CVAI London: Voice App Leaders Tease a Post-Text Era As AI Wall Fears Are Forgotten
We hosted the Cerebral Valley AI Summit in London Wednesday
At the Cerebral Valley AI Summit in London Wednesday, long forgotten was last year’s unending worrying about an impending AI scaling wall. This year’s post-training successes — showcased by models like OpenAI’s o3 — have proven that there is plenty of juice left to squeeze out of today’s model paradigm and there remains a lot of optimism that good things are still to come down the pike.
“If you were to literally freeze time right now and just let the tech that’s already been developed, models have already been developed, and techniques percolate, you would have so much change over the next five years,” Figma CEO and prolific AI investor Dylan Field said on stage.
Instead of worrying about whether model providers could keep delivering stunning results, at this year’s mid-year summit in London we spent a lot of time trying to figure out how to move beyond simply typing messages back and forth with large language models.
What is the best way to interact with a large language model?
Several speakers like Granola CEO Christopher Pedregal and Synthesia CEO Victor Riparbelli predicted the ascendance of speech, with Riparbelli painting a picture of an almost post-literate world where only hobbyists engage with text and most information is consumed through video.
Riparbelli predicted a not-too-distant future where children don’t need to learn to read. “Your kids’ kids are not going to be reading and writing. They’re going to be watching and listening almost exclusively,” Riparbelli said on stage.
Brought to you by Forethought
CX used to be built in pieces. But today, companies like Scale, Cohere, and Airtable use Forethought to connect every part of the customer journey.
Now, a team of AI agents can work across sales, onboarding, support, and retention. They understand, plan, take action, and improve.
No circles. No silos. No dead ends. This is what enterprise AI should look like.
Speech Without Speech
GV partner Tom Hulme took speech one step further, pointing to companies like Neuralink and an unnamed startup in Israel that’s “that’s doing sort of pre-vocalization or silent speech” as two examples of companies that are working to suss out someone’s intended speech before they even fully put it to their lips.
“I look at my kids. They send voice notes on WhatsApp all the time, because it’s really quick and really efficient,” Hulme said.
For his part, Field is hard at work building visual cues for getting the most out of artificial intelligence. We titled his fireside chat “Escaping the MS-DOS Era.”
Figma recently created its own no-code tool, called Figma Make, which it has also integrated into its more design-oriented website-building application. The company is helping consumers to bridge their design assets with the ease of no-code programming, especially when it comes to building product demos.
Certainly, OpenAI’s acquisition of Jony Ive’s mysterious hardware startup was in the back of people’s minds as they puzzled over what form artificial intelligence would take as we figure out this new technology.
To the extent someone was skeptical of interface-obsessed thinking, Fireworks AI CEO Lin Qiao said that the startups she saw really succeeding were doing a lot of post-training work, making sure that they got the most out of the models, not just putting a nice design on top of someone else’s unaltered large language model.
Qiao predicted that we would move on from talking about prompt engineers to “reward engineers” who will amend models for their employers and differentiate their offerings from base models.
Three European Venture Titans
Our conversation with Accel’s Philippe Botteri, GV’s Tom Hulme, and Index Ventures’ Jan Hammer.
A big thanks to our sponsors Index Ventures, Wilson Sonsini, Forethought, Crusoe, and Blue Yard!
We’ll be back in San Francisco with Deus Ex Medicina on Sept. 9 and our flagship Cerebral Valley AI Summit on Nov. 12.
Watch this newsletter for more coverage from the summit and videos from the talks.