Apple Hires Carnegie Mellon Researcher to Lead AI Team

[mi7chy]

To put things into perspective, Siri is kindergarten level back in 2011 and hasn't progressed while Google AI is beyond PhD level beating and teaching world ranked professional Chinese Go (more complex than chess) players and solving real world issues such as optimizing data center efficiency.

Siri's root issue is understanding basic words before it can tackle higher level awareness such as location and context.

https://www.reddit.com/r/apple/comments/56wk4m/i_challenge_all_of_you_to_try_and_get_siri_to/

Typical frustration of having a conversation with Siri:

Versus Google AI:

http://english.khan.co.kr/khan_art_view.html?code=710100&artid=201605031817377

 

[brianvictor7]

How long before Apple get bored, frustrated and/or disinterested and sack everyone in the team?

Won't happen. A.I. is the ultimate enabling benchmark for all of mankind's technological innovation. Once we have it figured out, then things can *really* get interesting. All the major tech players know this. If Apple doesn't make its own, it will have to lease it from someone.

 

[macmyworld]

Too little, too late....the competition is too far ahead.

Couldn't have said it better myself. Hiring a Phd this late in the game won't scare Amazon or Google one bit. I'm amazed that Apple gave Siri such little attention.

 

[truthertech]

Probably just a coincidence but it never ceases to amaze me how influential Mossberg is.

"Probably" just a coincidence??? So, you think Tim Cook might have read something on a blog and then made a phone call "Quick, Walt doesn't know anything about what we've got going on behind the scenes, but I want someone hired by the COB tomorrow to take over as director of AI.!"

 

[emm386]

While Microsoft, Amazon and Google had common sense in mind when they designed their digital assistants,

Apple requires highly regarded scientists and a dedicated research center.

 

[macs4nw]

Too little, too late....the competition is too far ahead.

Although partly true, it's never too late to make those needed Siri improvements, especially reliable speech recognition, without which any future Apple personal assistant such as the Amazon Eco, would never work satisfactorily.

A zillion additional machine-recognizable questions and commands wouldn't hurt either, another area where Siri currently is severely lagging behind, and of course a vast improvement in deep learning, the ability to correlate compound questions and arrive at an appropriate and correct response.

 

[Analog Kid]

It's really hard to know what's happening in Apple's labs because, unlike Google, Apple doesn't release stuff to the world that they don't think brings direct customer benefit. We've seen Siri, and we've seen some limited stuff in Photos, but they're not really pushing the cutting edge because the cutting edge tends to be unreliable for users.

And they're not going to spend time playing Go.

That said, I think Apple really does have some challenges ahead in staying alive in this space. Someone mentioned Apple's focus on individual privacy, and that's definitely hampering them a big-- Google and others have made moves toward better privacy, but they definitely see the line as more fluid than Apple has.

The big thing I think is a challenge though, is that Apple doesn't have a search engine. Without a that enormous information index, and the image and video archives that go with it, they just don't have as much information to train on and they don't have a mechanism to observe humans interacting with that data.

If the Verizon deal with Yahoo! falls through, I wonder if Apple should think about picking them up for that reason alone.

 

[Jakexb]

Once again, you guys are full of trollishness.

But having a legit, high-level researcher leading AI is a big deal, similar to Yann Lecun at Facebook. It means that Apple isn't just going to be a consumer of off-the-shelf AI technology, but is going to get into serious research itself. What that means for publishing papers and the like, we'll have to see. It seems unlikely that Apple would be able to recruit someone like this without a promise of some openness.

He's one of the very top deep learning guys out there. He could literally get a job anywhere he wanted.

 

[macUser2007]

Once again, you guys are full of trollishness.
...

Why "trollishness?"

Many are simply pointing out that Siri seems stuck in place, while Google, MS and even Amazon have moved the bar considerably higher.

I've been using Alexa/Echo for a year now, and while it has it's shortcomings, it's become a part of my routine, for music, podcast and news, as well as for answering random questions. I also use Alexa for home automation.

I have ordered Google Home, will see how it stacks up when it arrives.

Sadly, Apple is nowhere to be found in my home automation scheme, not because I don't like it, but because it has nothing competitive to offer.

Even more sadly, Apple has not offered anything exciting or innovative for years now.

So, yes, while it's great to see such a notable hire, there is a lot of truth in the "too little, too late" statement.

Apple's management seems to be afraid to take risks or innovate, focusing instead on short-term result, which while ensuring large bonuses for the very top executives and pats on the shoulder by the Board, will ultimately leave the company with mediocre mix of products.

