What is it?
An artificial intelligence, AI, is a system built to respond to complex stimuli with coherent, seemingly logical and internally motivated responses. In short, it seems to think, performing complex creative and analytical tasks once thought only possible by humans.
What does it do?
The AI programs making recent headlines are more accurately described as machine learning models. ChatGPT from the firm OpenAI, perhaps the most famous of this new generation of programs, was created to generate natural sounding responses from real language prompts. You can chat with it or even ask it to write a limerick to respond to an RSVP:
I’m sorry I can’t come to your party
I know it will be fun and hearty
But I have to decline
For a reason benign
I’m allergic to cake and confetti
Other AI technologies can create images on command, mimic an individual’s voice or expressions, or write music.
How does it work?
Machine learning models identify patterns of relationships within massive quantities of data. A language learning model may find that the words “appetite” and “reader” often follow the word “voracious,” and that the phrase “voracious reader” is more often found than “voracious appetite” when an article also contains the word “books.” By using enough of these connections, an AI could create a legible love story about a bookseller and a gourmand.
What are some benefits?
New perspective – Machine learning models understand the world in a much different way than people, which can give us new insight into complex problems including medicine, physics and resource management.
New tools – AI provides tools for people to create, regardless of talent, training or education. Want to make an operatic rom-com comic book set on an alien planet, but your parents favoured mathematics over art class? AI can help you with your magnum opus.
What are some risks?
Accuracy – Responses are based on patterns in information, not facts. These often coincide, but it’s not guaranteed.
Information security – AI models are “black boxes” with internal networks of connections that are impossible to unravel. If you ask an AI something private, you can’t know if that information will pop up elsewhere.
What’s next?
In the near term, expect these technologies to become more advanced and less error prone. Already, these technologies are being adapted to other tasks, like identifying diseased cells, answering customer support calls or protecting computer systems from viruses. Speculation about the long-term effects is rampant. Some suspect a great remaking of the economy, with knowledge workers experiencing similar disruptions to those the manufacturing workforce faced with the introduction of robotics. Some project vast societal changes, from dystopic to utopic. It is impossible to say, but as we leave the information age and enter the AI age, we will see it happen up close.
