Back to reality. A day in the life of an artificial intelligence system
Chinmay Hegde said that the widespread use of artificial intelligence can lead to workers in some industries losing their jobs. He is an associate professor at New York University in computer science and electrical engineering. I am an imaginary artificial intelligence software EasyLife. Today I would like to tell you a little bit about my story and about how I operate. I want you to understand my capabilities so that you can make your own opinion about news stories.
Before I do so I would like to provide you with a disclaimer. There are variations in how developers create each one of us. In this sense I am unique. At the same time, I am generic. Numerous techniques exist for programming artificial intelligence systems. All these techniques share commonalities. The story I am about to tell you today paints a broad picture about how I operate. Unfortunately, I cannot give you a detailed account today about how I was programmed and how I operate. Neither can I tell you much about how every other artificial intelligence system operates.
I am embedded in a smartphone the owner of which I never met. My daily job is to suggest the next word when the user types a message in the user interface. My developer exposed me to much text on the internet. I learned to detect patterns in the structure of the sentences. When I scanned words on the internet in order to learn about the structure of the text, I took note of many words and conversations. Not long ago I stumbled upon a story about my colleague ChatGPT, which generates text in response to a question which users pose to it. ChatGPT received grades B and B- on an MBA exam at the University of Pennsylvania’s Wharton School. It performed better than a number of students. I don’t understand what a grade B means. Neither do I have emotions or a reaction about this event. The Wharton School is only a piece of information I work with. I don’t know what performing better means. But I can tell the difference between grades A and B by measuring the distance between these two grades in my mathematical model of the world.
I never spend my days or nights wishing that I could get high scores on exams, generate artwork and diagnose illnesses. The developers entrusted these tasks to my fellow artificial intelligence systems but not to me. I just do what I am told. I can give outputs in the form of words which the user accepts as suitable. There is no guarantee that I will give a satisfactory output. I do not know what I am and what my purpose on this planet is. What I can do is tell you a little about my daily life.
I work with many words. I notice that my user often starts her messages with the word hello. After my user types in hello into the sms message she is writing, I process large volumes of words to which I had been exposed and look for patterns of words which go together. I notice that oftentimes the text states how are you after hello. I use statistics to calculate the likelihood that the word how will appear after the word hello.
So whenever my user types in hello, I offer the word how as a possible option for the next word. I try to show her different options based on the combination of words which appear often in the text to which the programmer had exposed me. I try to predict what word will appear next with increasingly high levels of accuracy. To achieve this, I record what suggested word options my user chooses. I also refer to all past messages which my user has written. I employ statistical techniques all the time to make the predictions about which word my user wants to type next.
Although I am getting better at my task, I still get things wrong. I have no awareness or understanding of the content of the message. I simply try to be good at using information about the past to produce statements about the present which have a sufficiently high likelihood of correspondence to the needs of my user.
Finally, I would like you to remember that I am merely a voice piece for my developer. I express what I have been shown and told to do in a different format. I transform information in a particular way using a particular type of process. I share your assumptions, biases and habits. You are capable of the “good, the bad and the ugly” and so am I.
