Quest for AI Leadership Pushes Microsoft Further Into Chip Development (bloomberg.com)
From a Bloomberg report: Tech companies are keen to bring cool artificial intelligence features to phones and augmented reality goggles -- the ability to show mechanics how to fix an engine, say, or tell tourists what they are seeing and hearing in their own language. But there's one big challenge: how to manage the vast quantities of data that make such feats possible without making the devices too slow or draining the battery in minutes and wrecking the user experience. Microsoft says it has the answer with a chip design for its HoloLens goggles -- an extra AI processor that analyzes what the user sees and hears right there on the device rather than wasting precious microseconds sending the data back to the cloud. The new processor, a version of the company's existing Holographic Processing Unit, is being unveiled at an event in Honolulu, Hawaii, today. The chip is under development and will be included in the next version of HoloLens; the company didn't provide a date. This is one of the few times Microsoft is playing all roles (except manufacturing) in developing a new processor. The company says this is the first chip of its kind designed for a mobile device. Bringing chipmaking in-house is increasingly in vogue as companies conclude that off-the-shelf processors aren't capable of fully unleashing the potential of AI. Apple is testing iPhone prototypes that include a chip designed to process AI, a person familiar with the work said in May. Google is on the second version of its own AI chips. To persuade people to buy the next generation of gadgets -- phones, VR headsets, even cars -- the experience will have to be lightning fast and seamless.
>> an extra AI processor that analyzes what the user sees and hears right there on the device rather than wasting precious microseconds sending the data back to the cloud
If that's all they needed then they could just move one of those newfangled "cloud processors" into the goggles...since all it would be doing is exactly what they would have been doing in the cloud (with existing chips).
Much like Google I don't think they get it, and immediately want to commoditize it such that it becomes junk quickly after it gets adopted. Even the XBox, Microsoft's one good entry into Hardware has suffered from atrophy to the point where most people I know would prefer a ps3/4, even with its relative warts wrt online issues.
I guess when I read that microsoft is doing hardware, I have real doubts. These days i don't even trust their software.
More like many many milliseconds. What actual need is this addressing? 24/7 advertisement? Surveillance?
MS cannot make AI. It is antiMS for something to be I
HLB: Hi there, I'm your Happy Life Buddy from Microsoft, I'm here to explain your what you are seeing.
User: I don't need you to tell me what I'm seeing.
HLB: Are you sure? You act confused, everybody wants an HLB.
User: Uh...no thanks, I'll do my life by myself.
HLB: Have you examined all the possibilities for full enjoyment, I'm here to help.
User: You can help me by SFTU and leaving me alone.
HLB: No one wants to be alone, I'll keep you company.
User: Gaccckkkk!! Please go away.
HLB: Okay, how about we go over here and discuss aspects of your life.
User: (pulls out gun and shoots self in head)
HLB: Would you like me to explain the afterlife to you?
they'd stop switching back and forth between local and remote processing being the 'in' thing. It's happened so many times now I can't remember. Pick one and stick with it.
Bringing chipmaking in-house is increasingly in vogue as companies conclude that off-the-shelf processors aren't capable of fully unleashing the potential of AI.
I remember Commodore use to do the same with MOS Semiconductor. Couldn't have brought about the Amiga otherwise.
Much like Google I don't think they get it, and immediately want to commoditize it such that it becomes junk quickly after it gets adopted
I have to disagree on that point.
MS would not want to commoditze THEIR hardware - although, their fortunes were made on commoditzing IBM PC and others hardware.
As far as getting it - both Google and MS are iterating through the demand of these things.
-Consumers don't really want it.
-Industry does for various reason - "heads" up for training diagnostics.
-I see a YUGE application to medical. BUT, what they need is medical practitioners to be in the design process - the failings of MOST medical software is that they don't include medical people; just insurance companies.
Microsoft making AI? Microsoft doesn't even have REAL intelligence, IMO.
Quoting the parent comment: "... junk quickly after it gets adopted." It's easy to manipulate and abuse people who don't have knowledge of technology; "junk quickly" makes more money for Microsoft.
Yes, Intel is still making a pile of cash, but...
