Saving Wildlife With Epigram AI
Can We Use AI to Help Monitor Wildlife?
Norwegian AI company Epigram was acquired in December 2018. The acquisition was made for $1.5M (NOK 13M) and resulted in tools, IP, and all AI engineers moving to work on Huddly’s intelligent cameras (delivering to Google). However the company did not stop there. I will give you some background information about Epigram and proceed to my interview with Olav Djupvik the current CEO. 1,000,000 million species are threatened by extinction and we question how the field of AI in collaboration with wildlife research can contribute to preventing any further loss of wildlife.
$1,5M is a decent acquisition after one year. Epigram was founded by Bendik Kvamstad and Olav Djupvik. Bendik decided to join Huddly together with seven other highly trained data scientists in the team specialising on computer vision. According to an article by the Norwegian industry magazine Shifter the two founders received four million NOK ($489,428) each while the five other employees shared five million NOK ($611,785) amongst themselves.
An epigram is a brief, interesting, memorable, and sometimes surprising or satirical statement. The word is derived from the Greek: ἐπίγραμμα epigramma “inscription” from ἐπιγράφειν epigraphein “to write on, to inscribe”, and the literary device has been employed for over two millennia.
Having met the people from Epigram I can definitely see the satirical edge and attitude they have about the way they go about their work. What I got when I entered their office was feelgood. Situated in Oslo Science Park they are working hard to make a difference.
As mentioned the acquisition was not the end of Epigram. Olav decided to continue onwards, and it is with him whom I had the pleasure of having an interview. Let us however prior to this backtrack a few years.
The Backstory of Epigram
When I am talking about Epigram it is not the South-Korean lifestyle brand nor the metal band. It is Epigram AI with the platform that provides fast, flexible and precise image analysis to scientists and domain experts.
The Small Machine Learning Agency Back in 2017
Before Epigram was founded they had been a high profile startup and investment by a large media conglomerate in Norway called Snapsale. This was an unfortunate event for the Norwegian startup ‘ecosystem’ due to the failure of collaboration between a small startup and a large company (‘corporate’) with its resulting proliferation in the news (it has its own topic tag in the largest business newspaper in Norway) and several entrepreneurs that I talked to following this cited a growing lack of trust to large businesses. More about this in the interview further down.
Epigram was founded in 2017 as a machine learning agency and had worked with multiple clients. This included Huddly (which acquired them), Gjensidige (asset manager), Tradono, and No Isolation. They had in this time been helping to develop bespoke solutions within the field of artificial intelligence, specifically computer vision as well as visual classification utilising machine learning techniques. Epigram was founded by previous and current students at the Department of Informatics at the University of Oslo (according to an outdated post on The Hub).
Jumping back to an old job advertisement from navet I could manage to spot their first logo. Back in 2017 it looked like this:
Their short pitch at that time sounded like this:
Epigram is a machine learning company based in Oslo, Norway. We’re currently solving machine learning problems for companies in Norway, but aiming to create our own product in the not so distant future! If programming + statistics + math + creativity = YOU, then do not hesitate to contact us — we’re always looking for talent!
They were looking for someone to be writing code in Python, using libraries like, tensorflow, numpy, opencv etc. By using Web Archive I can also spot that they wrote on their website the 6th of November 2017:
We are passionate about what we do and in addition to building our company we aspire to build and nurture the machine learning community in Oslo.
They seem to have been working in Majorstuveien and Gaustadalléen, however they later moved to Schweigaardsgate 34C also known as Greenhouse Oslo. The first time I met Epigram was when they had just moved to Greenhouse Oslo, the same coworking space I had been working out from. The focus of the coworking space was on sustainability and particularly greentech and agritech. When I talked to them the first time they had taken the classical early-stage approach (in my experience) of mixed consulting and product building.
They had been working to make solutions to detect diseases amongst fish and with an insurance company to attempt understanding the possibility of making it easier to detect damages on your cars and make the insurance process easier particularly for those renting out vehicles, yet also possible private customers. It is funny, because I have recently seen some companies launching this product elsewhere, so they possibly too early into this area.
