Barely put online last Wednesday, the new ChatGPT artificial intelligence is already disturbing. A lot. Its claim is to produce, in response to a very specific question, dense and detailed texts rivaling in their form those of post-secondary essays. Is this the end of end-of-term work?
What is ChatGPT? Let’s ask him the question directly. “ChatGPT is an artificial intelligence application that allows you to chat with a robot in a natural way”, responds the AI. It is a language processing model created by research firm OpenAI that uses the latest human language understanding technology to understand the meaning of words and sentences and generate responses based on context.
His field of expertise? The Web, in most of its languages. The AI can write short essays, produce a tailor-made cover letter that will appeal to future bosses every time, and solve relatively simple mathematical equations. We have seen examples of plugins for the WordPress website manager programmed from scratch by ChatGPT. You just have to ask.
It is a worrying potential. Along with programmers, teachers reacted strongly to the arrival of ChatGPT. Professor of the Institute of Educational Technologies of the London School of Economics and Political Science Mike Sharples warned before the arrival of ChatGPT of the impact of these technologies on the school: “Students will use AI to write their works. Teachers will use AI to correct them. Nobody learns, nobody wins. If there’s a time to rethink this model, it’s now. »
And not just the school model, but all those who this technology is likely to influence, adds Montreal AI expert and creator of the Waverly application, Philippe Beaudoin. “You can’t put the genie back in his lamp. Framing AI is like saying we’re going to frame mathematics. That does not make any sense. On the contrary, we must provide access so that it does not end up only in the hands of a few technological giants. »
Deceptive truthfulness
ChatGPT is designed to be easy to use and accessible to everyone. You can chat with the AI about any topic, and it will try to respond to you in a consistent and natural way. You can also ask him questions on specific topics and he will try to provide you with precise and detailed answers. The tool is online, accessible free of charge, subject to registration with OpenAI.
To the question “who is René Lévesque”, ChatGPT answers in less than ten seconds: “René Lévesque was a Quebec politician, known for having founded and led the Parti Québécois and for having been Premier of Quebec from 1976 to 1985. He was a staunch defender of Quebec independence and played a pivotal role in bringing about the referendum on Quebec independence in 1980. Lévesque died in 1987.”
Whether Jean-Jacques Rousseau believed in the separation of powers within the state, ChatGPT is even more affirmative. “Yes, Jean-Jacques Rousseau believed in the separation of powers. In his work “On the social contract”, Rousseau defends the idea that the separation of powers is essential to avoid the abuse of power and to guarantee the freedom and equality of all citizens. According to him, the different powers of the state — legislative, executive and judicial — should be separate and independent from each other, in order to ensure that no one can monopolize all the power and to preserve the rights of every citizen. »
It’s simple, direct, effective. That’s more than the passing grade for any Secondary 4 course. But it can be completely off the mark. Because ChatGPT has its limits. Let’s ask ChatGPT if Hobbes also believed in the separation of powers: “Yes, Thomas Hobbes believed in the concept of separation of powers. In his book “Leviathan”, he defends a strong central government, but he also argued that a single leader should not control all aspects of the state. »
However, this is false. In matters of political philosophy, Rousseau is generally opposed to Hobbes, who lived a century earlier, because the latter was actually a follower of absolutism. The ideal political system for him is a state headed by a single leader.
Rethinking school… and more
It is a problem. Because the apparent veracity of ChatGPT’s words is likely to deceive more than one student who is a little lazy or mischievous (or both) from high school, CEGEP or even university. In computing, in any case, it worries a lot. The popular Stack Overflow programmer forum has decided to ban it from its discussions.
“The main problem is that even though his answers are likely to be wrong, they usually look correct. They are also very easy to produce by people who have neither the expertise nor the desire to check whether the answer provided is factual. »
The fear is the same on the teaching side. And nothing can. The development of these applications like GPT-3 and ChatGPT is accelerating and everyone is struggling to keep up. DALL-E on the digital image production side is another example.
“You have to realize that we have no idea yet of what is coming,” notes Philippe Beaudoin. “You have to find the positive. For example, will we be able to create tailor-made study plans? »
There is no doubt that specialized versions of a tool like ChatGPT will make specific areas of activity more efficient, concludes Philippe Beaudoin. In fact, the solution to the lack of guidance may lie in the technology itself: probably just ask ChatGPT…