Meet the founder combining AI, pilots, and the metaverse
A colourful conversation with Diana Deca, CEO of Neurobotx.
Metaverse-powered AI pilots.
This is what Cohort 3 company, Neurobotx, is creating.
Data collection is crucial in training artificial intelligent (AI) algorithms. Yet data collection is often a challenging blocker to the entire process.
While large quantities of flight data are required to train AI algorithms capable of autonomous flight, the aerospace and aviation industry does not yet have the infrastructure to effectively extract and utilise existing flight data. “It’s very expensive to get data”, says Neurobotx CEO & Founder, Diana Deca.
Neurobotx is a neuromorphic AI platform for autonomous navigation, leveraging human pilot data to train AI pilots. With a PhD in neuroscience, Diana Deca is pushing the boundaries of artificial intelligence through an unorthodox approach of intertwining humans and machines. Fascinated by the way human brains work, Deca is exploring how we can train autonomous navigation algorithms in the same way the brain approaches spatial navigation problems, an approach called neuromorphic computing.
“I want to take that idea from the brain and bring it into Boeing, BMW, and Airbus, and have autonomous planes and cars, based on how the brain works.”
In this interview, we tap into Diana’s excitement for this field and imagine a world where aircraft are flown by machines and only machines. She starts by unpacking the autonomy problem faced by the aerospace industry, urging industry and academia to look towards adjacent fields (👾 gaming) for inspiration in solving critical challenges and draws from her experience to spotlight ways to effectively bridge the university-industry gap.
We also explore the role the metaverse could play in revolutionising data collection in AI. Deca says, “I think that simulation and neuromorphic computing are going to go hand in hand in the future for autonomy because there will be a lot of data, and there needs to be a more complex way to process it.”.
Finally, Deca shares a teaser about what’s next for Neurobotx, her prediction for autonomous flight, and her experience being part of Aerospace Xelerated’s Cohort 3.
What’s Neurobotx?
In one sentence: Neurobotx is developing metaverse-powered AI pilots.
At Neurobotx, we train human and AI pilots together in a simulation. The aim of our platform is twofold:
- On one hand, we want to use an Oculus VR game to get data from citizen pilots
- We want to use that data to help large corporations like Boeing power their level 5 and 6 autonomy to build autonomous AI pilots
What problem are you trying to solve?
It’s the autonomy problem and the fact that it’s very expensive to get data. So by using simulation, we provide a lot of data that otherwise couldn’t be obtained.
It will help companies push different levels of autonomy for regulators, which is a big hurdle. Most importantly, it can bring revenue into the field of aerospace using the new model of simulation and VR based gaming.
Do you see many challenges in the aerospace industry that relate to how data is collected for autonomous algorithms?
Most of the companies we’ve spoken with at Aerospaced Xelerated find this to be very interesting. It is obviously very novel, I mean, the metaverse itself is so novel that even people at Meta are not sure what it will look like.
At the same time, this is a great opportunity for us as a startup, because we have the pace to push this, we have the right team, and we can plug into larger corporations to bring that innovation, without them having to do it internally. So this makes sense to them as a potential investment and collaboration so they can keep up without having to change the entire structure.
“If we’re going to do large scale simulations, we’re looking at enormous amounts of data that need to be processed in order to achieve autonomy.”
You have a PhD in neuroscience! Could you share more about your academic experience and how that helped you start Neurobotx?
During my PhD in neuroscience, my focus was on the parts of the brain that deal with spatial navigation. I was lucky enough to work with the two Nobel laureates that discovered the navigation of the brain. And I was fascinated by it, I think it’s beautiful.
I also realised that I do not want this to stay in academia. I wanted to take that idea from the brain and bring it into Boeing, BMW, and Airbus, and have autonomous planes and cars, based on how the brain works.
What is the greatest learning from your time as a founder?
Being teachable at all times. In academia, there seems to be a certain way that is approved, but things are changing, especially in parts of academia that are connected to industry, like AI, robotics, and now neuroscience.
I’ve learned so much from business angels, finance, legal, and business development teams, so so much so many books that I read in the meantime, following their advice.
And this helped me achieve what I wanted to do initially, which was to bring that abstract idea about how the brain works into the industry.
So I think that’s probably the most important part to be teachable. And be curious. Even when one is super-specialised.
