4 ms·

Real Time Machine Learning (2019) [pdf]

156 points by apsec112 6y ago

aspenmayer 6y ago
I did a deep-dive into RTML on Twitter[0][1] just now after seeing this post, and I hope to bring some context and clarity to this project, which is done in conjunction with the National Science Foundation (NSF). @DARPA Twitter thread [2019] on RTML https://twitter.com/darpa/status/1108760215611236353 https://twitter.com/darpa/status/1108760215611236353 Report in OP’s linked PDF authored by Andreas Olofsson https://d60.darpa.mil/speakers/MrAndreasOlofsson.html https://d60.darpa.mil/speakers/MrAndreasOlofsson.html DARPA RTML program overseen by Serge Leef https://www.darpa.mil/staff/mr-serge-leef https://www.darpa.mil/staff/mr-serge-leef Electronics Resurgence Initiative (ERI) Summit 2019 talk on RTML https://www.youtube.com/watch?v=zQfddARnNB0 https://www.youtube.com/watch?v=zQfddARnNB0 ERI Summit site https://eri-summit.darpa.mil/ https://eri-summit.darpa.mil/ ERI @ DARPA https://www.darpa.mil/work-with-us/electronics-resurgence-initiative https://www.darpa.mil/work-with-us/electronics-resurgence-in... National Science Foundation (NSF) RTML program https://www.nsf.gov/funding/pgm_summ.jsp?pims_id=505640 https://www.nsf.gov/funding/pgm_summ.jsp?pims_id=505640 My unrolled Twitter thread, for convenience [0] https://threadreaderapp.com/thread/1292268278958694402.html https://threadreaderapp.com/thread/1292268278958694402.html Direct link to my Twitter thread [1] https://twitter.com/aspenmayer/status/1292268278958694402 https://twitter.com/aspenmayer/status/1292268278958694402 Edit: This late edit was informed by additional info about those awarded grants from DARPA to study RTML, from HN user WillSlim95. https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-funded-project-to-build-open-source-hardware-generators/ https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-fun... In brief, Professor Sachin Sapatnekar (ACM Fellow [2016], University of Minnesota, College of Engineering, Department of Electrical and Computer Engineering) has been awarded a $2.2M grant from DARPA, “to build open-source hardware generators for a range of machine learning algorithms that process data in real time.” His faculty page https://ece.umn.edu/directory/sapatnekar-sachin/ https://ece.umn.edu/directory/sapatnekar-sachin/ Quirky OG Web 1.0 faculty page https://people.ece.umn.edu/users/sachin/ https://people.ece.umn.edu/users/sachin/
- jmeister 6y ago
  Awesome, thank you for the details!
- smangold 6y ago
  Thank you for sharing all the relevant information!
pkage 6y ago
Is there any information on who competed for (and ultimately won) the contract?
- WillSlim95 6y ago
  Sachin Sapatnekar from University of Minnesota Twin Cities. https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-funded-project-to-build-open-source-hardware-generators/ https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-fun...
canada_dry 6y ago
> program seeks to create no-human-in-the-loop hardware generators and compilers to enable fully automated creation of ML Application-Specific Integrated Circuits (ASICs) Of course it's a fairly obvious use of ML/AI. There are just so many fascinating ways this can play out in the future - it's going to be an interesting ride! Note also, Olofsson is the founder of Adapteva [i]. [i] https://en.wikipedia.org/wiki/Adapteva https://en.wikipedia.org/wiki/Adapteva
person_of_color 6y ago
The author of these slides is the creator of Parallela.
- axegon_ 6y ago
  Interesting, I hadn't heard the Parallela name in years. I remember them making some utterly unrealistic promises about close to triple digit cpu core sbc's back in the first half of the 2010's and that was that.
- dehrmann 6y ago
  I remember when I saw Parallela, I thought it might have interesting applications for low-cost, low-power, embedded ML applications.
nn3 6y ago
The power/performance spectrum seems to be very wide: 200uW for 400Kbps throughput with 0.6 accuracy 200W for 400Gps with higher accuracy with 0.99 accuracy I guess will be interesting where in this range the result (if any) will fall.
- messe 6y ago
  It’d be nice if they implemented TCP over that 400Kbps w/ 60% accuracy. Would be a nice projection of what sort of loss and efficiency should be expected.
- jagtesh 6y ago
  I guess they want silicon for different uses. Embedded with low power consumption (on a soldier/handheld for example), in a vehicle, in a datacenter, etc. With each situation, your need for precision can vary, and along with it the requirement for processing power. A model tuned for 75% accuracy is lighter and faster to run (fewer layers) vs 85% (several times bigger, slower). I've made up the numbers to illustrate a point. This is anecdotal and I'm not an ML expert. But here's some evidence: 1. https://ai.googleblog.com/2019/03/an-all-neural-on-device-speech.html?m=1 https://ai.googleblog.com/2019/03/an-all-neural-on-device-sp... 2. https://ai.googleblog.com/2018/05/custom-on-device-ml-models.html?m=1 https://ai.googleblog.com/2018/05/custom-on-device-ml-models... 3. https://krisp.ai/blog/how-we-shrunk-dnn-to-run-inside-chrome/ https://krisp.ai/blog/how-we-shrunk-dnn-to-run-inside-chrome...
jarym 6y ago
Only related to ‘low-power’ AI I recently came across Ogma[1] And though the combination of online learning and something that works on a Raspberry Pi was fascinating. Surprised it’s not appeared on here before [1] https://ogma.ai/category/ogmaneo/ https://ogma.ai/category/ogmaneo/
apsdsm 6y ago
Despite myself I’m disappointed this wasn’t about time machines.
christofon 6y ago
I really was hoping this lets me learn about real time machines...
godelmachine 6y ago
Has Andreas Olofsson left Adapteva for greener pastures? His LinkedIN profile makes an impression he has moved on. Does anyone get to be a DARPA Program Manager while being a CEO of another company?
- WillSlim95 6y ago
  He left Adapteva to become a PM at DARPA and has now moved on from DARPA.

