Microsoft is proud of its work on AI, and eager to convey the sense that this time around, it’s poised to win. In June, it invited me to its campus to interview some of Nadella’s top lieutenants, who are building AI into every corner of the company’s business. Over the next two days, Microsoft showed me a wide range of applications for its advancements in natural language processing and machine learning.
The company, as ever, talks a big game. Microsoft’s historical instincts about where technology is going have been spot-on. But the company has a record of dropping the ball when it comes to acting on that instinct. It saw the promise in smartphones and tablets, for example, long before its peers. But Apple and Google beat Microsoft anyway. The question looming over the company’s efforts around AI is simple:
I know we’re just at the very beginning of this whole thing, but so far, I’m not particularly impressed with the fruits of all this AI work for us as end users. Things like Cortana and Siri generally just offer more cumbersome ways of doing something achieved quicker with other methods, and they demonstrate little to no “intelligence”. Knowing I have a translation deadline at 15:00 and reminding me of it is not really intelligence; it’s just a talkative alarm with an annoying attitude.
Much like VR, this just needs way, way more technological progress and breakthroughs to really be what its name implies.
What now? Does that refer to how they realized the importance of the internet after everybody else and then tried to make their own?
Or perhaps about how voice communication would be the primary way to communicate with computer in only five years… fifteen years ago or so?
How everybody would love to have the desktop experience on their smartphone, and then how everyone would want a tablet GUI for their desktop?
“spot on” is stretching it a bit, I’d say.
Steve Ballmer’s comments at the launch of the iPhone are a pretty glaring piece of evidence of Microsoft’s lack of skill in predicting where technology is going, and the subsequent total failure of Windows Mobile just adds to that pile of evidence.
VR is nowhere near ready for everyday “household” use. The mental effects are one aspect that will have to be looked at, some people just aren’t able to function normally after being exposed to VR for long stretches of time, that’s just one problem.
AI and such functions as Siri, Amazon Echo, Cortana, Hound, and the like are far more promising in my opinion.
Edited 2016-07-08 11:29 UTC
Doesn’t that Prove the point?
They saw the internet was the future of technological growth so tried to corner the market. In this case they didn’t succeed, but they were there, right at the forefront. They had TCP/IP support in 1984(with IBM) while Apple took until 1988.
“Trumpet Winsock” is what I used for TCP/IP in windows up til Windows95. So I guess they didn’t anticipate the need for the internet on consumer desktops until around that time.
Ugh, I remember having to configure the third-party TCP/IP stacks. Damned glad not to have to do that anymore!
Yes, they did anticipate the need for the internet on consumer desktops.. with Windows 95.
But 5 years before that they had released TCP/IP as part of OS2 1.2
So they were at the forefront of the R+D. Even after they released 95 the market was still only 0.4% which windows (and IE) then drove the expansion of
http://www.internetworldstats.com/emarketing.htm
I think maybe you and me just have different views on what “anticipate” is.
These are definitely artificial intelligence. The label is accurate – it’s not actually intelligent, but more simulates intelligence.
What I think Thom is looking for is real machine intelligence. That’s where we’ll get some useful tech.
And what’s really frustrating is that even for power users, these companies are taking control away and forcing this AI shit down our throats LONG before it is ready for prime time.
I mean, some of it is okay, like telling Siri or Google Now to set an alarm, but in news aggregators, for example, I don’t like that I can’t specify in any sort of detail what sort of articles I’d like to see. Instead, I have to rely on some algorithm that I have no control over, and which doesn’t really do shit the way I want anyway.
Amen.
Sometimes, I’ll have to write my own RSS scrapers to get what I want, which’ll just piss me off more because they’re forcing me to waste my time working around something they claim is supposed to SAVE me time.
LOL, I’m actually writing my own in Python, that has an Inbox-style rules wizard.
I find it easier to just use raw LXML API calls to define the scraper rules since I’m the only person intended to use it.
