Apple To Start Publishing AI Research To Hasten Deep Learning (bloomberg.com)

In what is a major deviation in its strategy, Apple will allow its artificial intelligence teams to publish research papers for the first time. From a report on Bloomberg: When Apple introduced its Siri virtual assistant in 2011, the company appeared to have a head start over many of its nearest competitors. But it has lost ground since then to the likes of Alphabet's Google Assistant and Amazon's Alexa. Researchers say among the reasons Apple has failed to keep pace is its unwillingness to allow its AI engineers to publish scientific papers, stymieing its ability to feed off wider advances in the field. That policy has now changed, Russ Salakhutdinov, an Apple director of AI research, said Monday at the Neural Information Processing Systems conference in Barcelona. One attendee posted a photo of a slide from Salakhutdinov's presentation stating "Can we publish? Yes. Do we engage with academia? Yes."

Comments

Recruiting

[ShanghaiBill]

The main reason that Apple is now allowing work to be published is recruitment. In the past, many of the best AI people refused to work for Apple because they couldn't publish their work. Excessive secrecy has a price.

Re:Recruiting

[dgatwood]

I came here to post the same thing. It's about time Apple finally started behaving like a proper member of the academic community. On the other hand, I'd give it five... ten years tops before some leak causes them to clamp down and become paranoid again.

Re:Recruiting

[bug_hunter]

Smart move from Apple - even if it was to replace a previous dumb move.

Apple can do fine living in isolation developing hardware enclosures and user interfaces, but the world of AI is much bigger than Apple and it's where the majority of progress in the computing world is going to happen now.

I'm happy Apple has realised this. Apple vs "the rest of the world" in developing AI was not going to end well for them.

Re:Recruiting

[tlhIngan]

Smart move from Apple - even if it was to replace a previous dumb move.

Apple can do fine living in isolation developing hardware enclosures and user interfaces, but the world of AI is much bigger than Apple and it's where the majority of progress in the computing world is going to happen now.

I'm happy Apple has realised this. Apple vs "the rest of the world" in developing AI was not going to end well for them.

Well, Apple still has a handicap that's self-imposed - they may be able to do AI research, but they can't get at the data sets they need. Amazon and Google have no problems using information gathered for other purposes in furthering their AI products. Heck, Google even enabled full sharing of information among all of Alphabet's companies so your "OK Google" responses can now include all the webpages you visited with advertising. Or that you blocked DoubleClick's popups.

Apple prohibits that - information gathered for one purpose cannot be used for another purpose. So if you want to try to use Apple Maps location and search data for Siri, that's not allowed. You have to guess based on the current location of the user, because location searches are not allowed to be part of Siri's data set as they are covered under different privacy policies.

Apple can try to hire the best, but they'll need to start raping user data if they want to try to catch up with Google.

Not the reason Apple is behind in AI

[Anonymous Coward]

This is just ridiculous.The real reason that Apple is behind on AI, or rather Siri features, is because Siri has always been a hack. And by hack I mean that it was written for very specific use cases. It was never designed to learn, or be user extendable. In fact, the guys who wrote Siri quit Apple after their contracts were up and are busy writing a follow up to Siri. Probably another cleverer hack, but also not very extendable. Has anyone notices that IBM's Watson has not been offered as a personal digital assistant? That's because although Watson can learn, and is extendable, it is not scalable. Watson runs on a supercomputer. To a PDA, it would need to run on a whole lot of supercomputers...

Re:Not the reason Apple is behind in AI

[LifesABeach]

Granted that Voice to Text is not a trivial endeavor; text input to an app is. The reason for this choking of Apple progress isn't because its AI endeavors have hit a creative wall, it's because the H1B zombies only do what they're told. Machiavelli stated this argument when he wrote of Mercenaries; they're good at their trade, but not dependable. The fact that Apple is looking for innovation from a pack of H1B zombies is comedy.

Re:Not the reason Apple is behind in AI

[ShanghaiBill]

Poor A/C, it's not a super computer.

Actually, it is. Until the early 1990s, a "super computer" was a really fast computer. That is not true anymore. Today a "super-computer" is big array of commodity CPUs and/or GPUs. You can rent these by the hour or minute from AWS or Google.

The current version of Watson runs on a cluster of 90 Power7 CPUs with 8 cores each, and a total of 16 TB of RAM.

Re:Not the reason Apple is behind in AI

[LifesABeach]

For example, if I want to solve the problem of a what size counter weight should be attached to a ribbon attached to the Earth at the Equator. Do I go to Cray? Or do I go to Cloud? Another example is, I want to sell cloths made of Hemp. Do I go to Cray? Or do I go to Cloud?

Re:Trickle-down research

[hackwrench]

Feed off as in unable to have outsiders look at their stuff and go, "Hmm, if you just made this little tweak..."

Scientific Research

[Jzanu]

Apple should look into leaving the US behind and moving HQ to a real democracy. Those are much safer for conducting research and being able to freely disseminate it.

Re:Apple encourages homesexuality

[PopeRatzo]

Apple encourages homesexuality

As opposed to worksexuality.