1Mby1M Virtual Accelerator AI Investor Forum: David Evans, Sentiero Ventures (Part 5)

David Evans: Most of the literature and research that I’ve read says we need to start fresh and clean. That’s part of the reason why companies like Thinking Machines have raised so much money. We can’t just take what we already have and build on top of it. If that were the case, you’d just see OpenAI and Anthropic continue to build on existing models.

But I think the longer tail, if you think about AGI, is really similar to LLMs. It starts to make tools and processes scalable that haven’t been scalable in the past. They’ve always been constrained by how many human beings we have. Globally, we’re on a downward trend in terms of population—most of the developed world is below the replacement rate. We don’t have enough people to do the work.

When you take an abundance view of the market and a positive view of what AGI could do, it reshapes the complexion of work and services based on where you’re at in the market and how scalably we can deliver those. One of the things we’re excited about, even with the current state of AI, is that you can now start delivering tools and services that were once only available to large companies. These are increasingly becoming accessible to smaller companies because we now have the scalability.

I couldn’t serve small companies before because it wasn’t profitable. But if I can automate 98% of interactions with a small enterprise, I can take them on as customers. That’s an active dialogue we’re having with many of our portfolio companies—what does down-market scale look like? When we calculate market sizing, we’re looking at segments that would have been inaccessible five years ago.

If you look at it through the lens of a positive-sum AGI future, the nature of work will change, but I don’t think it will completely supplant what we do as human beings.

Sramana Mitra: I’ll give you my take on this. There is a school of thought that there will be no work in a complete AGI world. That school of thought is also quite fleshed out.

In the development of AGI, the emphasis has been on language, right? The huge highly accelerated progress that has happened has been on LLMs and language.  In vertical AI, we see a lot of small language models doing much better work with less hallucination, less cost, and still providing tremendously effective automation value.

But there are other ways human beings learn. There’s visual learning, there is auditory learning, and there’s smell. I just recently read about a robot in Brazil that is harvesting fruits by smelling the ripeness factor with sensors. This is a good example of starting to bring in more of the five senses into learning as opposed to relying only on language or one-dimensional learning, which is not complete as far as how humans learn.

 So, I think in terms of the development of learning models in general, there’s still a lot of ground to cover, but it’s not impossible that that ground will be covered.

David Evans: Absolutely.

This segment is part 5 in the series : 1Mby1M Virtual Accelerator AI Investor Forum: David Evans, Sentiero Ventures
1 2 3 4 5 6