Meetups for people turning AI ideas to into tangible solutions
I caught up with Benaich after the event and asked him why he started doing these meetups.
“I started hosting AI meetups in 2015 starting in London,” he told me. “I was interested in applications of machine learning research and startups and couldn’t find a forum with people who were interested in the same thing.”
Benaich was particularly interested in connecting with people who had taken ideas from the “I’m interested in this” stage to “I’m building this” stage. Since then, it’s been a passion project for him. “Being at the frontier is important to understand where things might go and what works and what doesn't,” he added.
But that doesn’t mean Benaich’s interest in AI is purely academic. Air Street Capital is a venture capital firm that specializes in AI-first companies.
“From an investing standpoint, if you want to do the job properly, you need to be a contributor and be active in that ecosystem rather than from our proverbial 75th floor tower,” said Benaich. “I like to be on the ground level, be around peers and make contributions of our own, like our State of AI Report.”
https://delivery-p141552-e1488202.adobeaemcloud.com/adobe/assets/urn:aaid:aem:d50d6d45-f00b-41cf-8fa7-808d1def3224/as/celonis_air_street_nyc_ai_meetup_group.png
From left to right: Marc Kinast, Akshat Bubna, Nathan Benaich, Roshan Rao, Laurens van der Maaten, Eugenio Cassiano, and Cody Blakeney.
I asked Benaich what stands out or still surprises him about these events, having done dozens of them over the years.
“I think the transferability of the ideas and techniques is pretty astounding,” he said. “Today, we saw large models being applied to next word prediction, then being applied to next protein prediction, and then to next image prediction. It's amazing how similar ideas and system designs can be applicable in different domains, which you might think are very different.”
Benaich also shared the AI advancement he’s most excited to see come to fruition in the near future. He said:
“The number one thing that contributors in AI are most excited about is this multi-step reasoning and a multi-step action taking, which is of course very relevant to what Celonis does. This idea that I can tell a system in quite vague terms to go do something and it figures out how to do it and pings me when it's not sure and tells me about the different options and interfaces with lots of different software products. That is pretty holy grail right now.”
Looking further out to the frontier of AI research, Benaich said there's still a lot of debate as to whether large models can invent new things and develop new knowledge.
“Knowledge is in some ways a recombination of different facts and remixing things,” he told me. “And in a way you could argue, as we've seen in biology, some of these systems are generating new proteins that have new functionality. Some people who scrutinize them will say, ‘Oh, but the sequence, when you look at similar sequences across the genome, is not that different. I could have figured that out.’ But the thing is, you didn’t. So on the science side, I'm most excited about the origination of new ideas and testing new hypotheses.”