Their business produced a broad range of cutting-edge software and tools designed to collect and analyse big data, with a greater focus on the application of intelligence in understanding context and content. I could remember that they were one of the hardest working companies in the house, because there were almost always someone there no matter what weekday or time I came walking by (they took breaks, but I was always impressed).
Rolling into 2018 and the acquisition by Huddly
Luckily for me writing this article I came to learn (through Googling) that Epigram had started their own Medium account in 2018. Stian Lind Petlund and Therese Byhring wrote a few articles that I am so lucky to run through to attempt explaining part of their year.
Stian writes the 6th of March 2018:
When we started Epigram we had agreed to start a company with an inclusive environment for students and other that wished to understand machine learning. We also agreed that Epigram was to be a counter to all the hype surrounding machine learning and artificial intelligence, and communicate that behind all the seeming magic there is a steady and understandable academic field. In addition to being inclusive and relevant, it was also important for us to contribute to society […] Machine learning is engineering in the same manner as builing cars, houses and apps. In the same way that there is no app for everything, there are no general artificial intelligence for everything. At least not yet. [I have loosely translated this from Norwegian]
Epigram AI – hvem, hva of hvorfor?
At this point Epigram had started to adopt a weird yet interesting retro-futuristic aesthetic:
The company had around this time additionally made an update to their visual profile, with their logo accompanied by a lot of purple:
As described introductory Epigram had grown out of Snapsale which was overhyped (according to themselves) and heavily bought out by a large media company. It therefore makes sense to see the opposite approach this time around, wanting as little exposure as possible while insisting not to ‘overhype’ the technological implementation. Epigram AI had started to attempt their move into health. Esten Høyland Leonardsen held a presentation at Deloitte on the 10th of January 2018.
The benefit here for business purposes should be relatively clear. If you can track movements and predict certain behaviours of animals (such as a bear) perhaps the same could be said for humans?
Before the acquisition Epigram had been working on Huddly IQ — a product offering features such as automatic framing and advanced space analytics. They had helped Huddly use this offering to detect and count people.
Beyond counting people we could wonder if Huddly that specialise in meeting rooms could have an interest in quantifying and understanding human behaviour in meeting rooms perhaps with analytics. This is of course purely speculative statement, however we could wonder if this would be advantageous in large or small business deals or purely for collaboration.
Epigram was acquired by Huddly on the 4th of December 2018 for $1.5M (NOK 13M). However the company continued while most of its assets and employees had been transferred to Huddly.
Today Epigram’s website looks more like this focused on their collaboration with Norwegian Institute for Nature Research.
I managed to catch a picture with the current Epigram team on the 27th of June the day when I interviewed Olav.
Artificial Intelligence to Save Wildlife
“Never has it been more important to understand how the natural world works and how to help it,”
-David Attenborough, Our Planet
It is clear that what Epigram AI offers, and the company has been working with the Norwegian Institute for Nature Research already. Their offering could be an interesting tool to help monitor and possibly protect wildlife.
You have likely seen this picture from the IPBES report before if not it might stun you and make you think for a second or the rest of the year:
The IPBES Global Assessment Report on Biodiversity and Ecosystem Services is the most comprehensive ever completed. It is the first intergovernmental Report of its kind and builds on the landmark Millennium Ecosystem Assessment of 2005, introducing innovative ways of evaluating evidence. Compiled by 145 expert authors from 50 countries over the past three years, with inputs from another 310 contributing authors, the Report assesses changes over the past five decades, providing a comprehensive picture of the relationship between economic development pathways and their impacts on nature. It also offers a range of possible scenarios for the coming decades […] The Report finds that around 1 million animal and plant species are now threatened with extinction, many within decades, more than ever before in human history.
-Media Release: Nature’s Dangerous Decline ‘Unprecedented’; Species Extinction Rates ‘Accelerating’ retrieved the 15th of July 2019
I discussed with the current CEO as an example wether these application could be used prior to implementation of large scale energy wind projects or to understand the environmental impact before and after certain other large investments.
I will do my best to not make this piece an empowertisement. What Epigram offers is of course part of a possible solution not THE solution. In this same breath is easy to dive into solutionism.
Empowertisement: is a word coined by Andi Zeisler to refer to the coopting of political and social movements by corporations to sell products.
Solutionism (uncountable): the belief that all difficulties have benign solutions, often of a technocratic nature.
