A venture capitalist's take on generative AI investment

Despite the economic challenges of the moment, AI -- and particularly generative AI, also known as GenAI -- is attracting big investments from venture capitalists and other investors.

Investment in AI technology will approach $200 billion worldwide by 2025, according to recent research from Goldman Sachs.

However, despite steady investments in AI startups such as Anthropic, Cohere and Hugging Face, investors are still cautious about which companies to support.

Glasswing Ventures founder and managing partner Rudina Seseri has been investing in AI technology for a long time. In this Q&A, she discusses what some of her investment criteria are.

Editor's note: This interview was edited for length and accuracy.

What are some of the changes you've seen in generative AI and general AI?

Rudina Seseri: I've been investing in AI-native companies that solve problems for enterprise security since 2010. It's a market that has been evolving.

Rudina Seseri

The first semblance of AI emerged in the 1950s. There are probably several monumental points -- one being in 2006 with the emergence of deep learning and neural nets in academia, and then around 2010 making its way to industry, and then in 2017 with the rise of transformers through GenAI.

GenAI has been in adoption since then. From the emergence of deep learning in 2006, you see a pretty steep curve in adoption of regular AI. And then, since the rise and emergence of GenAI, that curve has actually become a step function, so it's been fascinating.

We're at the peak of the hype.

As we're hitting the peak of the hype, we will go through a trough of disillusionment where AI models and applications that have not been properly constructed -- whether it's on the algorithmic side, whether it's on the data side, whether it's matching of the algorithms with the right use case -- there will be disillusionment.

From a hype and investment perspective, we're going to get into the solution phase. What's exciting to me as a long-term investor in the space is the adoption in the enterprise. So, the actual demand for AI-native technologies and platforms will only increase.

As an investor, what are your main criteria when looking to invest in companies?

Seseri: I'm really looking for exceptional teams that can execute, but that's true for any space.

Secondly, I look for unique and differentiated novel models, and data access that's unique and proprietary.

I look for outputs in the products that are editable and verifiable. It's not simply a black box, and you can go in and make tweaks to improve.

If you're going into an enterprise trying to solve a particular problem, and you can solve it without AI, don't bother.

I look for AI that's indispensable for solving a problem. What I mean by that is, if you're going into an enterprise trying to solve a particular problem, and you can solve it without AI, don't bother. Solve it without AI.

Additionally, developing the right AI MVP. What I mean by the right AI minimum valuable product is, AI is least valuable on a product's day 1 when you're just starting to train or feed data to it. After you've trained it for many, many time windows, it's performing and it's incredibly powerful. But day 1, it's not. Even so, with MVP, that first day when you're going to the customer, it has to create enough value even though it's probably not fully trained for the customer to extract true value.

How do you know a vendor you're investing in will be long-lasting?

Seseri: Ultimately, it comes down to execution. That question is the perennial question in my industry. Will the beginning founders go early or go long? Can they not just innovate in a lab, in a theoretical context, but in industry? Can they build a big business that's self-sustaining?

In my role, we're looking for companies that solve problems that are big with a big market opportunity. Therefore, they have the potential to go long and to go big.

At the end of the day, it comes down to execution by the management team.

How do you as an investor handle some of the security and privacy problems that come with generative AI?

Seseri: It's putting guardrails and developing tools to support them. For example, we will not go after companies that use facial recognition at an individual recognition level first. However, we have a company that's literally assuring that enterprises follow privacy regulations and have a security system, etc.

Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.