MONTREAL – Conversing with an artificial intelligence creature to understand it better: that is the challenge of the installation CHOM5KY vs CHOMSKY, presented from September 6th at the Espace ONF in Montreal.
For 25 to 30 minutes, participants in the experience wear a virtual reality headset and immerse themselves in an immersive, futuristic-looking world. You can ask questions and chat with CHOM5KY, an artificial intelligence unit created in part from the digital traces of American linguist and Massachusetts Institute of Technology (MIT) professor Noam Chomsky.
“We can talk to this Chomsky, exchange ideas with Chomsky, ask questions and see how he answers. So far, so good. But this entity also invites us to go through different phases, sometimes alone, sometimes with others, sometimes in collaborative games, to better understand what distinguishes our human intelligence from machine intelligence,” explains Sandra in an interview. Rodriguez, creator of the CHOM5KY vs CHOMSKY experiment, who herself taught at MIT for several years.
His new project aims to demystify how artificial intelligence works, at a time when this technology is taking up more and more space and at the same time raising fears.
“It is still very easy to see how (artificial intelligence) works and then understand that it is not magic. There’s no intention behind it, there’s no intelligence behind it, it’s actually just a gifted way of phrasing sentences that seems to fill a need, explains Ms. Rodriguez. If we understand how it works, we won’t be fooled or imitate anyone’s intelligence.
The virtual reality experience therefore allows us to look “under the hood” of artificial intelligence to understand how it creates reactions. CHOM5KY shows people who interact with it what other wording it could have used to answer a question or what synonyms it could have used, for example.
After the virtual reality portion of the experience, visitors are invited to learn more about artificial intelligence in a physical environment.
“During the process, CHOM5KY invites us to discover how the machines work. But the little subtle message is that it also makes us understand ourselves better,” says Ms. Rodriguez, who has been working on this project since 2016.
“One of the most digital intellectuals in the world”
Why did you choose Noam Chomsky to do this project? The idea came about when Ms. Rodriguez was a professor at MIT and a young researcher told her about his research project.
“He told me: There are so many digital traces on Noam Chomsky, I’m sure that if I can understand and imitate the way he speaks, I can understand how his brain works,” she says.
At that time, the “paradox of this proposal” hit her “hard”. “The great irony is that Chomsky’s theories say exactly the opposite,” she continues. According to Mr. Chomksy, it is impossible to imitate human intelligence because we don’t know enough about it yet and therefore we can’t imitate what we don’t know.
It was therefore optimal for the professor to use the numerous traces of “one of the most digitalized intellectuals in the world”, including conferences, interviews and some of his courses, to create an artificial intelligence that would reveal its own shortcomings.
“These traces contain his legacy, namely his belief that human creativity is what defines our freedom. Suddenly it becomes a perfect medium to create a unit of artificial intelligence, reminding us that it is limited, reminding us of its potential pitfalls, reminding us of Chomsky’s message about our capacity for wonder, what limits us as human beings what makes it unique is creativity, (and) we can’t lose it,” explains Ms. Rodriguez.
Noam Chomsky’s digital tracks are available online royalty-free. Nevertheless, Ms. Rodriguez informed Mr. Chomsky, now 94, about her project, in which he decided not to participate.
More transparency needed
During a conference on Wednesday as part of the upcoming launch of CHOM5KY vs CHOMSKY, Sasha Luccioni, ethical and sustainable artificial intelligence researcher at the company HuggingFace, emphasized the importance of companies developing artificial intelligence being more transparent.
“Companies are saying less and less about how their data was collected, how many people trained their algorithm and (how much) they were paid.” (…) Unfortunately, instead of advocating for transparency, I think this ultimately leads to regulation would (that’s a secret),” says Ms. Luccioni.
It reminds us that artificial intelligence does not arise “by magic” but, on the contrary, contains a lot of human intelligence. Software like ChatGPT relies on data people post online (e.g. news websites), and multiple people are hired to perfect the software, she points out.
Ms. Luccioni emphasized that the dissemination of artificial intelligence models would be useful for researchers, but also for understanding how this technology works in the population.
The CHOM5KY vs. CHOMSKY experience will be presented until October 15th.