The engineer Mira Murati, at the head of the teams that developed the trendy “chatbot” and the Dall-e model, assures that the discussion of the present and future of artificial intelligence must involve everyone from artists, humanists and social scientists to regulators and governments, among others.
To write this article, I had many questions, so I turned to ChatGPT. I started by asking him what ChatGPT is, and he said, “It’s a great language model developed by OpenAI. It’s a type of artificial intelligence (AI) program designed to generate human-like text based on a given message or input.” We talked about that in English.
I asked him the same question in Spanish, and the last part of the answer seemed better this time, as it gives a clearer or more specific idea of the uses it can have: “ChatGPT has been used in a wide variety of applications, from virtual assistants and ‘chatbots’ to text analysis and creative text generation”.
While the answer is an understatement anyway, the “wide variety” part is serious. The problem that many see around ChatGPT is the excessive or inappropriate use that can be given to a tool like this, which has managed to pass difficult university entrance exams and has been consulted to give concepts in court rulings, among many other cases.
I asked her what it should not be used for, and her warning, of course, was similar to the one given by the woman who leads the team that created ChatGPT. She is Mira Murati, an American engineer of Indian descent born in 1988. Thirty years later, she joined OpenAI. She previously worked at Goldman Sachs, Zodiac Aerospace and Tesla, where she was product manager for the Model X.
You may be interested in: ChatGPT is now incorporated in Microsoft’s Bing search engine, here’s how you can use it
Failing to get an interview with Murati, I opted to ask ChatGPT about her. She replied that she doesn’t know everyone in the world, to give her more context. I told him I was referring to OpenAI’s CTO. She then replied that as far as she knew Greg Brockman held that position.
But no: Brockman is a co-founder of the company (which had among its creators Elon Musk and Sam Altman, current CEO) and its president since 2015. After probing him about the same in different ways, he finally explained to me that to his knowledge, with a cut-off as of September 2021, the CTO was Brockman. Perhaps the confusion stemmed from there, as Murati came into the position in 2022.
As Kevin Roose wrote in “The New York Times”, ChatGPT “does not cite its sources and has trouble incorporating updated information or events”, something that is expected from Bing chat, Microsoft’s search engine, a company that has been supporting OpenAI for several years. With this new development, they could finally dethrone Google (which also announced its own AI).
In any case, in the exercise with the “bot”, several of the points that Murati points out became evident. In an interview with “Time” magazine this month, he explained that, as it consists of “a large neural network that has been trained to predict the next word,” the challenges with ChatGPT are similar to those with similar models: basically that they can make things up.
That’s why, in its response, the system warned me to use it “with caution and avoid relying on it for critical decision making or applications where accuracy and empathy are crucial.” In other words, it is not designed to replace human judgment, since, in addition, the data used for its training may be biased.
When John Simons, who conducted the “Time” interview, asked Murati to name a song that identifies or inspires her, she chose “Paranoid Android” by Radiohead. It is the second song from the album “OK Computer” (1997), a title that alludes to the science fiction novel “The Hitchhiker’s Guide to the Galaxy” (1979), by Douglas Adams.
The expression is from the moment when the ship recognizes that it cannot defend itself against a missile attack. “OK, computer: I want full manual control,” the galactic president, in command, replies. As reviewed by Andy Greene in “Rolling Stone” magazine, that sentence marks the point in the narrative where “humans are saved by taking back control from the machines.”
“It’s not the most uplifting thing, but it’s beautiful and uplifting,” Murati commented on the song, which its author, Thom Yorke, wrote in an attempt to “reconnect” with other humans using metaphors or references to the technological world. And it is precisely that which seems key to not losing control of the machines: dialogue between human beings.
Murati suggests that it is necessary to be aware that technology is not just a matter for engineers. He acknowledges that developers are “a small group of people” who need more feedback and perspectives in view of all the social, ethical and philosophical implications that developments such as those of the team he leads generate.
“It’s important that we bring [into the discussion] different voices, such as philosophers, social scientists, artists and people in the humanities,” he said in the interview with “Time.” Among those, he adds, should also be regulators and governments.
As if that weren’t enough, when asked by Simons about a movie that lets her know or inspires her, she mentioned Stanley Kubrick’s “2001: A Space Odyssey.” “It keeps stirring my imagination,” a purpose that seems to have materialized in Dall-e, OpenAI’s other major development, which predates ChatGPT.
The system, which got its name from the mix between the surname of artist Salvador Dali and the robot from the movie “Wall-e,” not only creates images from a description, but, for example, can “imagine” the environment that an existing painting or drawing would have. “It’s an extension of the imagination without the limits of a canvas or paper,” as the CTO said on “The Daily Show.”
In another exercise of imagination, Murati has referred to the scope of a development like ChatGPT. “It has the immense potential to help us with personalized education,” he commented for “Time.” He refers to the fact that, given that we all come from different backgrounds, it doesn’t make much sense for us to learn in the same way.
“With tools like ChatGPT, you can talk endlessly to understand a concept in a way that suits your level of understanding,” he estimated. In response to the fact that the tool has been banned in educational centers (basically because it can do all the work), she insisted that the problem is not the technology, but what you do with it.
This engineer, by the way, has collaborated with spaces such as Geek Girl X, which seek to encourage the participation of women and girls in the world of technology, an industry that, in her words, has the ability to “shape society” and vice versa, but in which half of that society is underrepresented.
On AI, for many people, as Roose pointed out in his column, “there are legitimate doubts about how fast all this technology is being developed and deployed.” But they also agree that OpenAI, which has about 400 employees, has started “the AI arms race,” with young people like Murati at the helm.