Of course we cannot only track wildlife population — it has to be protected. Naturally we have to think about energy usage for the AI implemented. It is not guaranteed to be sustainable if we implement it widely. However it can be preferred in terms of usage compared to using ML to monitor your chatting patterns to more effectively sell you things. Is our ecosystem more important than your emojis? There is no absolute, rather an open question.
The ‘understand the bear’ exercise is equally amusing and fascinating. It is certainly interesting to see how AI can be used differently for or with nature. As such I was interested to understand Olav Djupvik and why he had spent so many years and so much time on Epigram as well as other startups. I will start to write out the interview here, and if you are interested you can check out the rest of the interview on my podcast at AI Social Research.
Interview with CEO of Epigram — Olav Djupvik
From the dotcom boom; to protection gear for gas and oil; towards using machine learning to track wildlife. The bumpy round of Olav-Andreas Djupvik for ease of use shortened to Olav. My name is Alex Moltzau shortened to simply Alex. The interview can be listened to through Soundcloud or read here in this article:
Olav: my name is Olav-Andreas with a hyphon or dash, it is very west coast in Norway where we tend to have double names. I come from a family of founders and entrepreneurs my grandfather was an entrepreneur, he worked in early telecom before the 2nd world war and building up transportation, buses and stuff. My father took over his business and ran it, he had to stop when he was 75 because then he was not allowed to drive a bus anymore at least for pay. So my whole life in my youth I could sort of see how they built their company and dealing with all kinds of challenges both to personnel, infrastructure and everything with running a business. I have learnt that it is a struggle and accepted that it is a struggle that you have to fight for staying ahead both when it comes to technology personnel, property, where you’re staying, financing, customers everything. It’s a very hard fight, sometimes you succeed and sometimes you don’t.
My first startup was in 1996, then I was… How old was I? Around 27 I think and it was a huge success. We listed on the Oslo Stock Exchange and the company was bought by a French hosting company during the dotcom boom. So my first startup was sold and I quite immediately started looking for the next venture and in that period our company was webhosting company. We also developed software for financial services. Right after the exit B2B marketplace (business-to-business) was very hot so my second venture was a B2B marketplace in personal protection equipment in the oil and gas sector. The second startup was in London together with…
I always had co-founders. In my first startups we had three. We quickly became more in the founding, we became several founders and merged the companies. If you asked publicly who are the founders of Infostream a lot of people will raise their hands and say I am one of the founders. After this startup I thought: “this is easy” so I failed miserably with the second one. I think the company still exists, but it was taken over by another company and yeah so…
Alex: What was the name of the second company?
Olav: Weartical. [co-founded with Olav Inge Røyset]
Alex: With a V?
Olav: It was not a very good name. I think the idea that the oil and gas was a vertical. We were focusing on personal protective equipment for the oil and gas industry and it has to do with wearing.
Alex: I am not sure all listeners understand what a vertical is. It is a very startupy term.
Olav: Think of it as an industry, some companies are in the early-phase, then you have different companies in the value chain all the way to the gas station where you can buy petrol, it is a vertical downwards from the gas station to the oil field. If you are thinking horizontally the other way it is all the gas stations, different gas stations.
Alex: So it is related to the value chain.
Olav: Yeah it relates to the value chain.
Alex: so they are like: “I’m going to catpure ALL parts of the value chain.”
Olav: yeah that’s great, then you’re an early-phase startup!
Alex: yeah, ’cause you’re super small and you’re going to do everything, but where did it go after that?
Olav: we ran the company for two-three years and we even bought some bricks and mortar companies. Some companies serving the personal protective equipment market with warehouse, staff and everything. I think we bought at least two companies and we merged those companies and we built the storefront a B2B storefront and we got customers to use our storefront to order personal equipment for the employees. If they were going to send 100 people to an oil rig they could just order individual packages for employees. You had 54 helmets, exactly the same size garments with the name on it. Usually they send everyone to a warehouse and it is a complicated process. With our storefront they could just order in a simple online solution.
Alex: so that time in London ended with you selling out from the company or the company being bought?
Olav: we sold the company to our largest supplier.
Alex: makes sense.
