What does Artificial Intelligence really mean?
What do you know about AI? However, if someone would ask you to name one example of Artificial Intelligence, would you be able to identify more than those smart self-driven cars that only exist in Elon Musk’s near future?
Don’t get me wrong! I’m not against Elon Musk, technology or Artificial Intelligence, on a contrary, I think that AI is a revolution in itself and I am a big admirer of those who are working in this area and trying to improve life through technology, but I also believe that we are in the position of not knowing exactly what Artificial Intelligence really is. We tend to think AI could only exist as a human shape-shift, and the society, as a whole, unfortunately sustains this fantasy.
Of course, there are high chances that AIs will sooner or later be seen walking beside us, but that doesn’t change the fact that those concrete and human-like forms of AI don’t exist as a rule in our present days.
And that also doesn’t mean that our days are not full of other examples of AI that we are not aware of, most probably because we are stuck in those speculative ideas that depict AI as a friendly robot that will soon sit next to your lonely gramma and help her get dressed and have a normal life.
Is the definition of AI enough to understand what artificial intelligence really is?
My question is in fact related to the recently drafted European paper named “Definition of Ai — Main capabilities and scientific disciplines” which relates to “Draft Ethics Guidelines for Trustworthy AI”, both of them released for consultation by European Commission until the 1st of February 2019.
With the risk of not being so efficient, I am used not to reflect on anything before getting to perfectly know who the main character of the story really is. And whether we are talking about Ethics or not in this area, the main character of our story is definitely Artificial Intelligence. That’s why this article will focus on the definition of AI considered to be of high importance to be debated and established before addressing the necessity of Ethics in the context of Artificial Intelligence. For the consideration of this purpose, all the following opinions will strictly regard the document “Definition of Ai — Main capabilities and scientific disciplines” issued by the European Commission’s High -Level Expert Group on Artificial Intelligence on the 18th of December 2018.
Artificial Intelligence (AI) refers to systems that display intelligent behaviour by analysing their environment and taking actions — with some degree of autonomy — to achieve specific goals.
Besides the above AI’s definition, the document also presents some explanations on what the components of an AI’s system would be, by identifying its “sensors”, the “reasoning and decision making” component/s and its “actuators”. But, surprisingly, the illustration is sustained by reference to only one example of AI, particularly — “The cleaning AI system — a system that automatically cleans the floor”.
Of course, the AI system for cleaning floors is a valid example, but that doesn’t mean that it will teach you more about Artificial Intelligence than you already knew. And above that, how many future non-expert readers of this document will be able to correctly identify different types of Artificial Intelligence by bearing in mind this example only?
The definition might also be a good one, we don’t contest that at this point (although there are different opinions regarding the essential components of an AI), but the example of the cleaning AI system will definitely limit the possibility of understanding an Artificial Intelligence system outside the sphere of those hardware embedded types.
Let’s put things differently.
Appraising “intelligence” in the context of devices is a difficult task, the same is the representation of how a device would be able to analyze something and even more difficult to understand how a computer will actually decide on anything. To clarify all the aspects and to set a common knowledge base for what an AI really is, the document should firstly focus on giving modern relevant present’s day examples of systems that incorporate Artificial Intelligence, especially of the software-based ones since their presence can’t be contested. Because it’s true, there are so many examples of software-based AI systems that already transit our lives that understanding and being able to identify them is almost vital.
Another reason to focus more on the software AI than on its hardware counterpart, would be that the former category has more chances to pass unobserved or to be misunderstood, most of the people being not aware of its “intelligence”. If concepts such as “sensors” or “perception” could be easily explained and accepted in the context of an embodied AI, which uses cameras and microphones as eyes and ears to perceive and collect data that surrounds it, the “sensors” of a face recognition systems will not be identified that easy, although the risk and implications of that type of AI obviously exceed the former’s ones.
The European document correctly identifies the paper’s purpose as to serve as a useful educational starting point for people that are not AI experts, which means that the information regarding Artificial Intelligence should be communicated in such specific way as to be easily absorbed by non-experts. Because of such purpose, the information has to teach the readers first and must include all the details and explanations necessary to educate them with technical terms, because being a “non-expert” in this area, it often means to not knowing absolutely anything.
However, how could that goal of sharing information on AI be regarded as complete or even correct, if the public will not have the correct representation of how an AI system works? For example, the concept of “reasoning” was explained by reference to the previously collected images that were interpreted by the cleaning AI system in order to understand and “identify” the best action. The “actuator” was also identified as being that signal that could activate the vacuum cleaner if the action was to clean the floors.
Nevertheless, how relevant are these examples in the case of a search engine, a voice assistant or a speech recognition system? Would any of the non-expert readers be able to identify the reasoning module of the reCaptcha system correctly? Are any of them ready to explain and define the actuation for predictive customer service? Are they even aware that purchase predictions, personalization of news feeds or spam filters are indeed services that incorporate Artificial Intelligence?
The answer is most probably NO, and the reason would be that the terms are meaningless, unless they are appropriately explained and adjusted to the target audience.
The audience is still waiting for the humanoid robots but is not yet ready to recognize and deal with the contemporary Artificial Intelligence around it.