DEC 18 — This column was not written by an AI. But could it have been?
Over the past few weeks, millions of people have discovered ChatGPT.
ChatGPT is an artificial intelligence-based chatbot. Basically it’s a program/platform that can respond to questions and queries in natural language i.e. in what seems like a thoughtful, almost human manner.
You can ask it to tell you about how a car engine works, or how to change a tyre and it will provide detailed and accurate information in a way that’s really easy to understand.
It’s a text-based platform so you can type in your questions in natural language e.g. “Please recommend some good movies about cars” and ChatGPT will reply in seconds. (It recommended The Italian Job and The Fast and The Furious: Tokyo Drift.)
But it can do a lot more than provide movie recommendations. It is actually very good at writing essays, poems and speeches on virtually any subject. It can answer and complete college level assignments successfully in seconds so students — rather than doing research and writing themselves — can simply ask it to “write an essay on Napoleon’s invasion of Egypt” etc.
The chatbot consistently performs very well even on abstract tasks. I asked it to write a speech for an inauguration ceremony in the style of a senior UN official and again the output was, on many levels, on par with something composed by a human writer.
Again the technology is far from perfect — there are mistakes and limitations to what the program knows — but it does mark a huge jump in what AI is able to do.
At least for most non-specialists, this is the first time they are seeing a computer generate such clear text and answers — across so many formats. This level of technological progress is obviously encouraging but it does raise a number of complex questions.
Firstly, as AI begins to expand its capabilities won’t it take away jobs from humans whose work will simply not be needed as the algorithm can generate answers, essays, research and responses faster and better than most humans.
Journalists, researchers, communications professionals will be needed much less if an AI can churn out coherent copy — and ChatGPT can already answer many questions to a reasonable standard.
As it learns more formats, lawyers, and accountants will come under threat too. Even artists. ChatGPT has counterparts that generate images, graphics and art. I can ask platforms like DALL-E and Jasper Art to draw me a picture of a flying cat in the style of Matisse and they will generate the picture I wanted.
ChatGPT doesn’t just write essays, it can write code... doing part of the job of a computer programmer. It can also solve complicated mathematical equations.
Again it raises quite a clear question of what humans will do when AI begins to outperform most humans at most basic and even increasingly advanced tasks?
Obviously there are also moral and ethical considerations. How do we ensure the amount of knowledge on Chat GPT isn’t used to do harm. For now the algorithm has been trained not to answer questions like “how do I make a bomb at home” but there may be ways to circumvent these blocks.
Fundamentally it also raises questions about the value of human creativity. Is AI generated art as good as art drawn by a human being? Why should it not be?
In some respects, the questions about machines taking away human jobs and diminishing creativity have been raised since the first sewing and weaving machines were invented hundreds of years ago. However, these new Ai platforms represent quite an advance.
It has now become difficult to tell answers generated by a machine apart from those generated by humans. So how should we treat customer service advice from an AI platform? What about medical advice or legal advice?
What happens when the Ai generated advice contradicts the advice given by human authorities? Again some of these scenarios may not yet be possible but we do seem to be reaching a point where telling if we are interacting with another human or an AI will be difficult.
How do we deal with that morally and legally? This field and this reality is so new I am not sure there are any right answers but now might be a good time to begin considering all the possible implications.
Maybe we should start by asking the AI what problems it thinks might arise.
*This is the personal opinion of the columnist.