Olav: it was not a financial success, but we got some of our money back. Then during the years I’ve had a few detours into normal worklife. I’ve been working around a year for Telenor. I have been working a few years in finance because I’m an economist. I have worked in finance for a few years, but I was more or less doing corporate work – corporate finance getting investors into companies that were going to be listed. That was typically a job I did. Then you are sort of on the financing side of entrepreneurship.
Alex: it is interesting because there are all these different stages. People segment it into pre-seed, seed, early, venture capital a lot of different areas of series A, series B. You are sort of towards the…
Olav: listing.
Alex: yeah listing side of it which is late early-stage — we are becoming a company.
Olav: all the companies we were raising money for they were quite mature, established companies.
Alex: as I would imagine when you go to the stock exchange.
Olav: but still risky.
Alex: there is always a degree of risk right? It’s what you get on every investment paper.
Olav: it is funny to read the disclosures in the prospectus on listing companies. It is sort of: “Everything can go wrong.” 40 pages of risk factors.
Alex: of course, they could sum it up: “everything can go wrong,” then hand you the papers. So when did you move back into the ‘starting-up-company-mode’?
Olav: 2011 I moved back into the Oslo Science Park. I was matched up with a founder of another company and we co-founded the next startup. So that was first a company called Skylib and it was sort of sharing economy company where people were supposed to put in stuff they wanted to share with others like a lawnmower or a screwdriver. People were going to list stuff to neighbours.
Alex: so OBOS took the idea?
Olav: yeah I was going to call OBOS to tell them that it won’t work. People will list stuff, but not borrow it. Maybe you are afraid of ruining the lawnmower.
Alex: it is like the sharing economy recent ventures close ventures. If anything goes wrong it’s going to be fine. With insurances etc. You want people to share things it is environmentally friendly not buying 10 or 20 lawnmowers and everyone does it once in a while. It is important for society being able to share things, but the practical side is not a straightforward manner.
Olav: where I live I have no problem asking for help, I was fixing a pipe the last week and I asked for a spare tube for a bike and he had it — and handed it over. You don’t need technology you can just shout.
Alex: how did it go and what other company did you transition to.
Olav: we found out it didn’t gain enough traction, so we decided to start selling stuff. It was a restart of Skylib, and we renamed it Snapsale and started to use machine learning. We used image recognition for recognising what people were selling. If you were selling a bridal gown the image recognition could recognise.
Alex: so it was categorical? That was your first venture into machine learning.
Olav: we used it to sort everything into the right category. If you took a picture of something it was categorised. The user did not need a lot of time to mark stuff they could just say what price. We could list a product in 15 seconds and all the information is there. I think there are other companies that have adopted the procedure and this kind of use of machine learning. Then a media company called Schibsted invested, but they didn’t want to invest further. They didn’t want to do the market launch, I think because it was in strong competition with other investments.
Alex: should we help you or kill it and steal it?
Olav: So they killed it and I kind of understand them.
Alex: from a market perspective or commercial sense? It is interesting to see these stories like with mCash and DnB, but tell me about the process at least a little bit at least. Because I think a few Norwegians know the story of Snapsale, but others might not be so familiar.
Olav: We had an app and a very good team. We were nine programmers and me. So I was the only one with business experience. When we took the investments I knew there was a risk that they could destroy the company.
Alex: with your experience I would imagine.
Olav: the thing I really regret from the process was that all the time I kept saying to my co-founder that: “Schibsted is a huge company they can’t do stuff like this, they can’t screw you.” I said that all the way up till the end, I said the same to Geir [co-founder], but they could. I don’t know what I was thinking, but I actually believed they wouldn’t screw us. I actually believed they would see the benefit from the service, and they would sit down and negotiate a deal favourable to both sides, but instead they chose the destructive ending.
That made my co-founder mad and when the story ended up in the newspapers in Norwya it was kind of a blow to Schibsted and raised a huge debate around startups and big companies in Norway.
Alex: yeah it did, it had huge repercussions for the whole ecosystem.
Olav: and I think that was my co-founders intention by doing it. Schibsted tried to bribe us to not tell the story. They tried to get us to accept some money and sign a document where we shouldn’t say anything about it, but Geir didn’t want to do it at all, and I was negative to it too: to sell out and shut up.
