Losing Ian Goodfellow to DeepMind is the dumbest thing Apple has ever done
There are many machine learning developers in the world. But only one has been mentored by both Andrew Ng and Yoshua Bengio, invented a new type of artificial neural network, contributed to or led research at Google Brain, OpenAI and Apple, and still has less than 40 candles to blow. on her birthday cake. .
And Apple just let him walk out and walk straight into the Google offices where he will soon be working for the DeepMind research team.
His name is Ian Goodfellow. And letting him walk away from your ultra-progressive tech company based in California, Silicon Valley, because he doesn’t agree with your work demands at the office is such a stupid scenario that I can’t not believe that GPT-3 did not invent this.
To put this in a sports analogy, it’s like letting Tom Brady or Michael Jordan quit your team because of a disagreement between them and the team owner over how the napkins should be folded.
Let Goodfellow and his team work wherever they want. If they think they can code a better machine learning model from the International Space Station, you should probably consider building a rocket.
Let’s step back for a second and put things into perspective. I can feel our C-suite readers reflexively whipping up a “good company” or “everyone is a rockstar in our organization” speech in their heads in response to this op-ed.
I don’t think anyone deserves special treatment at Apple. Or any other business, for that matter.
But Ian Goodfellow’s contributions to the field of machine learning cannot be overstated. Office hours are a stupid enough thing to ruin any talented developer, and it’s even more ridiculous to let your ML director walk around because you think in-person smiles are important.
Losing Goodfellow is a blow to Apple for two reasons:
- His talent cannot be easily replaced
- His work at DeepMind could put him in direct competition with Apple
Goodfellow’s greatest claim to scientific fame is as a member of the team that invented the Generative Adversarial Network (GAN).
A GAN is a neural network that learns to create content by trying to fool itself and, eventually, humans.
Whenever you hear of an AI capable of generating text, writing poetry, creating images, or producing its own original music, you almost certainly hear of a GAN.
The cool thing about GANs is that they work by pitting two neural networks against each other. Without GANs, humans would have to fine-tune each generative iteration, like trying to do coarse sanding with fine paper.
But a GAN has a network that creates and another that discriminates. The second network is essentially a bouncer that blocks many unnecessary exits before they even have a chance to show up.
General Artificial Intelligence (AGI)
After Goodfellow et al. invented the GAN, he continued his work at Google Brain where he helped solve problems in computer vision and ML security. And from there, he toured OpenAI, the Elon Musk and Microsoft-funded AGI think tank, where some of the world’s brightest minds are trying to figure out how to invent and control human-level artificial intelligence.
This is important to note because there’s really only one other company on the planet as deeply invested in AGI and full of well-known talent as OpenAI, and that’s DeepMind.
When Ian Goodfellow became director of machine learning at Apple, many of us in tech journalism were surprised. It seemed like a huge loss to the Google Brain team, but it made sense to Goodfellow (it seemed like a well-deserved promotion) and from what we know of Apple’s AI programs, it didn’t seem like not be something that would come back to bite Google in the ass.
Fast forward to today, and Google looks like the smartest player on the board as it welcomes Goodfellow back into the fold.
DeepMind is essentially Google’s version of OpenAI. Where OpenAI seems to be a bit more focused on ensuring that an AGI doesn’t stand against us, DeepMind is more obsessed with creating an generalist An AI that can do anything a human can do without needing to be retrained over and over again to learn new abilities.
It’s something that could come back to bite Apple in the butt.
Siri, Siri, why are you Siri?
About five years ago, if you wanted to make a joke about virtual assistants or anthropomorphic AI, you had to summon Siri. It’s the only named AI that all my readers knew about in 2017.
Now I better use Alexa for name recognition. But no one has forgotten Siri. At least not yet.
DeepMind is onto something big with its new GATO AI system. No, I don’t think it’s on the way to AGI with GATO (or, really, anything he’s currently doing, but that’s a discussion for a different article).
But I think GATO could be extremely marketable if DeepMind can overcome the problem of large-scale models and biases.
Imagine Siri, but a version of Siri that could do a thousand different tasks on your behalf. Right now, our virtual assistants basically do web searches and open apps for us. It may seem like Siri can do hundreds of different things, but telling you what time it is, how many messages you have, and what the capital of Nebraska is, it’s pretty much the same task.
I’m talking about a version of Siri that could control a robot that can wash your dishes, while simultaneously identifying weedy areas of your lawn, while generating a completely original cartoon for your kids to watch based on your specific prompts, and so on.
Currently, it would be an impressive feat for a team of AI developers to create a system that could do all of this in a simulated environment. The challenge of unleashing general purpose AI in the homes of random consumers is far greater.
What if DeepMind succeeded? What if, instead of Siri or Alexa, it was Google Assistant that became the first AI assistant in the world capable of really assisting you in your daily life?
If DeepMind and Google can turn the dull, boring 2D idea of what a virtual assistant is into something that could conceivably start to look like a real life assistant, everyone will forget about Siri. And Alexa. And any other “assistant” who can’t do what GATO can do.
I’m not sure DeepMind can pull it off, but I’m sure the odds increased significantly the moment the company signed a contract with the GANfather himself.
It’s your business, Tim
In the end, who knows what really happened at Apple. Maybe Goodfellow wasn’t happy, or maybe Apple wasn’t.
There’s no guarantee that DeepMind’s work will ever interfere with what Apple is trying to accomplish, although most of what everyone else is trying to accomplish in the field deals with some of Goodfellow’s ideas on the subject. deep learning.
And it’s also worth mentioning that big tech is constantly competing for talent. Let’s not forget that Goodfellow left Google twice, once to join OpenAI and the second time to join Apple.
But the moment he joins the DeepMind team is quite exciting. He would be on board as an independent researcher. It sounds a lot like he’ll be getting anything and everything he needs to do his best.
Maybe Apple CEO Tim Cook has good reason to let his star quarterback leave to join a rival team midway through the playoffs. It’s hard to see from our vantage point outside the walled gardens of Cupertino society, but it is possible.
Either way, it’s an exciting time for the field of AGI research. It’s unclear what Goodfellow and the DeepMind team can accomplish together.
Be sure to sign up for the Neural newsletter so you don’t miss a thing.