GUEST BLOG. Artificial intelligence fascinates. We deserve a well-deserved respect, even a certain force of attraction by its revolutionary nature, without however being able to prevent feeling a certain fear. The fear that it will end up going too far beyond the limits of what humans can really hope to control. All this without being able to grasp the essence of what defines it or to visualize its form, more intangible than concrete.
Inevitable, because the largest companies are already completely revising their business models around artificial intelligence (AI). Added to this are certain figures in the scientific field such as Stephen Hawking and Elon Musk who expressed serious fears. Not to mention the warnings issued by a multitude of theorists that humanity is threatened by hyper-intelligences which, at best, will steal all jobs, and at worst, will massacre humans. In the end, one wonders if it is simply possible to build truly intelligent systems, then cascading philosophical reflections questioning when artificial intelligence will become natural, and vice versa.
Well, it’s already heavy.
This paradox of ideas and messages inspired me to demystify AI, in order and in disorder. The whole is unpretentious and aims to make the subject accessible to all.
What is “unnatural” so-called artificial intelligence?
Let’s face it, the main use of artificial intelligence right now is to accomplish what humans are already doing, only faster, better and by taking it further.
Put simply, artificial intelligence is a sub-domain of computing. Its objective is to allow the development of computers capable of performing the things normally performed by human intelligence. In other words, AI is the science that makes machines smart. For a system to be considered an AI, it does not necessarily have to function in the same way as a human, but rather to be “intelligent” in the accomplishment of a precise task like winning a table game, correctly identifying a cat in a photo, use sensor data appropriately to drive a car, or any other process involving higher functions that one would characterize as typically human.
AI is the new electricity
Visualizing what artificial intelligence is is anything but easy. We often try, wrongly, to compare it to a “solid” technology. If it is true that it would be much simpler to visualize it as we do for a computer, an automobile or a robot, AI is rather a technology called enabling.
I often compare AI to electricity, which is also an enabling technology. Electricity is everywhere. We owe him the light, the heating, our electronic devices. Electricity propels these innovations without itself taking a solid form.
Similarly, artificial intelligence, through the use of algorithms leading to learning replicating human action, allows many applications. However, when isolated, it embodies no commercial “value”. The craze is not for AI per se, but rather the applications that will emerge thanks to AI.
Whether personal assistants like Siri or Google in our phones, through the proposal of “Tiger King” on Netflix, to the profiling, personalization and cyber protection that hides behind more and more of our consumer choice, artificial intelligence affects almost every aspect of our lives. And this is just the beginning. I also loved the following visualization about the impact that AI will probably have on the evolution of humanity over time:
One wonders what it should feel like and what it feels like to be on the brink of such a great technological revolution. Well, the answer is, not much. The reality looks more like this:
Carrying both admiration and fear of this technology is therefore well founded. The point is to find this balance between recognizing how much artificial intelligence applications can really improve the world, while being wary of bad motivations and oversight errors (hello biases).
What could go so wrong?
We talk a lot (too much) about AI in its most advanced conception (which we are far from reaching) and the worst forms it could take. It is true that minimizing the precautions to be taken and its importance could really have catastrophic consequences. Take the example of a robot like the one developed by the firm Boston Dynamics integrating facial recognition applications to identify and eliminate certain targets in the context of war. This creates both the potential for collateral damage reduced by highly targeted attacks rather than diffuse bombing, yet imagine the gravity if it ended up in the wrong hands.
A great example to demonstrate why AI needs supervision comes from “Tay”, Microsoft’s artificial intelligence chatbot, which was designed to speak “like a teenager”. The goal was for Tay to become more intelligent and more human by learning from conversations she had with real humans online. However, considering that many people indulge in degrading others online and that the Internet is a fairly corrupt place, in 24 hours, the sweet and innocent Tay has become racist and Nazi. Unsurprisingly it was disabled.
Here are some examples to demonstrate why AI needs supervision:
- Google’s photo recognition software marks dark-skinned people as “gorillas”;
- the voice conversion function of YouTube does not recognize female voices;
- HP facial recognition cameras cannot recognize Asians;
- Amazon classifies LGBT literature as “18+”.
Save the world with AI? No but…
The fact remains that there is not enough mention of the applications that use AI and that make our lives frankly easier and all the positive that humanity can get from this technology. Whether it is for assembly line optimization, retail customer experience or healthcare support, the potential for AI powered robotics is immense and truly promising.
For example, in Canada and everywhere else, AI scientists are mobilizing to fight the coronavirus. Yoshua Bengio appealed to all to develop an app that can track the likelihood of you being infected based on where you have been and the meetings you have had and would share this risk information with the phones of the people you meet so that their app can update their own risk assessment. In this pandemic context, artificial intelligence could truly save lives while preventing the downturn in the economy.
AI doesn’t reinvent the wheel anyway … Even if in fact, yes, in a way.
What is certain is that an innovation has never been so disruptive. Artificial intelligence, in all its forms, is here to stay. You might as well get used to it. What if, rather than fear it, we all learned to tame it a little better? You know, the question of setting up and choosing, as a consumer, well thought out applications with a positive impact.
The views expressed in this article are those of the author and do not reflect the views of the Government of Canada.