Oh, and my new phone will not be an iPhone 7 Plus as I expected, but a Pixel XL. Google is just more exciting nowadays.

 

[farmboy]

I get the feeling that several "visible" executives shouldn't spend too much time figuring out where their new office will be in the Apple Space Ship.

The thing that Wayne Gretzky said about skating to where the puck will be...the key thing to remember is that HE was the guy flying down there with or without the puck most of the time. He led in dramatic fashion. Why can't Apple do this anymore?

And that is less about Siri (enough said) and more about why the management team and the high-priced talent they pay for repeatedly fail to bring superior products to market. Why is almost everything an iteration and not an innovation? Is Apple hiring the wrong people, or hamstringing those they do hire?

Apple Management meeting notes:
"Today's Topic: Are we being cautious enough?"

Where are all the "fantastic things" that were/are supposedly in the pipeline, per management statements?

 

[DarthMoops]

Hello HAL.

 

[Sasparilla]

Glad to see Apple pushing here...they intro'd a great technology (Siri), years ahead of everyone else and basically put it into maintenance mode a couple years later (one of Apple's weaknesses) & now they're way behind.

Probably just a coincidence but it never ceases to amaze me how influential Mossberg is.

He really is. It was also the Google Pixel introduction the other day, they really showed off the AI chops they've been working on since Apple hung the Mission Accomplished sign on Siri years ago. Mossberg put a spotlight on the differences for everyone to see (who didn't watch the Pixel intro) - which was really beneficial for Apple (so they'd wake up).

 

[theheadguy]

Everything that'll help Siri to become more clever is welcome! desperately needed!

fixed that for you.

 

[Bubba Satori]

Siri are you an Apple hobby?

Here are the directions to Hobby Lobby.

 

[weup togo]

I wouldn't say the chances are zero. There are already counter-examples of Apple exposing past advanced research in patent applications, and in open source contributions (ResearchKit and Swift, et.al.) Advancing Swift requires a ton of open collaboration with the outside community, and the same can be said of some of the current trends in machine learning and DNNs.

AI research is fundamentally different than any of these. Patents require exposing work, in exchange for an artificial monopoly on them. ResearchKit is an API, and Swift is a language Apple is advancing to increase its influence on the state of software. Both of those have to be open by definition.

The kind of R&D that goes into developing AI is pure secret sauce in Apple's view. This is by far the most intensely competitive space in software today. Giving away one's advances to competitors is absolutely the last thing Apple will do or tolerate.

 

[KALLT]

AI research is fundamentally different than any of these. Patents require exposing work, in exchange for an artificial monopoly on them. ResearchKit is an API, and Swift is a language Apple is advancing to increase its influence on the state of software. Both of those have to be open by definition.

The kind of R&D that goes into developing AI is pure secret sauce in Apple's view. This is by far the most intensely competitive space in software today. Giving away one's advances to competitors is absolutely the last thing Apple will do or tolerate.

What firewood was getting at is that the nature of the technology may require Apple to engage with external researchers and developers. There is precedent for this. Apple has done this with LLVM/Clang and Swift. These projects would not have been possible or would have been held back. AI is a huge field of research and not all leading researchers are willing to work in a closed environment and have their work claimed by a commercial entity. Not everything needs to be shared, but some of it has to be.

I'd be happy with Siri just knowing public information first before diving into my private information.

I strongly agree with this sentence. I think what disappoints me the most about Siri is that you cannot use it in ways that the average human being would expect. Apple likes to show what Siri can do, giving the impression that it is more organic than it actually is. I remember being thoroughly impressed with the sophistication of Mac OS X Leopard. How much of it made sense and how there was always a shortcut, a hidden menu or an extension point for Automator or AppleScript. It felt like Apple thought of everything. Siri just feels conspicuously limited and scripted.

It is the same story with Maps. I know for a fact that many public transport organisations and companies have well-documented APIs for transport information available. Yet Apple chooses to work slowly, city by city, seemingly randomly across the world. They have to crowd-source some of this work, Google recognised this years ago.

 

[usarioclave]

Apple has the opportunity to do one thing really well, which in general has been Apple's MO for a long time. I don't really care about a general-purpose assistant. I'd rather have one thing doing AI well.

I've used Alexa, and I've seen what Amazon's AI stuff can do for me, which is nothing. Same with google. I don't really need better recommendations, or more appropriate ads. I really don't care about that crap. Talking to a device to manage my house is fun, but it's a parlor trick.

