I met Michal Wroczynski earlier this year at a friends’ party. I knew he was doing something excellent, but had no clue what it was. We met couple more times and I started to discover the amazing challenge he and his team were dealing with. Finally, we met on a Thursday afternoon for a comfortable chat on my sofa (which is famously known among our friends as a place where ideas are born and shaped!). We spoke about the story and magic behind Fido Labs.

SG: Fido Labs. Sounds pretty mysterious. What is Fido Labs?

MW: Fido Labs is an independent lab working on a radically different artificial intelligence (A.I.) that enables computers to finally understand us humans. We focus on language understanding – the ultimate frontier of artificial intelligence.

“We share our feelings, desires and experiences. This is the Internet of Humans, but until now it was invisible for computers. All they can understand is keywords and statistics. That’s why there is Google keywords. Computers are really bad at understanding the logic and the structure of the sentence. “

SG: It understands text on the computers?

MW: It’s very difficult for computers to understand human language. It’s much easier to understand images. There is also this great idea of Internet of Things. It’s about putting sensors almost everywhere, collecting data and interpreting it to understand reality. But at Fido we believe, that the best sensors are ignored. We humans. We miss a lot of information. We are thinking, blogging, reading, tweeting. We share our feelings, desires and experiences. This is the Internet of Humans, but until now it was invisible for computers. All they can understand is keywords and statistics. That’s why there is Google keywords. Computers are really bad at understanding the logic and the structure of the sentence.

Actually, a four year old child finally understands a structure of a sentence and grammar. Its ability to understand and learn grows rapidly. But computers are on a level of a much younger child, most of the artificial intelligence could be called animal intelligence. For example, Google car is in a way a horse’s eye, very advanced one. But language is so complex and difficult for computers.

Actually, a four year old child finally gains the ability to decode the complex structure of the language. From this point its ability to learn and understand grows exponentially. But computers are not even close to that point. Most of the artificial intelligence systems could be called an animal intelligence. For example, Google self-driving car is in a way a horse A.I., very advanced yet still animal. Human language is far more complex and difficult for computers.

It was extremely motivating but also very challenging to go against the status quo. Currently scientists and great engineers try to process text top-down, e.g. “What is the topic of this article?”, “What kind of words are used there?”, “How many of them are positive”, “How many negative?”. We believe that there are small atoms of knowledge in every sentence, in every text. We extract them and create a bigger picture for real understanding. We don’t count how many times you said “fear”. Instead, we extract exactly what you are afraid of or why you are afraid. We have chosen the bottom-up approach. This is the milestone for computers to truly understand us humans.

SG: On what level is it right now? How well does your technology understand the language?

MW: It’s three times more accurate than the state of the art on the whole sentence level. We are pushing technology, so right now you can actually ask any question to social media. Every question you might want to ask has been already answered somewhere over the Internet. No matter if you want to know what you should do to make your app better, what people think about global warming or how you should say “sorry” within an academic community in Oxford. All those answers are out there. The only problem is to get them. We live in a paradigm where search engines, instead of giving you the answer, give you a list of websites where the answer might be. Similarly Siri gives you the answers pre-programmed by Apple engineers.

We are so excited about the Internet of Humans and that we can harness this large untapped pool of knowledge. This is our focus. When we go back to the early 1960s, to the beginnings of the concept of artificial intelligence, there were actually two paths. The first path was to build an artificial human brain. The second path was called collective intelligence. Instead of mimicking humans, the idea was to understand millions of them, free to compete, collaborate and express themselves. The new intelligence would emerge automatically. In the movie Ex Machina there is an artificial intelligence that learned everything and actually became intelligent just by listening to all phone calls and conversations of us humans all over the world. We took the second path, to create collective intelligence by acquiring knowledge from all kind of people, their articles, blogs, forums and so on.

There are so many ways it could be used. We believe that this power should be trusted and people should really benefit from it. A.I. cannot be good or evil, it’s rather what kind of mindsets are working on it and for what purpose. People are afraid of A.I. and I can understand that because it is a psychopath. It has no emotions and can do anything. Imagine that you have a very sick child You need medications but you have no money. You can go to bank and convince people to give you some money. Try to do the same with ATM.

SG: How long have you been working on this technology?

MW: In 2003 we started Fido Interactive and created first dialogue systems for companies. Something like Siri now, but for business. Actually, I’m very proud because we were the first company that passed the “erotic Turing test.” It was an erotic chat, but we used our A.I. to hold talks with consumers who thought that they were talking to real women. This A.I. got two times higher conversion rate than real humans. I still remember a guy, who sent 160 text messages to the system and fell in love believing that a real woman was answering his questions. That was quite evil.

“We developed our own idea of how computers should understand human language and the first step was to teach computers grammar.”

SG: Was that a research or a business?

