Just how smart AI is?
I had the opportunity to participate in an interview on BloxLive recently, about whether we should fear AI or not. Interestingly enough, this topic is becoming more and more popular in the media. Netflix recently released a film called I AM MOTHER where a robot grows a child after humanity has been eradicated. We are witnessing a resurgence in this theme in popular culture, much like we witnessed in the 80s and 90s with movies like The Terminator. Also, there are voices of people like Elon Musk, that have said that we should all be very scared of AI.
So, should we really fear AI? The truth is that AI is far from creating an all powerful, omniscient algorithm like the ones you see in movies. Right now AI is good in solving narrow problems. While many of the things that we can do through modern techniques are pretty impressive, the algorithms are not of general use, and require many hours of training and fine tuning. For example, NVIDIA has released an amazing Generative Adversarial Network, called StyleGAN, which can produce very realistic artificial human faces, like the ones you see below.
In order for NVIDIA to make this work, they had to train the neural network for many weeks on multiple GPUs. On top of that, you have to take into account the engineering effort required to actually create and tune the network, and find the right parameters for it. Similarly, OpenAI released a language model that is arguably so good, that they decided to not release it to the public in fear of it being used to disseminate fake news. However, once again, this model has taken a long time to develop, and it can only be applied in a very narrow domain.
Is artificial intelligence going to steal your job and burn down your house?
It would be very easy to find 100s more examples of cases where machine learning is being used, but in a very narrow set of problems. Impressive results, but far from what you’ve seen on The Terminator. Hence, what can we make of this, and should we fear AI or not? Well, the answer is yes and no.
AI is evolving rapidly, and is disrupting multiple industries at the same time. This disruption will help some companies grow, and it will destroy others. It will destroy jobs, and create new jobs, radically changing the landscape of our economy. Also, it has the potential to be used against citizens by governments, but also terrorists alike. Therefore, AI it is an important and disruptive technology. However, the implications are not as simple as a super-intelligent machine destroying humanity. Rather, similar to other disruptive technologies of the past, like the steam engine, there will be multiple effects across many parts of the economy and society.
Dealing with the effects of AI requires citizens to be well informed about how AI works. This is why through my work, including my book and my courses, I have tried to inform the public as best as I can about AI. I believe that it is the responsibility of all experts to do the same, and there are some, like Yann LeCun who are doing a great job at it.