You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Intel has released an AI self learning chip based on 14nm. It will be interesting to see if Intel has really come up with something new or is just entering the race, only time will tell. No real performance specs yet and who will be writing the software. This is just another entry into the AI/ML race and the next 18 months we should be seeing some major advancements from a number of companies in AI/ML. Not only is the race on to build these systems, but the real winner is the company that develops and applies real world marketable applications. The AI/ML race is not just about the chip, but who can establish dominant platforms that take over previous manual real world functions from robotics in everything to self guided transport in sea, land and air. These areas will just be the starting point. Another serious area of development should be training/education. Medical though I feel will be the biggest area if it can handle everything from diagnostics to robotic surgery. With the "Great Acceleration" I feel we will see significant progress and actual applications applied within five years and the world as we used to know it will be gone.
The world will belong to the individuals and organizations that have and can apply the creativity, flexibility, resources and intellect to this world that has more and more variables as technology broadens and expands new and existing resources and what we can do with them. Never have we had so many options and opportunities before us.
Very early days here for Intel with the Loihi "test chip" targeted for use by Universities and Researchers later in 2018. Just a pre-announcement of something of interest only to University PhD's to start out with. Yes, time will tell if this is a breakthrough or not.
Daniel, Thank you for the comparisons and the perspective they give. The one advantage of man made AI/ML is focus, which I believe is a very serious consideration. I would appreciate your views on this for I highly respect your thoughtful input across a wide variety of technology.
I've been hearing about AI since the 1970's and the only thing that lives up to the hype has been SciFi movies like 2001: A Space Odyssey. The Japanese started and failed on their grand "5th Generation Project" to create general purpose AI.
I continue to see many narrowly defined, domain specific applications of AI, but the general purpose AI is still a twinkle in the eye of researchers.
Forcing an electronic system which is primarily digital to mimic neural activity is a worthy goal, however our brains use complex chemical and electrical technology which is inherently analog in nature.
You do realize this appears to be Intel's copy of IBM's True North?
In true Intel fashion, their inferior chip, a year or two later, is presented as a grand unprecedented breakthrough...
Here's IBM, describing not just the hardware but also what it can ACTUALLY do (as opposed to Intel's empty-headed claims of how it will bring about the rapture):
You do realize this appears to be Intel's copy of IBM's True North?
In true Intel fashion, their inferior chip, a year or two later, is presented as a grand unprecedented breakthrough...
Here's IBM, describing not just the hardware but also what it can ACTUALLY do (as opposed to Intel's empty-headed claims of how it will bring about the rapture):
My thought exactly. Furthermore there's a lot of skepticism in the AI/ML community regarding time-based "spiking" hardware (see Yann LeCun's thoughts on TrueNorth), but it sure makes for some sexy sounding press coverage. Case of Intel marketeers copying IBM marketeers, who are in fact the best tech marketeers in the world.
Very early days here for Intel with the Loihi "test chip" targeted for use by Universities and Researchers later in 2018. Just a pre-announcement of something of interest only to University PhD's to start out with. Yes, time will tell if this is a breakthrough or not.
Anyone from Intel on this forum can tell how can I as a university researcher register for early access of Loihi.
Does Intel have any equivalent emulator for researchers to start working from now on instead of waiting till 2018 to get access to actual chip.
I actually feel ML is here and it has already made serious and material advances possible in many areas, among them the early detection of skin cancers. AI doesn't need the connections of a human brain for the focus will be extremely tight compared to the wandering mind.