Alex: what sort of stake did they have at the time?
Olav: they had enough to block… Just across 34% so they had negative control. So they were in a position to stop the company if they wanted. We tried to have other investors, but they didn’t want to sell. So it was very…
Alex: stuck in limbo for the staff and for you… Moving on some of you started up a different venture right?
Olav: the team was split and we were all offered to go to Cognite where my co-founder went, but we also had an opportunity to start a new company and a customer that wanted to explore AI further, so about half of us decided to go for a new startup and my co-founder decided to go for the more, still risky, but well funded Cognite.
Alex: Cognite seems nice.
Olav: yeah they are doing well. They are well-financed and growing rapidly within the oil and gas sector, but I think the guys that joined me liked deciding for themselves what do and what to focus on.
Alex: moving away from oil and gas towards different areas.
Olav: our first client was within insurance, I guess the largest insurance company in Norway. They wanted to explore understanding the severity of a damaged car, or a damage on a car.
Alex: I remembered that you were moving into a digital and physical realm, right, that was interesting.
Olav: it had to do with the damages or the claims of the insurance company as a cumbersome process, it is easier to just collect the premiums.
Alex: I remember the prototype, and I was impressed.
Olav: the idea was to speed up their claims process, so instead of having the customer go back and forth between the garages and appreciate the damage as frustrating and want it fixed, but at the same time the insurance company want to control the cost, so they don’t want to dish out for damages that are not actually there. They want to understand the severity and control the costs. It was an interesting project and we are still close to the insurance sector. We are looking at property, farms – where the insurance premiums are high and were you want to control cost and inspire the customers to be careful and responsible.
Alex: I think the insurance companies have a really important role especially now, because I heard from an Australian environmentalist that a lot of insurance companies were questionable about underwriting large projects because of climate change, but when you don’t have an insurance it can be challenging to make projects happening.
Olav: it can be the same thing in Norway, you have more landslides and flooding. Last year we had a drought and this year we had a huge scandal with pig-farming industry. The public is asking the meat industry to provide real-time video from the slaughter house and pig-farming, so I think the… The suppliers, insurance is important because they are in several areas where there will be effects of climate change on the business and I have the impression that the insurance industry are eager to support the industries that they are in.
Alex: Yeah I think sustainability and climate change — it’s a huge risk if an insurer insures a company and they are not really compliant, because it is very damaging to the planet, the reputation and financially. Those things are very connected now.
Olav: it could easily have an impact of the pricing of risk. It is a very hard estimate the risk, so if you don’t know how likely is it to be a landslide, drought or all the fish in the fish farm is dead in one month, how can you insure it?
Alex: it is more unpredictable.
Olav: you can end up pricing the risk very wrong.
Alex: Epigram from start to now can you tell me about that journey?
Olav: We started out with insurance company as customer and we were five founders. Should we do more consulting within AI or should we do a product? After a year we decided to split up, so two of us bought out the other three-four and the other guys continued doing consultancy and some went into PhD programs and stuff. Me and my new co-founder decided to go for a more product direction, but just six months into that we got an offer from Huddly, another Norwegian camera producer.
They wanted to speed up their AI development, so they wanted to buy the company. I didn’t want to sell, so we decided noch ein mal (yet again) to split. So we sold of 90% of the tech team for cash and shares in Huddly and I decided to find a new co-founder and continue developing the image recognition platform and follow up the customers that we had. We have a very interesting customer in the Norwegian Nature Research Institute, so we are doing image recognition on all the wildlife cameras that they have placed around in Norway.
Alex: I see.
Olav: So all the images are going into a platform that are sorting removing images containing humans, because they are sensor operated. If you walk into a place where there is a wildlife camera you get taken picture of. For privacy protection they remove all the human related images. It could be tractors, people on a horse, walking the dog…
Alex: monitoring, but needing to be ethically coherent.
Olav: that’s how it started, because they, datatilsynet, data protection agency in Norway, they wanted the research institute to improve the privacy.
Alex: yes I think what you are working on now is so important, fascinating. With recent report of the drastic reduction in species and the risk posed to the world as part of that. We need to be able to figure out how we can keep our wildlife vibrant and that’s an interesting challenge.