156 points by apsec112 6y ago

I did a deep-dive into RTML on Twitter[0][1] just now after seeing this post, and I hope to bring some context and clarity to this project, which is done in conjunction with the National Science Foundation (NSF). @DARPA Twitter thread [2019] on RTML https://twitter.com/darpa/status/1108760215611236353 https://twitter.com/darpa/status/1108760215611236353 Report in OP’s linked PDF authored by Andreas Olofsson https://d60.darpa.mil/speakers/MrAndreasOlofsson.html https://d60.darpa.mil/speakers/MrAndreasOlofsson.html DARPA RTML program overseen by Serge Leef https://www.darpa.mil/staff/mr-serge-leef https://www.darpa.mil/staff/mr-serge-leef Electronics Resurgence Initiative (ERI) Summit 2019 talk on RTML https://www.youtube.com/watch?v=zQfddARnNB0 https://www.youtube.com/watch?v=zQfddARnNB0 ERI Summit site https://eri-summit.darpa.mil/ https://eri-summit.darpa.mil/ ERI @ DARPA https://www.darpa.mil/work-with-us/electronics-resurgence-initiative https://www.darpa.mil/work-with-us/electronics-resurgence-in... National Science Foundation (NSF) RTML program https://www.nsf.gov/funding/pgm_summ.jsp?pims_id=505640 https://www.nsf.gov/funding/pgm_summ.jsp?pims_id=505640 My unrolled Twitter thread, for convenience [0] https://threadreaderapp.com/thread/1292268278958694402.html https://threadreaderapp.com/thread/1292268278958694402.html Direct link to my Twitter thread [1] https://twitter.com/aspenmayer/status/1292268278958694402 https://twitter.com/aspenmayer/status/1292268278958694402 Edit: This late edit was informed by additional info about those awarded grants from DARPA to study RTML, from HN user WillSlim95. https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-funded-project-to-build-open-source-hardware-generators/ https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-fun... In brief, Professor Sachin Sapatnekar (ACM Fellow [2016], University of Minnesota, College of Engineering, Department of Electrical and Computer Engineering) has been awarded a $2.2M grant from DARPA, “to build open-source hardware generators for a range of machine learning algorithms that process data in real time.” His faculty page https://ece.umn.edu/directory/sapatnekar-sachin/ https://ece.umn.edu/directory/sapatnekar-sachin/ Quirky OG Web 1.0 faculty page https://people.ece.umn.edu/users/sachin/ https://people.ece.umn.edu/users/sachin/

