There are a lot of machine learning developers in the world. But only one person, mentored by Andrew Ng and Joshua Bingyu, who invented a new type of artificial neural network, and contributed or led research at Google Brain, OpenAI, and Apple, still has fewer than 40 candles exploding on their birthday cake.
Apple has let him out of the house and into the Google offices where he will soon work with the DeepMind research team.
His name is Ian Goodfellow. And letting him walk away from a California-based, progressive high-tech company because he doesn’t fit the requirements of working in-office is a pretty stupid scenario I can’t believe GPT-3 didn’t come up with him-she.
To put this into a sports analogy, this is like letting Tom Brady or Michael Jordan leave your team because of a disagreement between them and the team owner over how to fold the towels.
Let Goodfellow and his team work from wherever they want. If they think they can code a better machine learning model than the International Space Station, you should probably consider building a rocket.
Let’s go back and put things straight. I can feel our C-suite readers reflexively preparing the “good for company” or “everyone’s a rock star in our organization” speech in their minds in response to this opinion piece.
I don’t think anyone deserves special treatment at Apple. Or any other company for that matter.
But Ian Goodfellow’s contributions to machine learning cannot be overstated. Hours of work is too silly to lose sight of any talented developer, it’s even silly to let your ML manager walk because you think personal smiles are important.
Losing Goodfellow is a huge blow to Apple for two reasons:
- His talent is not easily replaceable
- His work at DeepMind could put him in direct competition with Apple
Goodfellow’s greatest claim to scientific fame is that he was part of the team that invented the Generative Adversarial Network (GAN).
A GAN is a neural network that learns how to create content by trying to deceive itself, and ultimately humans.
Anytime you hear about an artificial intelligence that can create text, write poetry, create images, or produce its own original music, you’ll almost certainly hear about 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 iteration – like trying to do coarse sanding with fine paper.
But the GAN has one network to be created and another to be differentiated. The second network is basically a guard that blocks a lot of useless output before it has a chance to show up.
Artificial General Intelligence (AGI)
After Goodfellow et al. He invented the GAN, and went on to work at Google Brain where he helped solve computer vision and ML security problems. From there he toured OpenAI, the Microsoft-funded Elon Musk and AGI research center, where some of the world’s brightest minds try to figure out how to create and control human-level artificial intelligence.
This is important to note because there is only one company on the planet that is deeply invested in Artificial General Intelligence and filled with well-known talent like OpenAI, and that is DeepMind.
When Ian Goodfellow became Apple’s director of machine learning, many of us in the tech press 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 upgrade), and based on what we know about Apple’s AI software, it doesn’t look like something will come back to bite Google in the ass.
Fast forward to the present day, and Google appears to be the smartest player on the board as it once again welcomes Goodfellow into the fold.
DeepMind is basically Google’s version of OpenAI. Where OpenAI appears to be focusing a little more on ensuring that General Artificial Intelligence does not play against us, DeepMind is more focused on creating specialist An AI that can do anything a human can do without having to be retrained over and over to gain new abilities.
This is something that could come back bite Apple in the ass.
Siri, Siri, why are you Siri?
About five years ago, if you wanted to make a joke about virtual assistants or holographic artificial intelligence, you had to call Siri. This is the only named AI that all of our readers were familiar with in 2017.
Now, it’s better to use Alexa for name recognition. But no one forgets Siri. At least not yet.
DeepMind is working on something huge with its new AI system, GATO. No, I don’t think he’s on his way to AGI with GATO (or, really, anything else he’s currently doing, but that’s a discussion of a different article).
But I think GATO can be very marketable if DeepMind can overcome the problem of massive models and bias.
Imagine Siri, but a version of Siri that can do thousands of different tasks for you. At the moment, our virtual assistants mainly perform web searches and open apps for us. It may seem like Siri can do hundreds of different things, but telling you the time, how many messages you have, and the capital of Nebraska, it’s all pretty much the same job.
I’m talking about a version of Siri that can control a robot capable of washing your dishes, while simultaneously identifying high weedy areas in the front lawn, while creating a completely original cartoon for your kids to watch based on your specific directions, and so on and so forth.
For now, it would be a great feat for a team of AI developers to create a system that can do all of this in a simulated environment. The challenge of letting a specialized AI into the homes of random consumers is much greater.
But what if DeepMind works? What if, instead of Siri or Alexa, the Google Assistant became the world’s first AI assistant capable of actually helping in your everyday life?
If DeepMind and Google can turn the boring two-dimensional notion of what a virtual assistant is into something that can practically start to look like a real life assistant, everyone will forget about Siri. And Alexa. And any other “assistant” cannot do what GATO can do.
I’m not sure DeepMind could pull it off, but I’m sure the odds increased by a huge margin the moment the company signed a contract with GANfather itself.
It’s your job, Tim
Ultimately, who knows what really happened at Apple. Maybe Goodfellow wasn’t happy, or maybe Apple wasn’t.
There is no guarantee that DeepMind’s work will go against what Apple is trying to achieve, although most of what everyone is trying to achieve in this field deals with some of Goodfellow’s ideas about deep learning.
And It’s also worth noting that big tech is catching talent from each other all the time. Let’s not forget that Goodfellow left Google twice, once to join OpenAI and the second time to join Apple.
But the timing of his joining the DeepMind team is very exciting. He is said to be working as an independent researcher. This looks a lot like he’s going to get anything and everything he needs to do his best job.
Apple CEO Tim Cook may have good reasons to let him leave the star midfielder to join a rival team in the middle of the playoffs. It’s hard to see from our vantage point outside the walled gardens of the Cupertino Company, but it is possible.
No matter what, it’s an exciting time in the field of artificial general intelligence research. There’s nothing Goodfellow and the DeepMind team can tell us together.
Make sure to sign up for our Neuro Newsletter so you don’t miss a thing.