From IEEE Spectrum:
Hoping to speed AI and neuromorphic computing and cut down on power consumption, startups, scientists, and established chip companies have all been looking to do more computing in memory rather than in a processor's computing core. Memristors and other nonvolatile memory seem to lend themselves to the task particularly well. However, most demonstrations of in-memory computing have been in standalone accelerator chips that either are built for a particular type of AI problem or that need the off-chip resources of a separate processor in order to operate. University of Michigan engineers are claiming the first memristor-based programmable computer for AI that can work on all its own.
(Score: 2, Insightful) by nitehawk214 on Wednesday July 31 2019, @08:41PM (8 children)
Why so we need to put a buzzword like AI into this to make people interested in it. Memristors [wikipedia.org] are pretty fantastic on their own. They are basically better than transistors in most ways.
"Don't you ever miss the days when you used to be nostalgic?" -Loiosh
(Score: 5, Informative) by KilroySmith on Wednesday July 31 2019, @09:00PM (5 children)
Because a memristor is an inherently analog device that would make an excellent synapse between neurons in a neural network ASIC. The resistance of the memristor would correspond with the strength of the connection.
(Score: 2) by ikanreed on Wednesday July 31 2019, @09:14PM (4 children)
Yeah, and the number of non-AI uses for any kinds of analog circuits in computing is actually pretty small.
Outside computing, I dunno, I'd have to ask an EE where a memrister would have applications. Maybe something to do with tunable circuits?
(Score: 2, Insightful) by Anonymous Coward on Wednesday July 31 2019, @10:31PM (2 children)
Betcha we could sell 'em to audiophiles. They'll buy anything.
(Score: 3, Funny) by ikanreed on Wednesday July 31 2019, @11:03PM (1 child)
My coworker paid $500 for an insolated USB cable that protects against noise.
(Score: 2) by RS3 on Thursday August 01 2019, @02:47AM
Was it "low oxygen" too? https://en.wikipedia.org/wiki/Oxygen-free_copper/ [wikipedia.org]
(Score: 2) by RS3 on Thursday August 01 2019, @03:05AM
Okay, I'll give it a shot. [trundles off to read wiki...] I have no clue. Wacky thing. Electrically reminds me of magnetic core memory- you need some fancy circuity to write and read, and by reading you change its state, so you have to put it back (re-write). But it's not something you can just look at and understand all the possibilities. If I was doing that kind of work, it'd be different. Hopefully some other EEs will weigh in.
(Score: 2) by corey on Wednesday July 31 2019, @10:05PM
Because it's an announcement by researchers, who always need to spruik their work, adding some fad buzzwords always helps for those who don't understand the research; "sounds good!"
(Score: 4, Informative) by takyon on Wednesday July 31 2019, @10:43PM
This has been a thing for many years now:
https://www.engadget.com/2009/07/14/are-memristors-the-future-of-artifical-intelligence-darpa-think/ [engadget.com]
https://spectrum.ieee.org/robotics/artificial-intelligence/moneta-a-mind-made-from-memristors [ieee.org]
https://www.nextplatform.com/2015/07/08/memristors-mimic-brains-for-massive-machine-learning/ [nextplatform.com]
https://www.kurzweilai.net/synapse-like-memristor-based-electronic-device-detects-brain-spikes-in-real-time [kurzweilai.net]
http://sliu17.mysite.syr.edu/Memristor_CSM_10152017v2.pdf [syr.edu]
2009 was long before the latest AI craze. And you can probably find older references than that.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]