“Human beings should not do the surveillance of 40 cameras at once. We are not meant to analyse that much data.”
So says Giedrė Rajuncė, co-founder and CEO of AI Innovations, a security and defence company, powered eponymously by artificial intelligence (AI). The Lithuanian startup came from Rajuncė’s experience at another company she co-owns, ES Security.
Go deeper with GlobalData
“I had this idea several years ago,” she said. “I was sitting and thinking that human beings should not do that in the first place.”
AI Innovations aims to use AI to streamline security for factories and offices as well as provide targeting and vision systems for military drones. Verdict sat down with Rajuncė during the Enterprise AI & Tech 2024 conference in Vilnius, Lithuania, at which she was speaking on how the company’s technology can make businesses safer.
This interview has been edited for clarity and concision.
Can AI really replace people in the security sector?
So I have several different businesses. One of them is a physical security company, where actual physical security guards patrol territory, do surveillance and other things. I remember the first time when I had this idea in my head, it was several years ago. We had the monthly meeting with our team from all across Lithuania and they were discussing that this person didn’t see that, this person didn’t go there, the third person forgot to do XYZ. And you know, I was sitting and thinking human beings should not do that in the first place; we are not meant to analyse that much data. Although we had help from technology, cameras with motion detection and other things, it’s not the same as recognising if someone is just passing by your fence, or if someone is climbing the fence – the motion is the same.
How well do you really know your competitors?
Access the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain competitive edge.
Thank you!
Your download email will arrive shortly
Not ready to buy yet? Download a free sample
We are confident about the unique quality of our Company Profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the below form
By GlobalDataThis was when I said let’s do what we do, but I will try to find technology that actually helps us work and provide valuable services. I began to search for the technology at security exhibitions, defence exhibitions and so on. But in three years, I managed to find only two software options similar to what actually we try to do now, but they weren’t as functional as I wanted because we need so many different integrations. We provide services to large manufacturing sites, so they not only have surveillance, they have fire systems, access systems, cargo weight systems and so many other things.
Are all those services provided by your companies?
We do the installation, but our main service with the security company is actually to place a security guard on-site that does the surveillance, checks the people passing and coming by and so on. I think technology can and should replace human beings for some of those functions but the current technology we could use in this field is too expensive for commercial use.
For example, Palantir is doing this, but it’s too expensive and too difficult to connect this one little system extra to it. What we are doing right now is building a technology that will disrupt this part of the market at least.
Your technology is focused on cameras that use AI to figure out what they’re seeing, right?
That’s the main source of data that we are connecting to our AI and testing at the moment, but not the only one. For example, we have the situation of when transport comes and goes from the manufacturing site, so we have to read the plate numbers, and then either allow it to enter or to leave. We also have to weigh it to check it’s carrying the exact amount of cargo that they signed up for, so we have to connect several different systems here to be able to do that. We have to connect cameras, of course, we have to connect the weighing device, we have to connect the access system. So it’s not only cameras, but yes, cameras are the main source.
How confident are you in the technology at the moment in terms of error rate?
Error is something that can be fixed by training our models more, and we’re currently testing on our own cameras in our offices. So it’s easy for us to test, analyse, adapt and actually build a technologically advanced software product. But of course, it depends. The efficacy of the system depends on several different things, but mainly the data that we can input.
In our company, we’re creating and building our own IP and our own code, which is very convenient for us. We can do any adaptations easily and fast. We also don’t have to pass the data to third parties like ChatGPT or whatever because around 80% of startups are using off-the-shelf solutions.
For example, we are now talking to one large international manufacturing corporation that is based in Lithuania, and we just signed a service level agreement of what effectiveness we will reach with our technology. We are only starting with all the real-world tests, but we are going to agree 90% with them.
To what extent do you think your technology can minimise some of the mistakes we see with AI or the potential for being tricked?
I think AI is ready to replace only some of the functions, especially those that include big data. Our goal is to replace human beings for those functions, but if someone is creative enough they will be able to do almost anything at least once. With AI though, we can train it, and we can tell it that this was the situation, learn from it, and don’t allow it to happen again.
That’s of course the best-case scenario, and there might be several different ways to trick the systems, but that’s where our ability to connect different systems is useful. We can have levels of security so if someone passes the first level of security, for example, cameras, we can move on to further levels. I think that’s the beauty of it.
One of the key parts of the EU’s AI act was the partial ban on AI facial recognition – what impact will that have?
Well, I wouldn’t call it a ban or partial ban, because there are so many different caveats. What the act is basically saying now is that our company would be on the highest risk level. That will come with some more regulations for us, but there’s still a long way to go until it’s actually valid. What we Europeans like to do is we like to regulate something that we haven’t seen or tested yet. This act was prepared before the AI boom, so we haven’t even seen many of the things this technology can do and we are already putting regulations on it. I understand the value of regulations of course, but still.
There are also different kinds of options to comply with the legislation. For example, we are talking now to one international partner and their country has regulations, not on AI but still on human facial recognition. They have cameras on-site. This is the perfect case for us because we are not installing any new cameras so we don’t have to get new agreements for people to be filmed and so on and so on. Our system can blur faces so that there is no recording or filming of faces, making it easy for us to adapt.
