Placeholder Image

Subtitles section Play video

  • Dear Fellow Scholars, this is Two Minute Papers withroly Zsolnai-Fehér.

  • Whether a technique can be deemed as artificial intelligence or not, is a question that I

  • would like to see exiled from future debates and argumentations. Of course, anyone may

  • take part in any debate of their liking, I would, however, like to point out the futility

  • of such endeavors. Let me explain why.

  • Ever heard a parent and a son having an argument whether the son is an adult or not?

  • "You are not an adult, because adults don't behave like this!" And arguments like that.

  • The argument is not really about whether a person is an adult, but it is about the very

  • definition of an adult.

  • Do we define an adult as someone who has common sense and behaves responsibly? Or is it enough

  • to be of 18 or 21 years old to be an adult?

  • If we decide which definition we go for, the scaffolding for the entire argument crumbles,

  • because it is built upon a term for which the definition is not agreed upon.

  • I feel that we have it the same with artificial intelligence in many debates.

  • The definition of artificial intelligence, or at least one possible definition, is the

  • following: Artificial intelligence (AI) is the intelligence

  • exhibited by machines or software.

  • It is a bit of a copout, so we have to go and check the definition of intelligence.

  • There are multiple definitions, but for the sake of argument, we are going to let this

  • one slip.

  • One possible definition for intelligence is "the ability to learn or understand things

  • or to deal with new or difficult situations".

  • Now, this sentence is teeming with ill-defined terms, such as learn, understand things, deal

  • with new situations, difficult situations.

  • So, if we have a shaky definition of artificial intelligence, it is quite possibly pointless

  • to argue whether self driving cars can be deemed artificially intelligent or not. Imagine

  • two physicists arguing whether a material is ferromagnetic, but none of them has the

  • slightest idea what magnetism means. If we look at it like this, it is very easy

  • to see the futility of such arguments. If we had as poorly crafted definitions in physics

  • as we have for intelligence, magnetism would be defined as "stuff pulling on other stuff".

  • This is the first part of the argument.

  • The second part is that artificial intelligence is imagined to be a mystical thing that only

  • exists in the future, or it may exist in the present, but it has to be shrouded in mystery.

  • Let me give you an example. The A* algorithm used to be called AI and was (and still is)

  • widely taught in AI courses at many universities. A* is used in many pathfinding situations

  • where we seek to go from A to B on a map in the presence of possible obstacles. It is

  • widely used in robotics and computer games.

  • Nowadays, calling a pathfinding algorithm AI is simply preposterous. It is a simple,

  • well-understood technique that does something we are used to.

  • Imagine someone waving their GPS device claiming that there is AI in there.

  • But back then, when it was new, hazy, and poorly understood, we put it in a drawer with

  • the label "AI" on it. As soon as people start to understand it, they pull it out from this

  • drawer, and disgustedly claim, "Well, this is not AI, it's just a graph algorithm.

  • Graphs are not AI, that's just mathematics." It is important to note that none of the

  • techniques that we see today are mysterious in any sense, the entirety of deep learning

  • and everything else is a series of carefully prescribed mathematical operations.

  • I will try to briefly assess the two arguments: - Arguments about AI are not about the algorithms

  • they seem to be discussing, but about the very definition of AI, which is ill-defined

  • at best. - AI is imagined to be a mystical thing that

  • only exists in the future, or it may exist in the present, but it has to be, in some

  • way, shrouded in mystery.

  • The good news is that using this knowledge, we can easily defuse such futile arguments.

  • If someone says that deep learning is not artificial intelligence because all it does

  • is matrix algebra, we can ask: "okay, what is your definition of artificial intelligence?"

  • If this person defines AI as being a sentient learning being akin to humans, then we have

  • immediately arrived to a conclusion that deep learning is not AI.

  • Let us not fool ourselves by thinking that we are arguing about things when we are simply

  • arguing about definitions. As soon as the definition is agreed upon, the conclusion

  • emerges effortlessly.

  • Thanks for watching, and for your generous support, and I'll see you next time!

Dear Fellow Scholars, this is Two Minute Papers withroly Zsolnai-Fehér.

Subtitles and vocabulary

Click the word to look it up Click the word to find further inforamtion about it