AI and Big DataBreaking News

Nobody understands how AI works

No one knows now AI really works an MIT reviewer states. A little while ago, at the beginning of 2024, we released a piece stating that the AI revolution would not take place. We understand that this was a very bold move. A few readers laughed, most dismissed the idea and when on ChatGPTing (or ChatGPTeing I’m not sure about the spelling, GenAI hasn’t yet hallucinated it). Yet a short piece by MIT is hinting that, maybe, our hunch wasn’t so biased after all.   

Nobody Understands How AI Works

How AI works
How AI works is a mystery according to MIT reviewer Melissa Heikkilä — Image produced with Midjourney aimed at representing generative AI going haywire. Quite frightening it seems.

Tech companies are rushing AI-powered products to launch, despite extensive evidence that they are hard to control and often behave in unpredictable ways. This weird behavior happens because nobody knows exactly how—or why—deep learning, the fundamental technology behind today’s AI boom, works. It’s one of the biggest puzzles in AI. My colleague Will Douglas Heaven just published a piece where he dives into it.

Last week I was talking with a lady friend who is working for a software vendor a stone’s throw from my office suite. As I was on my way to filling in the kettle, I decided to stop by and say hello. As we exchanged she explained to me that they were in the process of injecting AI all over the place, therefore revolutionising the way that solicitors, legal advisors and lawyers are working, providing insights that never were available before. The future is bright.

As Melissa Heikkilä has it, tech companies are indeed rushing headlong into something they do not quite master. The good news is that nobody really does. You won’t feel alone. Even Geoffrey Hinton seems lost.

The biggest mystery is how large language models such as Gemini and OpenAI’s GPT-4 can learn to do something they were not taught to do. You can train a language model on math problems in English and then show it French literature, and from that, it can learn to solve math problems in French. These abilities fly in the face of classical statistics, which provide our best set of explanations for how predictive models should behave, Will writes. Read more here.

How AI works
Nobody understands how AI works, MIT reviewer says. Even Hinton agrees… – Image produced with Midjourney (the character’s hands are as always very strange)

Unfortunately, though if one doesn’t understand how AI works, one grasps very well how humans do. They like pining for dreams and are always attracted to the improbable. Humans love their pipe dreams, in a way it’s reassuring, this is what makes us so human. We are downright erratic, in a way we wouldn’t permit our machines to be.

Read the MIT piece

Yann Gourvennec
Follow me
Latest posts by Yann Gourvennec (see all)

Yann Gourvennec

Yann Gourvennec created in 1996. He is a speaker and author of 6 books. In 2014 he went from intrapreneur to entrepreneur, when he created his digital marketing agency. ———————————————————— Yann Gourvennec a créé en 1996. Il est conférencier et auteur de 6 livres. En 2014, il est passé d'intrapreneur à entrepreneur en créant son agence de marketing numérique. More »


Back to top button

Discover more from Marketing and Innovation

Subscribe now to keep reading and get access to the full archive.

Continue reading