jmeister 6y ago
Awesome, thank you for the details!
smangold 6y ago
Thank you for sharing all the relevant information!

pkage 6y ago

Is there any information on who competed for (and ultimately won) the contract?

WillSlim95 6y ago
Sachin Sapatnekar from University of Minnesota Twin Cities. https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-funded-project-to-build-open-source-hardware-generators/ https://ece.umn.edu/prof-sachin-sapatnekar-to-lead-darpa-fun...

canada_dry 6y ago

> program seeks to create no-human-in-the-loop hardware generators and compilers to enable fully automated creation of ML Application-Specific Integrated Circuits (ASICs) Of course it's a fairly obvious use of ML/AI. There are just so many fascinating ways this can play out in the future - it's going to be an interesting ride! Note also, Olofsson is the founder of Adapteva [i]. [i] https://en.wikipedia.org/wiki/Adapteva https://en.wikipedia.org/wiki/Adapteva

person_of_color 6y ago

The author of these slides is the creator of Parallela.

axegon_ 6y ago
Interesting, I hadn't heard the Parallela name in years. I remember them making some utterly unrealistic promises about close to triple digit cpu core sbc's back in the first half of the 2010's and that was that.
dehrmann 6y ago
I remember when I saw Parallela, I thought it might have interesting applications for low-cost, low-power, embedded ML applications.

nn3 6y ago

The power/performance spectrum seems to be very wide: 200uW for 400Kbps throughput with 0.6 accuracy 200W for 400Gps with higher accuracy with 0.99 accuracy I guess will be interesting where in this range the result (if any) will fall.

messe 6y ago
It’d be nice if they implemented TCP over that 400Kbps w/ 60% accuracy. Would be a nice projection of what sort of loss and efficiency should be expected.
jagtesh 6y ago
I guess they want silicon for different uses. Embedded with low power consumption (on a soldier/handheld for example), in a vehicle, in a datacenter, etc. With each situation, your need for precision can vary, and along with it the requirement for processing power. A model tuned for 75% accuracy is lighter and faster to run (fewer layers) vs 85% (several times bigger, slower). I've made up the numbers to illustrate a point. This is anecdotal and I'm not an ML expert. But here's some evidence: 1. https://ai.googleblog.com/2019/03/an-all-neural-on-device-speech.html?m=1 https://ai.googleblog.com/2019/03/an-all-neural-on-device-sp... 2. https://ai.googleblog.com/2018/05/custom-on-device-ml-models.html?m=1 https://ai.googleblog.com/2018/05/custom-on-device-ml-models... 3. https://krisp.ai/blog/how-we-shrunk-dnn-to-run-inside-chrome/ https://krisp.ai/blog/how-we-shrunk-dnn-to-run-inside-chrome...

jarym 6y ago

Only related to ‘low-power’ AI I recently came across Ogma[1] And though the combination of online learning and something that works on a Raspberry Pi was fascinating. Surprised it’s not appeared on here before [1] https://ogma.ai/category/ogmaneo/ https://ogma.ai/category/ogmaneo/

apsdsm 6y ago

Despite myself I’m disappointed this wasn’t about time machines.

christofon 6y ago

I really was hoping this lets me learn about real time machines...

godelmachine 6y ago

Has Andreas Olofsson left Adapteva for greener pastures? His LinkedIN profile makes an impression he has moved on. Does anyone get to be a DARPA Program Manager while being a CEO of another company?

WillSlim95 6y ago
He left Adapteva to become a PM at DARPA and has now moved on from DARPA.