Philip Houska If technological rivals share data and knowledge with each other quite openly, this is the field of artificial intelligence, which is moving forward much faster thanks to mutual cooperation. Apple, which has so far remained on the sidelines as it usually tries to keep its initiatives under wraps, is now likely to join them. The Californian company wants to cooperate with external experts and academics around the world and, thanks to this, to gain additional experts to its teams.
Russ Salakhutdin, the head of artificial intelligence research at Apple, revealed the information at the NIPS conference, which discusses, for example, the issue of machine learning and neuroscience. According to the published footage of the presentation from people who do not want to be named due to the sensitivity of the topic, it can be read that Apple is working on the same technologies as the competition, only in secret for now. These include, for example, image recognition and processing, predicting user behavior and real-world events, modeling languages for voice assistants, and trying to solve uncertain situations when algorithms cannot offer confident decisions.
For the time being, Apple has made a more prominent and public profile in this area only within the voice assistant Siri, which it is gradually improving and expanding, but the competition often has a slightly better solution. Above all, Google or Microsoft do not only focus on voice assistants, but also other technologies mentioned above, which they openly talk about.
Apple should now begin to share its research and development of artificial intelligence, so it is possible that we will at least get a rough idea of what they are working on in Cupertino. For the otherwise very secretive Apple, this is definitely a relatively big step, which should help it in the competitive struggle and further development of its own technologies. By opening up development, Apple has a better chance of attracting key experts.
The conference also discussed, for example, the LiDAR method, which is a remote measurement of distance using a laser, and the aforementioned prediction of physical events, which are key to the development of autonomous technologies for cars. Apple demonstrated these methods in pictures with cars, although according to those present, it never specifically talked about its own projects in this area. Anyway, it surfaced this week letter addressed to the US Traffic Safety Administration, in which the Californian firm acknowledges the efforts.
Considering Apple's ever-increasing openness and the generally rapidly developing field of artificial intelligence and related technologies, it will certainly be very interesting to watch further developments within the entire market. It was also said at the mentioned conference that Apple's image recognition algorithm is already up to twice as fast as Google's, but we'll see what that means in practice.
jojo, marketing to make it look like something is happening :]
let's believe and see, unfortunately for apple, apple's main competition has far more data from which to build AI
:-)
I already wrote it here http://jablickar.cz/apple-v-oblasti-umele-inteligence-bud-spi-nebo-skryva-sve-karty, and according to this comment of yours, I obviously hit the mark.
I like to learn, so what did apple demonstrate?
First I recommend you look at the original articles, the local one is, well, local. After all, the editors do not hide the fact that the topic of AI, ML, DL, etc. is not their own.
What Apple will actually demonstrate at NIPS will be known with some delay, so for now it is only one step in a series.
You probably know that today there are a significant number of companies dealing with AI (more precisely, specific areas within AI). As a rule, they do not do basic research, but draw mainly from the results of the academic sphere. And only a few of these companies publish their results, even partial ones. Except as finished products :-)
Apple does/used to do the same. But apparently he wants to change that and start publishing his research (certainly only part of it). And he also stated in which areas AI works. That's all for now.
It doesn't say anything about what Apple has achieved so far in AI, who it's working with, etc. You need other sources of information for that. And they exist :-)
Well, I understood that I missed something and Apple "somehow" showed up, but I see that it didn't
I agree with the rest of what you wrote, but I think that Apple simply fell asleep and has nothing* (* in the sense of something equivalent to a company at least one-tenth the size of Apple)
In no way do I want to advocate for Apple, or claim that it's doing God damn well in AI compared to others. But since I'm from the "field", I'm interested, after all I have to, what's going on where.
And since your contributions to the already mentioned article http://jablickar.cz/apple-v-oblasti-umele-inteligence-bud-spi-nebo-skryva-sve-karty they came to be, well, let's say not much, I got the urge to rant at you a little :-)
Hmmm. I haven't had, don't have, and don't plan to have any Apple product anytime soon.
you can have compulsions and there is nothing wrong with that. I had the urge when I read the article and saw in it a typical Apple marketing, a lot of nonsense for the ignorant masses just to make an impression
as a person from the field, you must be well aware that apple is not even in the second, but in the third league
apple is a hardware company, it can't do software, what it does is not essential, it doesn't produce it (somewhat significant % of the profit),...
I doubt that he will invent something fundamental in AI, if he does not invent new things in hardware. apple was simply lucky to have a couple of attacks, a marketing genius (not even a technological one), and associating device manufacturers with the fact that they are expected to come up with something revolutionary in AI, that is simply naive. that company revolves around design and marketing, not research
and my personal opinion is that the first one to come up with some really big breakthrough in AI will be some third party, probably the army (funded by the army), the rest of the first agi (and any precursor to the development of it) should be used for cyber attacks/defense and that and from the point of view of the economic usefulness of such a system (and AI research is booming there anyway)
btw, what's your hw? :)
I will only answer the final question for now, I don't have much time now:
Dell regular ntb, 15″, 16GB, SSD 512GB, i7 quad core. Windows, Linux. It is enough for development in the company and at home (modeling, verification tests, lately mainly neural networks), real training, validation, etc., of course, run on a proper Linux station with TITAN from NVidia ;-)
I would like to add that the target platform of our systems is mainly customer HW (ARM based + FPGA), but also server applications for clients.
Apple is of course also a SW company, if we recognize that the OS for their HW is the SW developed by them :-)
In terms of application, it is weaker, I agree, and I would almost say that it is getting worse. But that's their business.
If I'm not mistaken, the experts will surely correct me, Apple has never brought anything groundbreaking based on its own research, but neither have most other companies. But he was able (is able) to solidly implement the existing one (even with some of his own invention) and sell it. But there's nothing wrong with that, right?
I don't expect anything groundbreaking from him, certainly not in the field of AI (but perhaps no one with a bit of sense, nor do I feel that anyone is proclaiming it). Nor from most other companies. In any case, the essential things happen at universities, and companies draw on them.
Mutual cooperation, now more of an "employment relationship", between people from the academic environment and companies has been especially strong in recent years. I will name a few of the most sloppy and probably the most visible here on this server - IBM, Google, Microsoft, Facebook, Baidu, etc. Yes, even Apple :-) And people from universities such as LeCun, Norvig, Thrun, Hinton, Ng, Bengio, Krizhevsky, Salakhutdinov, Sutskever, and dozens, hundreds of others. Sometimes something similar works in the Czech Republic, but unfortunately not much :-(
Btw. for developers, Apple already provides a low-level API for Deep Learning (not only) in the latest versions of iOS and macOS. For running both on CPU and within Metal. In Metal, it also offers a higher-level API focused on the currently most popular field of computer vision (convolutional neural network). Thanks to the development of SW/OS to "fit your HW", you can believe that performance will be better, or certainly not worse, than similar third-party frameworks (Caffe, Tensorflow, etc.).
Details can be found, among other things, in videos from WWDC 2016 and developer documentation.
I don't have practical experience with Apple products, so I can't personally evaluate the quality.
Apple can't compete with Google in AI at all...