Give me an AI that helps me to my job better. Have it keep track of what I've done, what I'm doing, and what I'm supposed to do.

I can wait for Apple or someone to do it right. I'll grant them permission to any data they need.

Really, Apple's not behind, because nobody is doing anything of value in the AI space right now. They're making talking dogs, which is neat. A talking dog is a lot different than a dog that has something meaningful to say.

 

[mi7chy]

Couldn't have said it better myself. Hiring a Phd this late in the game won't scare Amazon or Google one bit. I'm amazed that Apple gave Siri such little attention.

One academic PhD isn't going to make difference when other companies such as DeepMind have 250+ scientists with academic and real industry experience.

 

[usarioclave]

One academic PhD isn't going to make difference when other companies such as DeepMind have 250+ scientists with academic and real industry experience.

It's about focus.

Let's put it this way: one thing people are terrible at is financial management. Who is doing a brain for financial management so you don't have to do it? Maybe Intuit should be doing it, but they aren't. Something that reminds you and auto-pays your bills? Keeps you on-budget? Tells you that you won't be able to make rent?

The AI field is still full of snake oil. Heck even NLP, which is generally used for voice input, sucks. You can't multiple-round-trip google translate. And if Watson was so great, why isn't IBM making bazillions of dollars off of it?

Right now there no "behind", there's just "hasn't gone down as many blind alleys and dead ends."

Really, the most useful AI-like thing Apple is doing is telling me when it's time to leave given the estimated traffic conditions. That's not even AI, but who cares? It might as well be.

 

[macUser2007]

I was just running an experiment on this site....

Ah, I see. An experiment to see how many beers one can drink and how many beer bottles one can shoot shoot before someone notices the nutty post...?


[doublepost=1476776373][/doublepost]

... I've used Alexa, and I've seen what Amazon's AI stuff can do for me, which is nothing. Same with google. I don't really need better recommendations, or more appropriate ads. I really don't care about that crap. Talking to a device to manage my house is fun, but it's a parlor trick....

Huh?! Is this why the Echo is the currently second best selling-selling device on Amazon US?

I use it constantly. Because of it, I almost never use switches any longer to turn my lights on or off. I do a number of other stuff daily on Alexa, such as playing music, radio stations, podcasts, ask the time or just random questions I want to know the answer to.... It's actually pretty great, if you get used to it. It's "parlor-trick" only to someone who thinks a smart phone is a "parlor-trick."

 

[Ramchi]

I think Apple is doing the right thing! As I have mentioned earlier for an article for SIRI, they need scientists to lead this initiative. Not Engineers!

 

[usarioclave]

I use it constantly. Because of it, I almost never use switches any longer to turn my lights on or off. I do a number of other stuff daily on Alexa, such as playing music, radio stations, podcasts, ask the time or just random questions I want to know the answer to.... It's actually pretty great, if you get used to it. It's "parlor-trick" only to someone who thinks a smart phone is a "parlor-trick."

That's not AI, that's keyword matching with voice recognition. I know, I'm integrating a couple of things into Alexa with the SDK right now. If you like, you can see what Alexa really is by reading their SDK.

The public really has no idea what AI really is. You don't know what's behind the curtain; you're being fooled into thinking it's an AI. It's good enough for you because you're using it for nothing important. Try asking Alexa something that's not a tarted-up google search and she falls over.

 

[Muzzakus]

Feels to me as if Siri has devolved over time. Of course perhaps I'm just seeing it in the context of being grossly left behind by the competition.

Perhaps it's so fundamentally flawed that attempts in improvements are just a band aid approach. Scrap it and start again with solid, modern foundations. Can't keep going like this, or Apple will one day wake up and wonder 'where did everybody go?'.

 

[nazaar]

How long has Siri been around now, 5 years? And its still for lack of a better word garbage. Even children learn from their mistakes

 

[macUser2007]

That's not AI, that's keyword matching with voice recognition. I know, I'm integrating a couple of things into Alexa with the SDK right now. If you like, you can see what Alexa really is by reading their SDK.

The public really has no idea what AI really is. You don't know what's behind the curtain; you're being fooled into thinking it's an AI. It's good enough for you because you're using it for nothing important. Try asking Alexa something that's not a tarted-up google search and she falls over.

And your point is...?

Nope, it's not Ex Machina. I am well aware of it, but for my purposes, and for the purposes of many others, Alexa is a good start. In terms of usefulness, Alexa is way ahead of Siri. I expect Google Home to be ahead of Alexa.