no people there.
Do you think they removed theit famous 'Don't be evil' accidently?
Posted on Reply
#21
TheoneandonlyMrK
kiriakostGoogle, failed be a quality search engine, failed at YouTube as administrator, failed as email hosting service.
Failed
in its and every sector of engagement with real people.
Who knows, they might succeed in sectors that there is no people there.
Yet again you Failed to be correct.

And millions of searchers, YouTube watchers and email users prove.

Oh and Fff Google, I agree , they're monopolist shit and data mining sucj balls but they are successful.
Posted on Reply
#22
Frank_100
Chrispy_I'd laugh if Google rebranded itself to Cyberdyne Systems for April 1st this year.
I believe Cyberdyne is a spinoff of Acme subsidiary Yoyodyne.

One of the chief competitors of Weyland-Yutani and Tyrell Corp.
Posted on Reply
#23
DeathtoGnomes
lexluthermiesterSorry. Utter nonsense. Saying that robot became "self-aware" would be like saying an ant became self-aware. It is not possible. The computing power for such complex algorithmic constructs does not exist yet and will not for at least 2 decades given how advancements have slowed down. Moores Law is currently broken. Whatever happened with that robot was not consciousness.

EDIT;
If you note the date of publish on that article, you will find the correct answer... Didn't see it until I read through it a bit..
I'd wager we are closer than 2 decades to seeing a self aware AI, only to have the advancements slowed down, or even stopped, due to fear of success. If Elon Musk is against further development there must be something wrong! :kookoo::rolleyes: I wont lie, I am a bit fearful too. In 2 decades, the robot body, android by then, will have also advanced enough that, along with the AI, will be almost indistinguishable from a real human.

Yea I didnt see this at the very bottom either... LOL..
(Happy April Fool’s Day! In fact, the video was from an experiment to see how long humans would hesitate in this situation.)
Posted on Reply
#24
TheoneandonlyMrK
DeathtoGnomesI'd wager we are closer than 2 decades to seeing a self aware AI, only to have the advancements slowed down, or even stopped, due to fear of success. If Elon Musk is against further development there must be something wrong! :kookoo::rolleyes: I wont lie, I am a bit fearful too. In 2 decades, the robot body, android by then, will have also advanced enough that, along with the AI, will be almost indistinguishable from a real human.

Yea I didnt see this at the very bottom either... LOL..
To be fair, going on past history I might have rage attacked the ai computer sooner not just turned it off, I am not into this extreme case of the computer says No.
And a pc has lost its life to my right foot before, and could again. :D :p :)
Posted on Reply
#25
zlobby
Frank_100I believe Cyberdyne is a spinoff of Acme subsidiary Yoyodyne.

One of the chief competitors of Weyland-Yutani and Tyrell Corp.
You forgot Aperture among others but top kek anyway! :D

If we don't die in some global war or some cosmic catastrophy, the next pretendent for global human extinction is AI.

The moment AI gain the ability to replicate and sustain itself outside humans, the most logical thing for it would be that humans are its biggest threat. And that ain't some scifi...
Posted on Reply
Add your own comment
Nov 21st, 2024 09:55 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts