August 29, 1997; 2:14 am ET. Skynet, an artificial intelligence, becomes self-aware after just 25 days of being “alive”. What follows next is a war the world has never seen before.
The action/sci-fi movie industry loves to make up stories – exaggerate facts – and, with a few exceptions, they don’t necessarily follow the laws of physics. Still, movies like “Terminator”, “Wargames”, “The Matrix” and “A Space Odyssey” are used to feed the fear we have: what could happen when artificial intelligence becomes too smart and self-aware?
Our biggest hopes and fears … for pizza?
A while ago, I read an article from the NY Times and loved the quote: “Artificial intelligence represents the manifestation of humanity’s biggest hopes and fears for technology. But for now, it can help us order Domino’s […]”. Basically, what AI is doing at the moment is augmenting our life by making it easier for us to deal with everyday decisions. In the example of Domino’s, it allows us to tell Dom, the friendly assistant in the Domino’s app, what we are hungry for instead of looking through a long item list of different pizza styles and making a decision on one of them.
The same is true for business decisions. Artificial intelligence can help us scan through large amounts of data and present an answer to a question we might have. It can draw conclusions because it’s capable of indexing and categorizing any amount of data, thus helping us to make a decision on investments, treatments, vacation destinations, etc.
The truth is …
What rarely anyone tells us is the fact that all these decisions the specific artificial intelligence is capable of making are trained into the AI upfront. Humans add all the data to the AI’s knowledgebase, humans help the AI to understand and categorize that knowledge and humans train the AI to make it more confident in the answers it delivers. In the case of a virtual assistant, for example, humans can add existing transcript data to a machine learning algorithm that then creates a first model. This model is able to understand and interpret some of the context already, but not necessarily all of it. To become more confident, it has to learn, and that literally happens while humans are talking to it.
Yes, a lot of today’s AI processes, setup, learning, etc. are automated by sophisticated algorithms or “intelligence”, but in most of these cases there are humans supervising the process, helping the AI when it is stuck and making sure that it only does what it is supposed to do.
And just in case Domino’s Dom becomes too self-aware and orders you a salad instead of pizza, we know how to deal with it – by tossing it into a pit of molten metal.