Can machines think? Or are humans machines?

Tuesday, 20 December 2016

Can machines think? Or are humans machines?

This is the last in a series of three post about some fundamental notions of AI. The objective of these series of three posts is to equip readers with sufficient understanding of where AI comes from, so they can have their own criterion when reading about the hype of AI. If you missed any of the two previous posts, you can read the first one about what Artificial Intelligence is here, and the second one on how "intelligent" can Artificial Intelligence get, here.



Machines
Figure 1: Can machines think? Or ... Are humans machines? 


Symbolic vs non-symbolic AI

This dimension for understanding AI refers to how a computer program reaches its conclusion. Symbolic AI refers to the fact that all steps are based on "symbolic" human-readable representations of the problems which use logic and search to solve problems. Expert Systems are a typical example of symbolic AI as the knowledge is encoded in IF-THEN rules which are understandable by people. NLP systems which use grammars to parse language are also symbolic AI systems. Here the symbolic representation is the grammar of the language.

The main advantage of symbolic AI is that the reasoning process can be understand by people, which is a very important factor for taking important decisions. A symbolic AI program can explain why a certain conclusion is reached and what the intermediate reasoning steps have been. This is key for using AI systems that give advice on medical diagnosis; if doctors cannot understand why an AI system comes to its conclusion, it is harder for them to accept the advice.

Non-symbolic AI systems do no manipulate a symbolic representation to find solutions to problems. Instead, they perform calculations according to some principles which have demostrated their capability to solve problems without exactly understanding how to arrive at their solutions. Examples include genetic algorithms, neural networks and deep learning. The origin of non-symbolic AI comes from the attempt to mimic the workings of the human brain; a complex network of highly interconnected cells whose electrical signal flows decide how we, humans, behave. Figure 2 illustrates the difference between a symbolic and non-symbolic representation of an apple. Obviously, the symbolic representation is easy to understand by humans, whereas the symbolic representation isn't.

Symbolic and non-symbolic representation
Figure 2: A symbolic and non-symbolic representation of an apple (source http://web.media.mit.edu/~minsky/papers/SymbolicVs.Connectionist.html).


Today, non-symbolic AI, through deep learning and other machines learning algorithms, is achieving very promising results, championed by IBM's Watson, Google's work on automatic translation (which has no understanding of the language itself, it "just" looks at co-occurring patterns), Facebook's algorithm for face recognition, self-driving cars, and the popularity of deep learning. The main disadvantage of non-symbolic AI systems is that no "normal" person can understand how those systems come to their conclusions or actions, or take their decisions. See for example Figure 2: in the left part we can understand easily why something is an apple, but looking at the right part, we cannot easily understand why the system concludes that it's an apple. When non-symbolic (aka connectionist) systems are applied to critical tasks such as medical diagnosis, self-driving cars, legal decisions, etc, understanding why they come to a certain conclusion through a human-understandable explanation is very important. In the end, in the real world, somebody needs to be accountable or liable for the decisions taken. But when an AI program takes a decision and no-one understands why, then our society has an issue (see FATML, an initiative that investigates Fairness, Accountability, and Transparency in Machine Learning).

Probably the most powerful AI systems will come from a combination of both approaches.

The final question: Can machines think? Are humans machine?

It is now clear that machines certainly can perform complex tasks that would require "thinking" if performed by people. But can computers have consciousness? Can they have, feel or express emotions? Or, are we, people, machines? After all our bodies and brains are based on a very complex "machinery" of mechanical, physical and chemical processes, that so far, nobody has fully understood. There is a research field called "computational emotions" which tries to build programs that are able to express emotions. But maybe expressing emotions is different than feeling them? (See Intentional Stance in this post).

Computers and emotions
Figure 3: Can computers express of feel emotions?

Another critical issue for the final question is whether machines can have consciousness. This is an even trickier question than whether machines can think. I will leave you with this MIT Technology Review interview with Christof Koch about "What It Will Take for Computers to Be Conscious", where he says: "Consciousness is a property of complex systems that have a particular “cause-effect” repertoire. They have a particular way of interacting with the world, such as the brain does, or in principle, such as a computer could."

In my opinion, currently, there are no scientific answers to those questions, and whatever you may think about it, is more a belief or conviction than a commonly accepted truth or a scientific result. Maybe we have to wait until 2045, which is when Ray Kurzweil predicts technological singularity to occur: the point when machines become more intelligent than humans. While this point is still far away and many believe it will never happen, it is a very intriguing theme evidenced by movies such as 2001: A Space Odyssey, A.I. (Spielberg), Ex Machina and Her, among others.

No comments:

Post a Comment