There are many machine learning developers in the world. But only one was a mentor of both Andrew Ng and Joshua Bengioinvented a new type of artificial neural networkhas contributed or led research at Google Brain, OpenAI and Apple and still has less than 40 candles to blow on his birthday cake.
And Apple just let him out the door and right in Google’s offices, where he will soon be working for DeepMind’s research team.
His name is Ian Goodfellow. And let it get away from your California-based, high-tech elite Silicon Valley technology company because he does not agree with your requirements for office work is a scenario that is so dumb that I can’t believe GPT-3 didn’t invent it.
To put this in a sporting analogy, it’s like letting Tom Brady or Michael Jordan leave your team because of disagreements between them and the team owner about how the towels should fold.
Let Goodfellow and his team work 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 for a second and put things in perspective. I can feel our C-suite readers reflexively preparing a speech “good for the company” or “everyone is a rock star in our organization” in their heads in response to this opinion.
I don’t think anyone deserves special treatment at Apple. Or any other company, in that sense.
But to Ian Goodfellow contribution in the field of machine learning cannot be overestimated. Working hours is a pretty stupid thing to lose a talented developer, and it’s even funnier to let your ML director walk around because you think personal smiles are important.
The loss of Goodfellow is a huge 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 part of the team he invented generative competition network (GAN).
GAN is a neural network that learns how to create content by trying to deceive itself and ultimately people.
Every time you hear about AI that can generate text, write poetry, create images, or produce your own original music, you’re almost certain to hear about GAN.
The great thing about GAN is that they work by pitting two neural networks against each other. Without GAN, people will have to refine any generative iteration – as an attempt at rough sanding with fine paper.
But GAN has one network that creates and another that is discriminatory. The second network is essentially a killer that blocks a very useless exit before it even has a chance to manifest.
Artificial Common Intelligence (AGI)
After Goodfellow et al. invented GAN, he continued his work at Google Brain, where he helped solve problems with the security of computer vision and ML. And from there he toured the OpenAI, funded by Elon Musk and the Microsoft think tank AGI, where some of the world’s brightest minds are trying to figure out how to invent and control artificial intelligence on a human level.
This is important to note, because there really is only one other company on the planet so deeply invested in AGI and full of well-known talents like OpenAI, and that is DeepMind.
When Ian Goodfellow became Apple’s director of machine learning, many of us in technology journalism were surprised. It seemed like a huge loss to the Google Brain team, but it made sense for Goodfellow (it looked like a well-deserved promotion) and based on what we know about Apple AI programsit didn’t look like it was going to come back to bite Google in the ass.
Scroll fast to today and Google looks like the smartest player on the board as it once again welcomes the Goodfellow on the fold.
DeepMind is essentially Google’s version of OpenAI. When OpenAI seems to be a little more focused on ensuring that AGI doesn’t rise against us, DeepMind is more focused on creating generalist AI that can do everything a person can do without having to retrain again and again to acquire new abilities.
This is something that can come back to bite Apple in the ass.
Siri, Siri, why are you Siri?
About five years ago, if you wanted to joke with virtual assistants or anthropomorphic AI, you had to call Siri. This is the only named AI that all my readers were familiar with in 2017.
Now it is better to use Alexa for name recognition. But no one has forgotten about Siri. At least not yet.
DeepMind is on to something huge with it new GATO AI system. No, I don’t think he’s on his way to AGI with GATO (or, really, something else he’s doing right now, but that’s a debate for another article).
But I think GATO could be extremely marketable if DeepMind manages to overcome the problem with giant models and addiction.
Imagine Siri, but a version of Siri that can perform thousands of different tasks on your behalf. At the moment, our virtual assistants are essentially doing web searches and opening applications 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 which Nebraska capital is almost the same task.
I’m talking about a version of Siri that could control a robot capable of washing your dishes, while identifying areas with a lot of weeds in your front lawn, while generating a completely original animated film for your children to watch based on your specific prompts, and so on.
At the moment, it would be an impressive achievement for a team of artificial intelligence developers to create a system that can do all this in a simulated environment. The challenge of letting universal AI flourish in the homes of casual users is much greater.
But what if DeepMind does? What if instead of Siri or Alexa, Google Assistant becomes the world’s first AI assistant capable of actually helping in your daily life?
If DeepMind and Google manage to turn the playful, boring, 2-D idea of what a virtual assistant is into something that might start to look like a real helper in life, everyone will forget about Siri. And Alex. And any other “assistant” who can’t do what GATO can.
I’m not so sure DeepMind can handle it, but I’m sure the odds increased by a big margin the moment the company signed a contract with GANfather himself.
It’s your job, Tim
At the end of the day, who knows what really happened at Apple. Maybe Goodfellow wasn’t happy, and maybe Apple wasn’t.
There is no guarantee that the work of DeepMind will ever interfere with what Apple is trying to achieve, although most of what everyone is trying to achieve in this area deals with some of Goodfellow’s ideas for deep learning.
And it is also worth mentioning that great technologies draw talent from each other all the time. Let’s not forget that Goodfellow left Google twice, once to join OpenAI and a second time to join Apple.
But the time for him to join the DeepMind team is quite exciting. He is reportedly on board as an independent researcher. It sounds a lot like he will be given everything and everything he needs to do his best.
Perhaps Apple CEO Tim Cook has good reason to let his star quarterback leave to join a rival team in the middle of the playoffs. It is difficult to see from our point of view outside the company’s fenced gardens in Cupertino, but it is possible.
Nonetheless, this is an exciting time for AGI’s field of research. It is impossible to say what Goodfellow and the DeepMind team can achieve together.
Don’t forget to register the Neural bulletin so you don’t miss anything.