On the AI beat, sometimes you need to dance the mambo | The AI Beat

Fancy footwork and perfect rhythm were musts to keep up with last week’s rush of AI news.

There was OpenAI’s Monday rollout of ChatGPT plugins to Plus subscribers. Sam Altman’s Senate testimony on Tuesday, calling for an AI regulatory agency. Open source news from Meta and Google on Wednesday. A ChatGPT app made its debut in iOS on Thursday.

By Friday, when I limped to a Manhattan dinner hosted by William Falcon, CEO of LightningAI and Emad Mostaque, CEO of Stability AI, in advance of their Unite to Keep AI Open Source meetup, I was toast.

The meetup was meant to celebrate open-source AI’s recent wave of success that has led to Big Tech mulling over their potential moats while new open-source LLMs debut weekly. But while I looked forward to being able to chat about the latest AI trends in an intimate, IRL gathering with Mostaque and Falcon, as well as several AI researchers and VCs, I was even more thrilled to talk about the mambo.

Mostaque and I were the first to arrive at the dinner, but we were quickly joined by Alfredo Canziani, an assistant professor and deep learning scientist at the NYU Courant Institute of Mathematical Sciences, where he works with Meta’s chief AI scientist Yann LeCun and NYU data science professor Kyunghyun Cho (who is well-known for his work in machine translation). Canziani immediately let both of us know that he would be leaving right after the dinner to go to ballroom dance class, where he would be spending the rest of the night dancing the mambo.

During the day he was an AI researcher, he explained, but at night he was a dancer, often spending four to six hours in the evening at training and competitions. I let him know that I, too, had studied social dance in Manhattan, enjoying swing lessons for several years.

I wanted to talk more about dance — but as the other attendees arrived, including LightningAI’s Falcon, NYU researcher Kyunghyun Cho, FirstMark’s Matt Turck and Myle Ott, who left Meta to help found Character AI, I had to shift my focus to chatting about AI and open source. I kept my rhythm smooth — moving deftly from asking for the attendees’ opinions about the Senate hearing on AI regulation (Mostaque emphasized he turned down the opportunity to testify but shared written comments) and LeCun’s comments about Meta‘s open source efforts in the New York Times (he doesn’t have anything to lose, said the NYU researchers) to pausing to listen for any spilling of tea (sadly, nothing on the record).

A Tito Puente mambo beat played in my head as I sipped my wine and dug into my steak frites. I watched Falcon on his laptop as he did some kind of technical demo for Canziani. I listened as the sole other tech reporter at the dinner peppered the VCs with questions. I wished it was a little quieter (at the table and in my head) so I could hear Mostaque’s rhapsodic predictions more easily as well as Cho’s dry, hype-free take.

When we left to walk a couple of blocks to the Lightning/Stability meetup (except for Cho, who reminded us all that he has a life, and Canziani, who was off to dance class), I couldn’t help but put a slight mambo beat to my steps — quick-quick-slow, quick-quick-slow — as I chatted with Turck about the AI scene in New York City and isn’t it nice to have neighbors who work in other industries besides tech?

At the event venue, hundreds of open source developers, founders in stealth, investors and researchers were already celebrating to thumping music and enjoying a vast view of the Hudson. By the time Mostaque gave a short, signature speech to audience cheers about how everyone would soon have an open-source AI model in their pocket — “free models for everyone!” — I was still thinking about the mambo.

Whenever you social dance, whether the mambo, salsa, foxtrot or swing — you have to both anticipate what’s coming next and be relaxed enough to go with the flow. You have to keep the beat with quick footwork while still moving your hips and staying energetic. You have to learn new moves as well as practice the ones you already know.

That’s how it is for those of us covering the AI beat. That is, going with the flow of what has become a nearly-unsustainable news cycle, while keeping an eye on what’s coming next. Keeping the rhythm of daily news takes while also moving with the biggest trends. Being ready to shift to another conversation when you’re in the middle of one. Learning what’s new as well as building on what you already know.

This week, I’m ready to keep dancing to the AI beat.