Olav: I think the main priorities in the project are the wolf, lynx and arctic fox where we are trained on 100'000s of images to get as good a model as possible, and it’s now sorting all the images very effectively. If you have a camera in the mountains with snow, rain and different… The images can be from disastrous to fantastic, we use AI to sort through to take away the blur, not usable images and I think the nature research institutions can use the platform for many different projects.
Alex: in essence it keeps track of the wildlife population. What is the purpose of this collaboration and what are you trying to achieve?
Olav: what we are doing in collaboration with nature researchers is to observe much more than we were able to do before. Earlier I get the impression that they used less image and video, they used more fieldtrips and studying different areas of nature. Now they can generate much more statistics by means of using video and images so it is easier to document and you can use the… you can have weeks and months of data and video that you can sort of use a machine to handle and then get a quite good impression of what’s happening. For instance we mounted a video camera to the bears and they walked around in the summer, and we used machine learning to guess what the bear was doing.
Alex: wow.
Olav: in one section the AI said now it’s hunting, eating berries, now it’s having sex.
Alex: bear tracker!
Olav: we could put it in statistics and show it on a graph. I was impressed by how much the bear likes to swim.
Alex: so you get a closer understanding of a bear, classified?
Olav: yeah, when you use AI to distinguish between hunting and eating berries, but then we saw it after a while that when it was hunting it was charging forward towards the deer, but when the bear was eating blueberries it made this swinging head movement. It was eating like a one meter band through the nature, it made the swinging by the head recognized by AI. So that project is interesting and the arctic fox program.
Alex: suddenly there has to be possibilities internationally to understanding our wildlife population around the planet. We get this report and we know the wildlife is in decline in terms of the fish and a lot of other… The amount of species, but also the general wildlife population, I cannot remember the number, but most mammals being human owned and bred, there is not much left of the ‘wild’.
Olav: that has been my idea the whole time working with nature researchers in Norway we need to work with the scientists and what we can offer is our skills in computer science, they don’t have the time to acquire the programming skills, because it is too much work. We need to put our knowledge into system they can actually use. To take it internationally we need to work with data scientists abroad, and it could be… I was suggesting to the researchers that we should study the videos that were taken when they were planning before putting up the windmills looking at birds, and it would be totally possible now to use our tools to study the old data.
Alex: well historically, but as well it would be possible to create regulations that made sure you had an impact study done with cognitive machine learning of tracking and understand the wildlife in an area before you go in with a $200 million dollar project. It would be a good investment to understand the implications of such a large scale project.
Olav: we are doing an aquaculture project with a camera company in Trondheim, you shoot video below the surface in pens where the salmons are bred. We observe illnesses and problems with the salmon.
Alex: do you do any collaboration with IBM on this?
Olav: we work with a camera producer and work with them to develop algorithms that can understand the welfare of the salmon.
Alex: software company now and you are naturally working with hardware companies.
Olav: it’s a good collaboration. Researchers have cameras, Huddly have cameras, sea farmers have cameras. It is very… You need to work with people that collect a lot of data.
Alex: on that note, chips and development of new hardware that is custom made or better adapted to machine learning. Any thoughts on that?
Olav: I think the Huddly camera is a good example. There is Movidius chip, so we develop models that actually runs on the camera, and I think that will be interesting for the wildlife camera, the pens in the aquaculture, you could end up with quite a bit of processing power inside the camera house, and there is a lot of advantages by doing that, because you don’t have to move it to the cloud to process you could process at the edge, and I strongly believe that would be more widely used. I think so.
Alex: so that brings us to now in a way in your current situation here at Forskningsparken, Oslo Science Park.
Olav: There is a vibrant community here of researchers and startups. There are both different institutes for water research, transportation research and there are very small companies to the more mature. I like to count us towards the mature since we are 10–12, yet is good to be close to the Institute of Informatics and other startup companies that we share knowledge with.
Alex: good co-location, do you do any collaboration?
Olav: we are part of projects where we cooperate with research institutions especially in CT-images for lungs and SINTEF. So we are within walking distance of our partners. I think even in… There are companies and institutes in this building that we have not talked to yet.
