AI: Suspicions of Doping!
AI: Suspicions of Doping! Citius, Altius, Fortius, three Latin words meaning "faster, higher, stronger," form the original motto of the Olympic Games, symbolizing an endless pursuit of increased performance that some may interpret as an invitation to doping. The parallel with artificial intelligence is intriguing on multiple levels. AI is a revolution, a tireless force led by our now deified machines—computers that calculate without error, without complaint. In all fields, AI is advancing. Consider the following:
Technical Praises Chorus
Fortius: Thanks to GPUs, TPUs, and the like, computing capabilities continue to support the strength needed to train models consuming astronomical amounts of data (and electricity) within clouds to serve deep learning. The latest triumph is the Large Language Models (LLM), the core of ChatGPT, the conversational assistant. It's unfortunate that fewer people are reading these days.
Altius: Indeed, these LLMs are formidable! Their remarkable performance in mastering language, especially the purely syntactical aspect, is surprising. The illusion does the rest. The computer speaks to you, shattering the Turing test (though, in fairness, it was a hoax). If not careful, soon, these "machinic" capabilities will replicate all human cognition. Why should AI have limits?
Citius: This pace, propelling us towards the culmination of AI (perhaps also towards the end of the night—rumor has it the '80s are making a comeback), seems to be accelerating in turn. The technical repercussions of LLMs are self-sustaining in a somewhat cacophonic mix of open source and profit-driven endeavors (but there's no harm in that). And we rush forward, driven by unquestioned progress, at full speed towards... Who knows what!
Intermezzo: Returning to the Olympic motto: it is already known that the 2024 Paris edition will employ AI, working behind intelligent cameras meant to ensure the safety of the people, augmented by generous foreign tourists. Surveillance—or as some say, "security"—cannot rest, unlike controllers and athletes (perhaps one day they will need replacing too!).
Ethical Praises Chorus
The end of the recently concluded year carried a preview of the new year, largely confirmed since:
Prosperity! Fortunately, AI is not confined to ever-advancing technical feats. The business is flourishing, with mind-boggling amounts raised in recent funding rounds or invested sums. Unless measured in billions of dollars—Altius!—one doesn't truly belong to the interesting players. In the hundreds of millions of euros, one has a place in the scene or may be promoted to a small supporting role, perhaps. Of course, the naysayers will point out that an investment is only an expectation of future gains (and for the present, a very real expense). They will also note that the stock value is a highly volatile measure of a company's actual worth. But well, one must keep busy...
Happiness! Since not everyone is a trader, and there's something for every taste, AI regulation is also in full swing. The global AI summit on November 1-2, 2023, in England saw political figures discussing with researchers and leaders from inevitable big tech companies. The meeting, dedicated to the well-being of people and nations (and under the tutelage of Alan Turing—different times, different customs), led to the signing of an agreement—or was it just a declaration?—known as the Bletchley Accord, occurring around the same time as Biden's decree aiming at the beginning of regulation in the field and the issuance of a code of conduct by the G7. Things are finally clear: it was a conspiracy against the European AI Act, which will be born (?) only a month later! Citius: everyone wants to be the first to say the same thing—to make AI responsible, ethical, and secure (a pious, non-exhaustive wishlist). Even China could sign. Fortius!
Audible Dissonance or a Choir Singing Out of Tune?
Yet, from a "permanent revolution" (note the oxymoron) to a "crisis," there isn't much distance. How else to explain these recent media-political enthusiasms? Is the technical success of LLMs enough to revive old notions of great replacement, universal income (which will be "elevated," according to Elon Musk, accustomed to the discomforts of the minimum wage), and humanoid robots that could turn hostile if their AI derails?
A crisis of conscience, perhaps, as people feel dizzy contemplating the extrapolation of what tomorrow's AI could be. All the more reason to ensure what today's AI is, the AI that is being created rather than what is being spoken of by the "fortunate ones." Let's return to a bit of science and education, to the making of AI, and not just the communications about it. Embarking on a journey to the end of the night is always possible; undertaking one to the end of AI will undoubtedly prove more exciting, we hope!