From the readings, what is artificial intelligence and how is it similar or different from what you consider to be human intelligence?
Are AlphaGo, Deep Blue, and Watson proof of the viability of artificial intelligence or are they just interesting tricks or gimmicks?
Is the Turing Test a valid measure of intelligence or is the Chinese Room a good counter argument?
Finally, could a computing system ever be considered a mind? Are humans just biological computers? What are the ethical implications are either idea?
Artificial intelligence (AI), simply put, is the ability for machines to act intelligently. That is to say, if a machine can act intelligently – similar to how a human would act – it has artificial intelligence. There are several varying degrees of AI, ranging from strong AI – the ability for a machine to think like a human and explain how a human thinks – to weak AI – the ability for a machine to think like a human in some small, particular way. For example, there are instances of AI all across our world today. Machines can think and reason like humans in simple ways. Google’s AlphaGo has become extremely proficient at the ancient board game of Go and can behave similarly to (well, better than) a human, but in only one very specific instance – the game of Go. AlphaGo cannot reason like a human across multiple aspects of life. As Kris Hammond so eloquently put in his article, “I may not want the system that is brilliant at figuring out where the nearest gas station is to also perform my medical diagnostics.”
Given the definitions we have above, AI is simpler to achieve than I previously thought. Machines simply need to reason like humans or give an outcome similar to what a human would choose, and it can be technically deemed as intelligent. But AI in that sense is very different from human intelligence because it is following a very narrow set of rules and generally has no ability to reason well with new, obliquely contextual information. For the sake of the rest of this reflection, I’m going to consider AI to be that kind of strong AI listed above – the ability for a computer to think and reason similarly to a human and apply that reasoning to a wide array of situations.
There have been many interesting developments in the past several years in the world of AI. We’ve seen spotlight performances from IBM’s Watson and Google’s AlphaGo, and there’s been an emerging utilization on software assistants such as Siri and Cortana. So are these kinds of machines proof of AI? No. In the case of Siri and Cortana, they simply use speech recognition software to translate text to speech and generally just perform a Google search. Sometimes they can pull out key words like “send” or “set alarm” and will perform an alternate, hard-coded action. This is not AI, but just a helpful interface to the Internet. Google’s AlphaGo is a little better, but it has such a narrow use, and the way it reasons out a good move in Go would most certainly not help me decide which classes to take next semester (if I had that option… 😥 ). IBM’s Watson is probably the closest thing we have to AI at this point, but even that simply uses sophisticated software to understand human language and perform refined encyclopedic searches.
So how do we know if we have something with AI? Some argue that the Turing Test is a good measure of whether or not something is artificially intelligent. In this test, a human user is communicating through text with another human and a computer, and the identities of both are unknown to the participant. If the user cannot correctly assess who the human is and who the computer is, the machine is said to be artificially intelligent. To me, this is an acceptable measure of intelligence. If we want a machine to be able to reason and judge like a human to consider it intelligent, then the ability to behave like one conversationally is surely a good example of human reasoning. Some argue, though, that the Chinese Room thought experiment refutes this notion. In the Chinese Room, a person, who knows no Chinese and has only English instructions on how to convert English characters to Chinese symbols, could appear to know Chinese to the outside user based on inputs and outputs. I don’t believe this is a sound counterargument, however, for the same reasons why I believe that strong AI is ultimately possible, just not yet attained.
To me, we are all just biological computers. (I will not dare to enter into the conversation of souls, which are unique to humans. The ability for a machine to reason like a human is separate from the soul, in my opinion.) We are all born with a basic set of rules, and we call those instincts. We’re generally unsure of how to control our outputs – we see babies flail and cry unintelligently – but we have the ability to receive inputs and slowly piece together inputs and outputs. Eventually we learn how to balance, walk, and speak. I believe that we could program a machine with a basic set of rules, sensors and motors for inputs and outputs, and the ability to rewrite its own code to adjust. This would give rise to human understanding and a human thought process.
While this is a cursory justification for the possibility of strong AI, I certainly believe it’s possible, just not yet attained in any sort of way. When you reduce what we know and how we act to inputs and outputs, I believe we act similarly to a very sophisticated computer. Needless to say, this is a controversial subject.
Jesse's Ethics Blog 