How advanced foundation models will expand what AI can do (and other predictions for 2025)

Why read AI predictions from not-Sam-Altman and not-AI-PhD? Well, young venture capitalists spend their days working with people much smarter than us, who are using AI at the cutting edge of what’s possible. This allows us to get an idea of ​​what the future might hold.

In my 30 years of experience in technology, I have never seen a technology – particularly advancements in and around foundation models – have so much impact, in so many ways, and so quickly. It reframed my expectations for how teams are built and how companies generate value.

With that in mind, here are my predictions for 2025.

We are trying to contain AI

I love the one by Mustafa Suleyman corollary of digital species for AI because it correctly poses the question of confinement. Much of today’s AI in products suggests courses of action. But increasingly, AI will act on our behalf. We must frame what we don’t do it I want that to be the case.

Expect a lot of loud cage noises in this area. If we use the failure of California’s SB1047 legislation as an indicator, Elon Musk’s pro-regulation stances and Trump’s anti-regulation stances will offset the sanctions. Look for things like Anthropic’s incorporation model on steroids as the industry attempts to “self-regulate” the safety of the foundation model to avoid government regulation. Likewise, expect a flourishing of AI security startups to contribute to forms of cyber containment.

Foundation models as brand DNA

Some of you may remember the days of “Intel Inside” as a big campaign. Core models will increase the brand power of the products that build on them. This goes beyond tone of voice, which Claude says today sounds like this:

• GPT-4: The premium and measured consultant.
• Claude: The prudent and ethical advisor.
• Lama: The scrappy, flexible generalist.

The tone is adjustable. The way comprehension and recall are applied in a model is not. Models will inevitably shape the way products “feel”.

AI for the Nobel

Not quite, but the ability to iterate in scientific discovery will increase dramatically. Today, AI enables advances in collaboration with humans. I agree with Vinod Khosla on this subject: AI scientists – AI agents who are hyper-specialists – will soon regularly outperform their human counterparts in information discovery.

Indeed, AI scientists don’t need breaks. They have infinite patience for iterations. They have perfect memory for data. And they can work 24/7. Get the first glimpses of this relentless work ethic in medical discoveries and healthcare advancements, with a slew of startups helping companies profit from them.

Audio interfaces redefine “hands-free”

If Alexa, Siri, and Google Assistant were all v1.0 audio interfaces, audio – with the added context of visuals and text – would create entirely new product experiences that would recalibrate our standards.

Specialized audio-only base models race with larger base models to push what’s possible here. We’re already seeing huge progress in Meta’s smart glasses and how far they’ve come compared to Google Glass. “Hands-free” operation will have a whole new meaning — and new expectations.

We are crossing the tipping point of trust with Generation AI

I have a good sense of direction. It only took me getting stuck in a few traffic jams that I didn’t know existed – but my mapping app did – before I gave in, finally trusting that the mapping apps know the routes better than I do.

The same thing will happen with generative AI results. Currently, experts still achieve superior results. This gap will narrow and we will capitulate to more results from Gen AI, assuming they are at least on par and sometimes better than even the experts. They will get credit the same way we cite those we cite. Concrete example: see the brand section of this article.

We pay a premium for craftsmanship

With the creation threshold lower – Dall-E and a host of other tools create new designs in seconds – non-digital goods with evidence of human creation will increase in value. Think original photography or visual design, but also that ceramic mug made by an artist, not stamped by a robot in a factory.

My adult children marvel at VCRs and love vinyl, but it goes beyond vintage being cool. In such a digital world, the analog and the physical take on a meaning that will surprise, especially for Generation Z and Gen Alpha.

“Humans found here”

Many of my friends like Waymo because it doesn’t require them to talk to anyone. But when push comes to shove, the ability to interact with a human who can exercise judgment not based on black box logic is important. Which approach – AI or forward-thinking – has more value will become a dividing line.

Expect backlash in areas like customer service, where the rush to AI will frustrate many. Involved humans will be considered preferable to fully autonomous systems. Businesses (and consumers) will promote and value this distinction.

What a year…

I’m ending 2024 with humility and expecting a lot of chaos in 2025. But I’m optimistic and rejoicing in the fact that I’m still surrounded by really smart people who are trying to build great companies and make things better. I wish you humility, joy and much grace in the year to come.

Martina Lauchengco is a partner at Costanoa Ventures.