Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster
Apr 12, 2001
68,627
39,507



Apple in March hired Ian Goodfellow as its director of machine learning in the Special Projects Group, according to Goodfellow's LinkedIn profile that was shared by CNBC.

Prior to joining Apple, Goodfellow worked at Google as a senior staff research scientist, and before that, he was an Open AI research scientist.

coreml-apple-800x400.jpg

As outlined by CNBC, Goodfellow is known for creating GANs, or Generative Adversarial Networks, a concept where two deep neural net architectures (one generative and one discriminative) are pitted against each other to generate creative content.

GANs have been used to create fake art and deepfakes, aka fake digital images that are indistinguishable from real images.

While at Google, Goodfellow worked on GANs technology and security. It's not clear what he's doing at Apple, but the company's machine learning techniques are used for many elements of its software.

Apple has been working on bolstering its AI and machine learning efforts in recent years, with another former Google employee, John Giannandrea, heading up those divisions.

Giannandrea, who likely oversees Goodfellow's work, was named Apple's senior vice president of machine learning and AI strategy back in December, just months after his original April hiring.

Article Link: Apple's Special Projects Group Gains Former Google Machine Learning Expert
 
  • Like
Reactions: DeepIn2U
we are experimenting with GANs at work. they can be described as the creative part of the brain. compared to “classic” neural nets, they don’t just classify some data - they create it. so for example: if you don’t have enough data, GANs can create new data which could be pure gold for a company like apple which tries to not collect a huge amount of data.
 
Companies who don’t research / invest in themselves will no doubt die. Many large tech companies, I.e Microsoft partner with universities for research, that will benefit everyone.

I love hearing that Apple has people working on projects without an immediate pay off. That’s the kind of stuff that great companies like Bell Labs had for years.
 
we are experimenting with GANs at work. they can be described as the creative part of the brain. compared to “classic” neural nets, they don’t just classify some data - they create it. so for example: if you don’t have enough data, GANs can create new data which could be pure gold for a company like apple which tries to not collect a huge amount of data.

You’re developing circuits that can make up stuff better than humans. I read an interesting article awhile ago suggesting it was the background noise in the human brain that accounted for new inspiration, are these circuits being made noisy deliberately?
 
  • Like
Reactions: bayelrey80
This is one of the benefits of hiring someone like Giannandrea—he knows who to take out from his old group. There are probably dozens more from Google at less senior levels as well.

Giannandrea is a Srouji-level hire, these guys are worth their weight in Apple RSUs. But of course it’s not really about compensation; the important thing is the opportunity to do the best work of your life, surrounded by others who are equally passionate about their work.
 
Last edited:
so far, all this "deep machine learning" has gotten us is the "For You" tab in the photos app and weird collections of algorithm chosen collections.
It seems that "machine learning" is the current buzzword (like "internet website" was back in the eighties)
I appreciate the work these guys do, but I’m also struggling to see what is ‘smart’/‘deep learning’/AI’ in ios right now.

All I see are some random guesses that software was able to do in the 2000s without ‘neural processors’.

See 2007’s Superfetch in windows vista that basically did the same as today’s ‘neural’ insights: https://techcommunity.microsoft.com...sta-SuperFetch-amp-amp-ReadyBoost/ba-p/372337

I’m not saying neural nets are useless, because companies like google do great things with them. Just that apple is being useless at implementing them usefully
 
  • Like
Reactions: martyjmclean
so far, all this "deep machine learning" has gotten us is the "For You" tab in the photos app and weird collections of algorithm chosen collections.
It seems that "machine learning" is the current buzzword (like "internet website" was back in the eighties)

It's a phase that we sometimes have to go through in order to workout/discover exactly how to integrate new technology into our daily lives.

The early days of electricity, entrepreneurs put it into everything. Some things worked out, some didn't. Same thing with radios and TVs. As the tech improved and reduced the physical size, they were slapped into anything as people just threw things against the wall to see what would stick. Same thing here. Enjoy the ride knowing that somewhere, someone is working on Alexa enabled public urinals.
 
I think there’s some huge potential coming in revamping Apples artificial intelligence sector. And I Don’t mean just Siri, I mean there’s more potential here to see how Apple creates artificial intelligence on another level where it’s beyond just iOS. I’m not sure exactly where this will lead, but I have a feeling within this next year, we should see more where Apple is leading under Giannandrea.
 
Last edited:
  • Like
Reactions: bayelrey80
I appreciate the work these guys do, but I’m also struggling to see what is ‘smart’/‘deep learning’/AI’ in ios right now.

All I see are some random guesses that software was able to do in the 2000s without ‘neural processors’.

See 2007’s Superfetch in windows vista that basically did the same as today’s ‘neural’ insights: https://techcommunity.microsoft.com...sta-SuperFetch-amp-amp-ReadyBoost/ba-p/372337


You can search for image-content.

Photos found the pics of the passport of my GF that I had.
Someone had told it how a passport looked like.
 
I hope Apple learns to become more open with its research. I think Apple's lag with ML is in part because not being able to collaborate and be open with your work is a big turn off to AI talent.
 
  • Like
Reactions: martyjmclean
It's a phase that we sometimes have to go through in order to workout/discover exactly how to integrate new technology into our daily lives.

The early days of electricity, entrepreneurs put it into everything. Some things worked out, some didn't. Same thing with radios and TVs. As the tech improved and reduced the physical size, they were slapped into anything as people just threw things against the wall to see what would stick. Same thing here. Enjoy the ride knowing that somewhere, someone is working on Alexa enabled public urinals.
Remember in the 80s when they just put a digital clock into everything? Or a little solar panel?
 
  • Like
Reactions: jeremiah256
So, is Google doomed now that two people in their AI division have jumped ship?
No. Doom only happens when someone leaves Apple, silly.
[doublepost=1554419170][/doublepost]
Remember in the 80s when they just there a digital clock into everything? Or a little solar panel?
Remember?!?! I still have the toaster-clock my bank gave me!
 
My Apple devices with Siri, a PC with Cortana, Alexa on a Kindle, and two Google Home Minis. It's all Downton Abbey up in here, except there's no endowments and more sass from the assistants.
 
Considering their engineering force outnumbers Apple’s by numbers alone, I’d say yes.
[doublepost=1554429892][/doublepost]

Fortunately for CA employees, there are no anti-compete clauses, so people in tech companies jump around at will
So why can’t Google produce a single decent piece of hardware? Or software for that matter?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.