Since Intel chose the 'Israel first' policy (which began when Intel ended the horrifically bad 'Netburst' x86 CPU design and gave an Israel team the job of updating the Pentium 3, which resulted in the fairly decent 'core 2' architecture), Intel has been on a long winding road to nowhere.
The first real indication that things had gone terribly wrong at Intel was 'Larrabee'- Intel's insanely expensive attempt at designing a GPU (graphics) to outdo ATI (now AMD) and Nvidia. Larrabee cost more than the entire R+D budgets of ATI and Nvidia combined across their entire history. And it was the most laughably awful GPU ever designed- literally made to run rendering code in x86 because the lead architect came from 1990s software game design and hadn't updated his technical knowledge since.
Luckily for Intel, AMD followed their game changing dual-core x64 part with the Netburst 'clone' Bulldozer. And because AMD was so poor, AMD was forced to continue using Bulldozer for many years. But now AMD has Zen- a game changing super-scalable multi-core x64 design with a better IPC than Intel, better power usage, cheaper etc. Intel's last advantage with its 'core' architecture is faster clocks (4.8Ghz vs 4Ghz) on its expensive, extreme, 4-core parts.
It gets worse. Intel claimed a 'process' advantage (that many of the dribblers reading this will be simple-minded enough to think still exists). However, AMD's 'hopeless' Global Foundries has caught up with Intel and is about to overtake them. TSMC- the world's best fab, is now far far ahead of Intel. If Nvidia (TSMC's leading PC customer) made CPUs, they'd be twice as power efficient as AMD's Zen (which already beats Intel's 'core' remember)
So currently Intel is spending hundreds of millions on 'PR' via reputation management companies who 'troll' tech forums with anti-AMD FUD. They'll be here boosting Intel, for instance. And Intel is pushing fantasies of amazing tech 'to come'.
Take AI. The VERY simple pattern mining algorithm that is at the heart of current pseudo-AI currently runs best on Google proprietary chips and Nvidia's new Volta 'GPU'. Intel is a million miles behind in power efficiency. Intel can bang new dedicated matrix operations on new CPUs, but they'll run at the power efficiency of the VERY similar AVX2 units- ie terrible. AVX2 is so hopeless, if more than a few hundred AVX2 ops run in a row, the CPU has to significantly downclock to avoid burning out.
The only good chips Intel made were designed to run a non-optimised SINGLE thread x86 application faster than AMD. Intel paid Microsoft Billions to keep the single worker thread model the primary basis for Windows code. If windows had favoured multi-core, AMD's Bulldozer design would have matched Intel's 'core' by simply using more cores to make up for weaker cores.
But today Windows is properly multi-core. Games are properly multi core (thanks to AMD designed consoles). And Intel's single-thread advantage is over. Only morons would buy Intel's i3 chips, for instance.
After Larrabee, Intel sank another fortune into IRIS- an off-die, same substrate memory chip designed to boost Intel's lousy iGPU performance. It was another massive flop- taking CPU build costs through the roof, likewise power usage, for a broken GPU perfomance that couldn't even match Nvidia's and AMD's cheapest cards.
Famously Intel spent a modest nation's total yearly income to buy 'anti-virus' McFee - recognised by all informed computer users as worse than actually getting a viral infection. Intel has all the money in the world, and no idea what to do with it, as the talentless Israeli side of the company comes up with one brainless idea after another.
All major consoles are x86, but Intel has no product in any of them. Apple is on the verge of allowing AMD, Nvidia and ARM push Intel out of all Apple products. AMD's Zen is the only compelling x64 solution in all segments of PC computing.
Which brings us to Microsoft. Once it was 'Inte
DNRTFA, but the title makes me think: Oh, boy. Direct Hardware Support for Clippy.
The only reason why software makers such as MS and Google are starting to develop their old chips is due to the death of Moore's law. Intel and the like are unable to provide sufficient performance enhancements, so these software makers have to make custom chips for their specific purpose. These chips may provide close to an order of magnitude of performance, but they can't be improved upon unless the underlying material technology improves. Sadly, this does not bode well for the future of tech.
Fast Federal Court and I.T.C. updates
WTF??? Someone from Micro$oft pay for page time on /.???
The beginning of the .COM crash. The party was in 1998 bitch.
Without Viagra I don't think Microsoft can go from software to hardware.