I completely disagree on the terminology. IM(NS)HO, “artificial intelligence” and “machine intelligence” are basically synonyms – actual intelligence in an artificial (i.e. manufactured) device.
And I completely agree with Thom that it is a misnomer to call any of the products available today “AI”. If the marketers win this battle of definitions, then it’s just another perfectly good phrase/description down the drain.
The definition has long shifted: what people think of AI is in fact “general AI”, and the AI we’re all being peddled are just quite stupid algorithms for very narrow tasks.
Microsoft can’t make so much as a Twitter bot without it turning into cyber-Hitler.
“Knowing I have a translation deadline at 15:00 and reminding me of it is not really intelligence; it’s just a talkative alarm with an annoying attitude. ”
Get the annoying, Thom.
On the other side of things: Would you prefer Cortana to start learning your translation style. Preparing the translation Work Books for you, send a preventive note to your Client, justifying your delay. Present you with a default translation according with what She has learned from your style. And prepare a product document as soon as you approve? Maybe sending pre-product for revision also?
The real question is: at what point is what Thom does replaced by computers ? 😉
I gotta disagree when it comes to VR needing more progress before it is what its name implies. Could VR be better? Yes. Can a good experience in a VR headset make you feel like you’re really there? Without a doubt. The HTC Vive (and I assume the Oculus Rift as well) has some amazing experiences available.
VR “does” need a lot more progress before it can be named as such.
VR does not and will not exist until holographic projectors are available and mainstream.
That will be VR, not this crap 2D stuff that pretends to be and needs a bucket strapped to your face!
Well, things will get worse before they get better. And that because the meaning of “AI” has been drastically diluted over the years. Cortana, Siri, and basically everything we have today that’s grouped under the AI term are not AI. They are artificial, yes, but we have nothing “intelligent” in the proper sense of the word. All we have are some very basic methods and solutions that can form the beginnings of AI. We have nothing that understands, infers, connects, creates, debates, evolves, transfers knowledge, and really nothing cognitive. Yes, we have learning algorithms that are getting pretty good for specific constrained tasks (e.g., playing a game, categorizing images/signals, etc.), but that’s it. There has been tremendous work put into it, but we still have a very long way to go. All things considered, Cortana, Siri et al. are just enhanced chatbots, but even that may be an overstatement.
As I understand it, people use the terms “Strong AI” or “Artificial General Intelligence” to cover what AI used to mean.
I’m sorry to say it, but that just sucks.
I fully understand if terms get used for outsiders, but I’m always surprised the people in a that field also need to change their terms.
“Dear aunt, let’s set so double the killer delete select all”
“We have nothing that understands, infers, connects, creates, debates, evolves, transfers knowledge, and really nothing cognitive.”
Last thing Huge AI will want to do is to debate or teach Us. Simply because there will be no logic on following Us. Yes, true AI is volitive.
Journeyman AI today is almost entirely about understanding the local environment through vision and voice. Both things that computer at the moment are pretty much spectacularly bad at.
Autonomous cars is a vision problem, not a control problem. We’re already pretty good at control. We can turn the cars, it’s the avoiding everything in the street that’s the big problem.
Computers are still poor at grokking “what they see”. They’re almost to the point where they can reliably identify shapes.
Similarly with voice. They’re just getting the basics of voice recognition down. Computers can suss out words, but understanding sentences and multiple sentences with implied context is the challenge today. Ye Olde “Time flies like an arrow” “Fruit flies like a banana” problem.
Once those fundamentals get better, once the computers can better grok “what we mean” from “what we say” (heck, they have problems grokking what we mean from typed text), then “AI” will explode and start becoming pervasive.
But we’re still in the 33K modem era of the internet in terms of AI.
but I do know one thing though.
I will never trust an AI system run by Microsoft, Google, Facebook etc more than I could throw their HQ buildings.
For them AI means a way to gather more accurate data on us mere plebs. We are the product to be marketed at.
Please, stop the world, I wanna get off.
{Well, I’ll probably be pushing up the weeds before we get truly viable AI’s.}