Interview

AI Innovations wants to use semi-autonomous drones to save lives in Ukraine

AI vision company AI Innovations believes the risks of the tech must be balanced against its life-saving potential. Isaac Hanson speaks to CEO Giedrė Rajuncė.

Giedrė Rajuncė, CEO and co-founder of AI Innovations

Following Russia’s invasion of Ukraine, the Baltic region has seen an explosion in dual-use companies. These businesses sell to military clients alongside the civilian sector, both helping the war effort and profiting from the huge demand for military technology it has sparked. 

One such company is Lithuania’s AI Innovations, which aims to use artificial intelligence (AI) to enhance security systems for factories as well as drone vision for use in the military. During the Enterprise AI & Tech 2024 conference in Vilnius, Army Technology spoke to the company’s co-founder and CEO Giedrė Rajuncė about how AI could reduce the losses in Ukraine. 

This interview has been edited for clarity and concision and follows questions about business security.

Isaac Hanson: What does your company's drone vision look like?

Giedrė Rajuncė: We are true believers in business-to-government friendship, and I think the best timing for that is now. What we see are dual-use funds, defence funds. We already have two in our small country. We are proud to be one of the companies that actually managed – at least at the minimum-viable-product stage – to squeeze our technology into a cheap $9 camera and try to help drone operators to be more effective. 

We're building cheap semi-autonomous drone software because we think that technology must be affordable today. Our goal is to make the last from 100m to 300m of the drone flight autonomous and to help operators easily and effectively hit their targets. We’re building this for military drones, including those using first-person view. 

Isaac Hanson: In terms of partnerships, who are you in talks with currently?

Giedrė Rajuncė: ​​​​​​​We’re currently talking to one of our Lithuanian sponsors, a dual-use defence fund that’s part of NATO’s DIANA program. Hopefully, we will be able to land this investment to grow faster and further. 

We are also talking to different people in the military sector. I cannot tell you a lot but yes, of course [we are in talks with militaries]. You know, I'm a business person. I am a huge believer in first of all finding the actual needs of the customers and then building a product. I don't want to do this vice versa where you think of a new life-saving technology then we present it to a military and they say that's not what we need, we actually need something else. 

So, we talked to many different people, and then we did internal analysis of our technology to see if we could make it affordable. This was a very important thing for us because we didn't want to be another Palantir where you have to do the government-level purchase for several million and then it may or may not get used. We want to make it affordable and we only want it to reach the field if it works. That's our major goal. 

Isaac Hanson: Is the software something that can be added to pre-existing drones, or does it need to be baked into new ones?

Giedrė Rajuncė: We sell the technology inside a very small computer, lightweight enough for a drone to carry, so that could be attached to existing or new drones as an add-on.

Isaac Hanson: You offer smart target selection, which is a contentious technology. Are you concerned about the use of AI in targeting?

Giedrė Rajuncė: ​​​​​​​Yes, of course. I had to have huge discussions first of all with myself, then with the team, with partners, because this is a very sensitive topic for everybody, especially for dual use. We're civilians, business people, technology creators and so on. So, we spent lots of time here and what we agreed is to do several things. First of all, to have extremely significant and detailed end-user checks. This will be a very strict process. 

Then, we are building our software with all possible safety items in it, so that the enemy won’t be able to add any new function. For target selection, we are working with targets, not people. So that's where and how we are training our AI. 

Isaac Hanson: Is there much danger of an enemy combatant like Russia intercepting these drones and replicating the technology?

Giedrė Rajuncė: ​​​​​​​With technology, you can on a theoretical level replicate anything. Practically, though, we are going to make it very, very, very difficult to do. It's early to talk about some kind of numbers because we are only at the pre-minimum-viable-product stage. I cannot tell you anything further, but I can tell you that we are not comfortable enough to show the technology to anyone, even for a demo, before we have all the safety features in it. 

Another thing is we would be very naive to think that bad people are not going to do bad things. Russia uses enormous amounts of AI-based drones now. [The number of AI drones used by Russia is not known, but believed to be small – ed]. Many bad things are happening already, so we have several ideas in mind for how to prevent that happening. 

Isaac Hanson: Ultimately, do you think that the technology is worth the risk?

Giedrė Rajuncė: ​​​​​​​Yes. What we are seeing in Ukraine is a technologically advanced army, especially with drones, fighting not that advanced of an army and that costs lives. Why not use technology to prevent that? I understand the risks of course, but the fight is not fair at the moment. I always try to remind people to actually evaluate questions with numbers, for example, how many people are being killed by the AI drones of Russia in Ukraine? This is a numbers question, not a values question. 

Fear of technology, especially of AI, takes so much energy and time for some of us and we haven't got that time.