How advanced foundation models can expand the capabilities of AI (and other predictions for 2020)

What’s the point of reading AI predictions by someone who is not Sam Altman or an AI PhD? Early stage VCs work with people who are far smarter than we are, and they use AI to the bleeding edge. This allows us to get a glimpse of the future.

I have worked in the tech industry for over 30 years and have never seen technology have such a rapid impact on so many levels. It has changed my expectations about how teams are built and how companies deliver their value.

With that in mind, I’ve made some predictions for 2025.

We try to contain AI

Mustafa Suleyman’s digital species corollary is a great way to frame the question of AI containment. AI products today suggest many possible actions. AI will increasingly take action on our behalf. We must define what we do not wish it to[do].

Expect to hear a lot more cage rattling. If we use California’s failed SB1047 law as a bellwether for Elon Musk’s pro-regulation and Trump’s anti-regulation positions, then the penalties will be offset. As the industry tries to “self-regulate” foundation model safety in order to avoid government regulation, expect things like Anthropic’s constitution model on steroids. Expect a boom in AI security startups that will help with cyber containment.

Foundation models as brand DNA.

You may remember “Intel Inside”a large campaign. The foundation models will give brand power to the products that build on them. This goes beyond the tone of voice which, according Claude, today looks like this:

• GPT-4 : The premium, measured advisor.
• Claude, the ethical and careful advisor. Llama is the flexible, feisty generalist.

Tone can be adjusted. It is not clear how understanding and recall are applied in a product model. Models will shape how products “feel.”

AI for the Nobel Prize

No, not quite. But the ability to iterate on scientific discovery will increase dramatically. AI and humans are working together to make breakthroughs today. Vinod Khosla is right: AI scientists – AI agents who are hyper-specialists – will soon regularly outperform their human counterparts when it comes to discovering insights.

It’s because AI scientists do not need breaks. They have an infinite amount of patience for iteration. They have perfect memories for data. They can work 24 hours a day. You can see the first signs of this work ethic in medical discoveries and healthcare breakthroughs. A slew startups are helping companies to take advantage of these advances.

Audio interfaces redefine “hands-free”

Alexa, Siri, and Google Assistant are all v1.0 audio interactions. Audio — with the context of visuals and texts — will spin out whole new product experiences which will recalibrate standards.

Specialized foundation models that only play audio are racing to keep up with the larger models. Meta’s Smart Glasses have already made huge strides since Google Glass. Hands-free operation will take on a new meaning and expectation.

We have crossed the trust tipping points with gen AI.

I have a good understanding of directions. It only took a few traffic jams that I didn’t even know about, but my mapping app knew about, before I finally gave in and trusted that mapping apps are better at navigating than me.

This will also happen with generative AI. Experts still produce superior results. This gap will close and we’ll accept more of the outputs from gen AI, assuming that they’re at least equal and sometimes even better than experts. They will be credited in the same manner as we credit those we quote. Consider the brand section in this article.

We pay more for hand-craftsmanship.

As the threshold for creating new designs is lower — Dall’E and other tools create them in seconds — goods that are not digital but show evidence of human creativity will gain in value. Think about original photography, visual design or ceramic mugs that were not stamped out in a factory by robots.

Although my adult children love vinyl and VCRs, this goes beyond being vintage. In a digital world, analog and physical things will become more meaningful, especially to the Gen Z and Gen Alpha generation.

‘Humans Found Here’

Many of my friends like Waymo, because they don’t have to speak to anyone. When it comes down to it, the ability of a human to exercise judgment based on black box logic is important. The question of which approach is more valuable — AI or human-forward? — will be a dividing point.

Expect a backlash in areas such as customer service, where the rush towards AI will frustrate many. Humans-in the-loop systems will be preferred to fully autonomous ones. Companies (and consumers), will value and promote this distinction.

What a year…

As I end 2024, I am humbled and expect a lot of chaos to occur in 2025. I’m optimistic and take heart from the fact that there are still many smart people around me who are trying to build great businesses and make things better. I wish you joy, humility and a lot of grace for the new year.

Martina Lauchengco works at Costanoa Ventures.