From your perspective, what is required to bring academic research into industry in a way that can be commercialised? How do we overcome and bridge that research-industry gap?
Great question. I’m now expanding back into the research lab as Neurobotx is doing more R&D. But at the same time, we’re also moving more into industry by bringing researchers into it.
There are many levels — from emotional to technical to industry-related. However, I think it starts with the funding. One strong argument that I was able to use to create partnerships, was that it is very difficult to obtain government funding for this type of research, which in AI and robotics can be very expensive. Bringing in funding from alternate sources apart from governments is great for professors because this frees up more of their time from applying for grants so they can focus on their students and the research again. By bringing in industry funding, researchers can do more science. This has been one of the main arguments.
I think it’s also about empathy, on both sides, similar to how we’re trying to bridge the sales and technical teams within startups. Because I think both parts can develop in their roles by learning from each other, as opposed to, assigning roles to each other.
Neuromorphic seems to be a very new and cutting-edge concept. What is neuromorphic computing?
This is a great segue from the last question about bridging academia and industry because neuromorphic computing has been developing in research for 30 years. It only penetrated the market 5 years ago by two startups who came out of Zurich with neuromorphic cameras. And it was only last year that the cameras went into mass production.
So it’s basically just starting in industry, but it’s coming in really fast with 30 years’ worth of research going, so it’s more a matter of business development and trying to find the right use cases.
So, what is neuromorphic computing? It’s an umbrella term, a bit like autonomy and AI. Essentially we are taking insights from how the brain works to help us produce something that has benefits for industry. For example, more sustainable energy-wise, or faster and more efficient.
The way neuromorphic cameras work, for example, they only extract the pixels that are relevant out of every frame. Our brains work very much like this. For example, when you look at me now — you’re not processing every pixel in the background, because your brain already knows that it’s not relevant to our discussion. Instead, you’re focusing on my facial expressions while I’m talking, because you’re listening and looking at me at the same time, and you’re already perceiving in that way. So most of the parts of the images that you’re looking at now are not even entering the rest of the brain, it’s basically stopping just behind the eyes. So, we’re taking some part of that.
This technology of neuromorphic perception is allowing us to detect incoming obstacles at microsecond resolution in defence with AI pilots, which is something that only neuromorphic can do.
What are your predictions for autonomous flight, or AI and autonomy in the near future? What exciting things would you like to see more of in the next 5 years?
As the whole world focuses on AI and autonomy, I’m obviously interested in the neuromorphic approach because of the connection with the metaverse. If we’re going to do large scale simulations, we’re looking at enormous amounts of data that need to be processed in order to achieve autonomy. So now that I explained neuromorphic perception and why it saves so much space, it makes sense for them to work in conjunction. Actually, there was a great talk from Yann LeCun about why neuromorphic is the future, and obviously because he’s the VP & Chief AI Scientist at Meta. But I do think that simulation and neuromorphic computing are going to go hand in hand in the future for autonomy because there will be a lot of data, and there needs to be a more complex way to process it.
On that note, what’s next for Neurobotx? Could you give a slight teaser on what you’re working on right now?
Oh, I am very excited about this and this is a direct result of many of the discussions we had at Aerospace Xelerated! Soon, we are going to launch our first VR product that people will be able to test. There is a long long waiting list of people that cannot wait to try it out on Oculus, and hopefully, it will also help the field to start getting the data that they need. And this is based on the things that we have noticed — access to data, helping people go through regulation faster, and so forth. We improved our product-market fit and I cannot wait for the app to be released.
Finally, how was your experience at Aerospace Xelerated?
It’s been very interesting — it’s been a great cohort and it’s been intense. There have been many different perspectives on many levels. Some of them are emotional, for example communicating between technology and sales.
At a deeper level, the technical discussions have been fantastic. Especially, the way we managed to get into deep conversations with technical people at Boeing and GKN Aerospace. I don’t think this would be possible if we were not in the programme. If we tried to reach out to these people in a different way, we would not have that kind of synergy at this pace.
Hear Diana Deca share more about Neurobotx and their next steps at Aerospace Xelerated — 2022 Demo Day
Stay up to date with Neurobotx’s journey on LinkedIn. Deca is also an avid writer and regularly shares updates on her personal LinkedIn.
Sign up to the Aerospace Xelerated mailing list to hear the latest news from our network, follow us on Twitter or connect with us on LinkedIn.
