This week in AI #18: AI on the mind
It seems like every week AI takes one step further away from futuristic fantasies – forged in our minds from watching too many science fiction films – and one step closer to becoming a very real and very essential part of our society. As you’ll see from this week’s picks, the applications can however be just as exciting!
1 — Leave it to the machines
Last week, Andrew Ng urged the deep-learning community to focus less on research and more on practical applications. As far as society is concerned, it has one significant implication: certain jobs will disappear or drastically change. There’s no point denying it. Some sources even go as far as saying that it could impact up to 30% of the workforce. The tough question we have to ask ourselves is: are we afraid because some of those jobs are essential and will disappear or because this will change society as we know it? Bear with me for a moment. If you look around you, there are plenty of jobs that feel more robot than human, try working in a factory for 8 hours, doing the checkout in a mall, or filling in payrolls. 99% of the time, people don’t do these repetitive jobs because that’s what passionates them, they do them because they need a job. The main focus point then becomes to make sure people still have means to live but more importantly that they fulfill their potential as humans, which doesn’t necessarily mean protecting their current job. In this sense, I think we’ll see more and more collaboration between humans and AI where AI takes care of the grunt work and leaves the more creative stuff to us.
2 — The future is here!
For those who don’t like driving and prefer to spend their time on more interesting stuff, we have great news: Waymo announced last week that their very first fully self-driving car fleet is hitting the road. No driver in the front seat anymore! You don’t believe us? Try it out yourself… You want to know more? Well, then keep on reading. Elon Musk just rolled out his new electric semi-truck with an enhanced autopilot. His mission: to quite simply save the world (by reducing toxic emissions and accidents). That’s still not enough? Then maybe you’ll love the latest Uber idea: on-demand aviation. Yes, you read that right. Flying taxis are coming…
3- Worry about the short-sighted greed
AI is everywhere nowadays. Even if we don’t always see it, it is becoming more and more a part of our lives. For now, it’s mostly dedicated to economic goals: more clicks, more purchases. A short-sighted vision which only benefits a few people. In addition, AI is influencing our society, but its implications are not at all studied. That’ll change with the newly founded AI Now Institute. One possible solution to the black-box vision would be to democratize AI and educate its users: let everyone see its inner workings and decide by themselves if they trust the AI decisions.
4 — Once upon a time…
Apple made a big splash when it got rid of its fingerprint reader for face identification on the iPhone X. Although, before identifying a face you first need to detect and locate it. This problem has been tackled for some time now, if you have the computing power, it is even considered solved by some. What if you don’t? What if you want to use face recognition on your smartphone but can’t stream it on a central server because of privacy concerns? That’s an entirely different story. One that Apple is ready to tell: how they went from computer vision of the last decade to deep-learning. There is one interesting fact in this article; they are still using OverFeat, a network that’s already four years old. It proves that it’s no easy task to transition a model running on a desktop computer with a GPU to an embedded device with a constrained environment while at the same time making it production-proof!
5 — Deep-learning diet!
Don’t despair! Things might be about to change. Google has already put a fair amount of work into making life easier for you when it comes to using neural networks on mobile and embedded devices and… they’re stepping on the gas pedal! They just announced the release of TensorFlow Lite, the evolution of TensorFlow Mobile. The objective? Give them a trained network, and they will optimize it to run on mobile. At the moment you’re limited to three models, but hopefully, we’ll soon be able to use a custom one! More broadly, this could be a decisive turn in the race to becoming the standard tool for deep-learning. Lately, Google has been pushing hard to be the first on the finish line, and it might just be starting to pay-off!
Woah ! It’s hard not to feel a mixture of excitement, curiosity and maybe a pinch of fear after this week’s top picks ! AI is new, innovative and has HUGE potential… this means that every week there are new advances in AI research and its applications. We’ll help you stay in the know, all you need to do is subscribe to This Week in AI below !