Alex: it is so challenging to find people in the community and take the time to meet up, or even know what people are doing. I guess that it is partly why I am doing this. I had no idea you were tracking bears. Bear tracker! It’s like understanding the bear, being one with the bear.
Olav: I remember when I first looked at the bear videos I was so scared because I had the sound on. It recorded the breathing of the bear when it was hunting, it was like a scary movie!
Alex: this is like B2C you know you could just put on a VR headcam and get the bear experience. People doing meditation anything like that: “don’t, just forget yourself, be the bear.” Great business opportunities here, untapped market. Nature lovers gone wild, or digital… So any questions that I haven’t asked you that you would have liked me to ask you. It’s a typical question towards the end of the interview.
Olav: I think I would have asked what do you think about the hype of AI? Do you think it is overhyped? Will we sort of… What sort of benefit will we have from AI? I think the worst example of AI that I have experienced is this robot from Dubai.
Alex: Sophia. Everyone talks about that. I haven’t had many interviews, but it comes up every time.
Olav: I think it is so annoying and destructive the way it draws attention to AI because it is simply a scam. At the same time it is sort of… People don’t know how difficult it is the image recognition, oh my face is recognised, but if you are thinking about predicting if an image is blurred or not can this be used or not. Is it a CT image, a dangerous illness, it is normal? It is such a challenging task with serious researchers, so I think this Sophia robot is destroying the seriousness of the challenges we are facing.
They believe this or that is solved, but we know there is so much work to be done to have proper AI and actually reaping the benefits of your efforts.
Alex: this is the dream of sort of AGI, artificial general intelligence gone wrong in a way. Because everybody knows we are not there, who have dipped their feet into the field of AI. With Sophie it is this anthropomorphism, wanting the technology to be human. Wanting that reflection of ourselves in the technology that we build when it is different. I think there is so much important work that is being done in the field of AI although there are plusses and minuses.
Olav: if you take the radiology, the lung project we have now. A lot of lung diseases go undiagnosed, it is chronic and very dangerous, so it is so important putting your resources into discovering the illness early. When it becomes late stage it is only a transplantation that can fix it. If the solution is that you don’t have the data set. You can’t have the… Some of the diseases are very seldom so there is not many radiologists that remember all of the diseases. It is because they haven’t seen all the different types. If you lump all the cases and experienced radiologist to annotate the data then it is a huge leap forward for treating lung disease.
Alex: yeah and there is a lot of people that smoke and they get lung cancer, so it is a common size of disease.
Olav: I think the respiratory diseases is the third most common disease.
Alex: there is more pollution, it is not a trend on decline, sadly.
Olav: you know then they transplant the lungs they don’t want the city lungs, they want the rural lungs are the best.
Alex: that is fascinating.
Olav: it is scary for us that are urban.
Alex: yeah I mean we live on average quite a few years less, and I think that has been measured in Netherlands, but probably done a lot of places. Most people do live in a cities, so the average health will go down, so we need to detect these things… But to finish off, just to say it is fascinating what you do and some amazing applications of the potential of narrow AI and what it can be purposed to do in terms of nature and health. Trying to help to manage the resources that we have.
Olav: I think it has to do with what we can do with AI combined with expertise is to handle and analyse more data to guide the scientists towards serious problems. In the lungs to scan all of the radiology and images put through the model. If there is something scary it highlights it and gets the notice from the radiologist. Here is something early stage, so the radiologist can check if it is actually true or not.
Alex: saving lives, it is not the worst thing you can do with your day, and trying to save parts of nature at least through the help of narrow AI.
Olav: you can see it with climate change that you can see things that surprises scientists. Observing something that is peculiar or strange.
Alex: it has been symptomatic that quite a lot of large companies have been self-policing their environmental responsibility, and it is a trend that needs to be counteracted. There is a possibility with cognitive machine learning to help in that regards. Great!
Olav: Great, thank you.
Thank you for staying with me or simply visiting my profile and this article. As always your thoughts are always appreciated and if you want to stick with me for this journey I would be very happy.
This is day 43 of #500daysofAI.
What is #500daysofAI?
I am challenging myself to write and think about the topic of artificial intelligence for the next 500 days with the #500daysofAI. Learning together is the greatest joy so please give me feedback if you feel an article resonates.
