In AI, it pays to pay close attention to what leaders of major AI labs are saying in public interviews.
Their insights can give you crucial insight into where AI is going—straight from the people shaping that future.
Recently, Aidan Gomez, co-founder and CEO of Cohere, a leading large language model company, shared his insights in an hour-long interview on the popular 20VC podcast.
What can we learn from his comments?
I got the scoop from Marketing AI Institute founder and CEO Paul Roetzer on Episode 112 of The Artificial Intelligence Show.
Before diving into the insights, it's worth understanding why Gomez's opinions carry weight in the AI community.
Here’s why he matters, says Roetzer.
According to Roetzer, a handful of key points are worth paying attention to in this interview.
Gomez confirmed that the "scaling laws" in AI—the idea that models get smarter with more compute power and data—are still holding true. However, he sees this approach as inefficient:
"It's definitely true that if you throw more compute at the model, if you make the model bigger, it'll get better. It's kind of like it's the most trustworthy way to improve models. It's also the dumbest [...] I just think it's extremely inefficient. There are much better ways."
While acknowledging the continued growth of large, general-purpose models, Gomez emphasized the importance of smaller, more efficient models designed for specific use cases:
"We live in this world of unbundled verticalized models, which are much more efficient and smaller, designed for specific use cases [...] There will be both [bigger and smaller models]."
Gomez has changed his mind about the importance of data quality in the last 12 months, saying:
"Pretty much all of the major gains that we've seen in open source space have come from data improvements. Models getting much better by taking higher quality data from the internet, better scraping algorithms, parsing those webpages, pulling out the right parts."
This focus on data quality allows for more efficient training and potentially smaller yet more capable models.
One of the most intriguing areas Gomez discussed is the development of "system two" thinking in AI—the ability for models to pause and reason through complex problems:
"The status quo with models is: I ask you a question and the model is expected to respond immediately with the right answer. That's an incredibly high burden to place on the model [...] There's this very obvious next step for models, which is you need to let them think and work through problems."
The challenge, Gomez notes, is that the internet is full of outputs, not the reasoning processes that led to those outputs. Teaching machines how to think is a major focus for Cohere.
Gomez is particularly excited about advancements in voice-based AI:
"Anyone who has tried having a voice-based conversation with one of these models, it's a stunning experience," Gomez said. "You're left in shock when you hear the model exhibiting emotion and inflection, and you hear it breathe before it speaks, and you hear its lip-smacking."
While acknowledging the fears around data security and IP loss, Gomez sees a significant shift in enterprise AI adoption:
"Last year was a lot of proof of concepts. This year, he's seeing an urgency to adopt," says Roetzer.
Gomez is bullish on the potential of AI agents.
"He says the hype is '100% justified'," says Roetzer. "These things are going to transform productivity, and they're coming. [They've] all been working on them for the last year to two years, and we're going to see a massive change there."
Gomez sees robotics as a major frontier for AI breakthroughs.
"I think robotics is the place where there will be big breakthroughs," Gomez said. "The cost needs to come down, but it's been coming down. But soon, someone's going to crack the nut of general-purpose humanoid robotics that are cheap and robust."
Perhaps most interestingly, Gomez's vision for AI is less about achieving artificial general intelligence (AGI) and more about enhancing human productivity:
"What I really care about with this technology is driving productivity for the world and making humans more effective, able to do more," he said in the interview.
Gomez's insights paint a picture of an AI future that's not just about building bigger models, but about creating more efficient, specialized, and thoughtful AI systems. From the rise of voice AI to the potential of robotics and AI agents, the next few years promise to be transformative.
For anyone interested in the future of AI, Gomez's full interview is well worth a listen to better understand where we're headed—and how we can best prepare for the future rushing our way.