1Mby1M Virtual Accelerator Investor Forum: With Ankit Jain of Gradient Ventures (Part 5)

Sramana Mitra: In the case of Algorithmia, what kinds of use cases are they? It sounds like it’s a horizontal platform that could be applied to all sorts of use cases. You said it’s a Fortune 500 target market. What kinds of use cases are they going after?

Ankit Jain: The simplest way to describe Algorithmia is in the context of a company that has hundreds of algorithm developers. There’s a lot of companies that are doing this, especially in banking and finance where different people are trying out different models. That institutional knowledge doesn’t get shared across a company.

Often, in companies that have internal code-sharing infrastructure, everyone publishes their code on GitHub. Let’s say you and I work at the same company and you have written a sentiment analysis model. If I want to use a sentiment analysis model, maybe I can go on GitHub and search for sentiment analysis. You and another of our colleagues has one. In order to use that, I have to clone that code, compile it, and make sure it works within the context of my code. There’s a lot of work in getting it done.

What people end up doing is they’ll say, “I’ll get the basic idea and go build my own model.” What Algorithmia enables is for you to say, “I’ll write the code and I’ll publish it as an API endpoint.” If I want to use your code, I don’t even have to look at what the code looks like. I just have to call the API with a couple of lines and tune it as needed. That’s the enterprise use case.

The public market use case is that anybody can publish new breakthrough research onto the public marketplace and offer it for free or for a small charge. There’re research teams at many universities that are paying their bills by publishing their research as live APIs. There’s a very interesting use case. In the past, folks would just publish a paper but not publish the work. Everyone goes and tries to recreate it.

Sramana Mitra: These enterprises are not only publishing their own to one another. Within the enterprise, they’re also accessing algorithms from all these different researchers. They’re getting access to the marketplace as well.

Ankit Jain: Absolutely.

Sramana Mitra: Very cool. It reminds me a bit of Kaggle. You are relatively a young fund. You’re just starting out. I imagine there’s not yet any exit that deserves to be discussed.

Ankit Jain: We’re super young. Our investments are all happily growing.

Sramana Mitra: In the time that you have been entertaining deal flow, what trends are you seeing in AI?

Ankit Jain: There’s quite a few. I’ll go through a few of my top ones. One is taking the existing breakthroughs and applying them vertically by vertical. There isn’t a vertical where someone is not trying to apply them. We’re also seeing a few companies that are still trying to figure out what the right business models are. I think they’re very exciting.

Just stepping back for a moment, we do two things. We understand the world, we learn the world, and then we help create the world. We help create conversations like this one. We help create a product. We help create art and music. We create everything around us. So far, a lot of the breakthroughs around artificial intelligence have been around perception and around understanding the world whether it’s natural language, speech-to-text, or image understanding.

There’s a whole wave of breakthroughs that are happening on the generational side which is generating data, art, and music. There’s going to be a whole wave of companies that will fundamentally change the way in which we experience the world, the way in which we create products.

There is a research paper around creating HTML and CSS from a Photoshop file. It changes what a designer can do. There is a lot of stuff around text-to-speech. It’ll fundamentally change the way we view a lot of fields. That’s an area I’m very excited by. Another trend that I’ve seen is, over the last few years, we have seen a transition from private data centers to the cloud.

There is still this disconnect in how those two entities transition into each other, especially when it comes to AI. A lot of the models that are built assume that the data is in one place. When you’re doing image understanding, you send the image to where the model is. If you want to train a model, you bring all the training data and you build a model.

That is going to change in a pretty fundamental way as we have more data in more distributed spaces whether it is on our cellphones or whether data is on Blockchain in a distributed way where you can’t bring it all to one central place. What does that mean for the world of AI and for the world of model-building and inference?

This segment is part 5 in the series : 1Mby1M Virtual Accelerator Investor Forum: With Ankit Jain of Gradient Ventures
1 2 3 4 5 6