MW: A business. At the beginning it was fun and we made good money on that, but over time it sounded fake . What we understood from these conversations was that people were not looking for erotic experiences. They actually felt lonely, looking for contact with other people.

Then, in 2007, I started another company with the same mission – enabling computers to understand us humans. We moved to the “light side” and created first anti-pedophile system. It was almost the same system as for the erotic chats. It was able to understand chats with children and spot a predator who is an adult acting secretly and having very bad intentions. I did this project with Polish police and other Polish agencies. It was amazing! And this is how the final idea has arisen – to understand what’s happening on the Internet and in social media since it provides a tremendous value. We didn’t want to follow the path of “let’s help companies make better advertisements.” Instead, we search for other values in the Internet of Humans.It’s not about marketing and sales. It’s about creating a completely new technology. We created a lab, hired the best experts in quantum physics, logic, mathematics. We actually didn’t work with linguists. We developed our own idea of how computers should understand human language and the first step was to teach computers grammar.

SG: When was that?

MW: We started this project in 2009. The mainstream science was figuring out how children acquire language. We focused on another problem – how foreign adults learn English. There are textbooks from which in one way you learn the semantic, the meaning of words, but on the other hand you learn syntax – the grammar or the structure of the sentence. The grammar can be taught to humans, so it should be to computers.  We started to code it – rule by rule – and at the end of the day we have created a solution that enables computers to understand grammar, intent and logic coded in human language. It turned out to be far beyond the state-of-the-art.

SG: So you are focusing on English language right now?

MW: Yes, we are working only on English language. This is a language-dependent solution but we are very excited about building Spanish and Chinese Language Decoder over time. Actually, we started with Polish, but after moving to Silicon Valley we understood that the biggest opportunity is for the English language. Also, we were misunderstood in Poland so we decided to move to SV.

SG: Let’s talk about the Silicon Valley story in a moment. You are a cognitive psychologist, right?

MW: I was a medical doctor who started Artificial Intelligence Society for young doctors. We were so passionate about what we can do with our brains. I wanted to be a psychiatric doctor to understand the most beautiful and complex thing in the universe – the human mind. Unfortunately academia is not a great place for entrepreneurs who want to change something so I decided to leave and start my own lab to work on A.I.

SG: How much does your psychiatric knowledge and expertise help in working on this technology?

MW: I like this question. It has two answers. In one way, cognitive science tries to uncover how we think and act. This is the foundation of what we do at Fido. On the other hand my psychiatric background helps to survive in the startup world. Sometimes I don’t see much difference between working in the psychiatric world and working with young geniuses. It helps me to stay calm and have deeper understanding of people and what’s going on. I think it helps in both ways.

SG: Its not easy to find any information about Fido Labs on the internet. I could not even find your website. Why is that?

MW: We have a beautiful landing page with smiling people from Fido (laugh)! Well, we are in the moment of creating our technology. At first we wanted to patent this so we didn’t show off. In fact, our technology can go in every direction. We can make a new search engine, a better recommendation system, and there are so many other tempting fields. The market is changing so fast that we just decided to focus on our work. And remember, often times artificial intelligence is the most bullshit thing in the world. We see companies that claim they use AI, but in fact there are humans behind it. In the early Fido Labs years, when we did systems for flirting or emailing, we had to hide AI and pretend it’s humans. We don’t want to create that biased communication saying that we can understand language like a human mind, blah blah. I think that PR is what we need a lot, but hype mostly distracts young stage researchers.

SG: How did the story with Silicon Valley start? When was it and how did it happen?

MW: It was around three years ago. There was a government program that invited Polish businesses to the USA. We got approved, but initially were afraid to talk about our technology in SV, because people here are so smart. Now I also see that other great entrepreneurs in Poland are somehow avoiding SV. But there was a 10 000 euro penalty if we didn’t go!

Once we landed here, we completely fell in love with California. Four days after our arrival we opened an office, we immediately decided that we would stop our research in Poland and would focus on English. This language is very easy for humans to learn, but it is very hard for computers. In fact, the Polish and Russian languages are the opposite, they are very difficult to learn for humans but easy for computers. It took us a while to find out how we can decode information from the English language. When we were ready, we came here for fundraising. We had this young, naive, idealistic view of the Silicon Valley that it’s all about the beauty, pure innovation and technology. We came here and started pitching to investors. I remember my first pitch explaining that we convert text to verb related function with syntactic and semantic parameters, which for us was the next big thing, but of course no one understood us. We were lucky enough to be invited by Greg Papadopoulos and Forest Baskett from NEA, and during the second meeting we heard: “Yes, we want to invest.” Since then our life was much easier.

SG: How much did they invest?

MW: The total round was $1,8 million. They invested $500,000 and we got another investment from Scott Raney from Red Point and Ben Narasin from Triple Point.

SG: Did you go straight for the round A?

MW: No, we first raised a VC seed round, before we bootstrapped for 5 years and had first global clients — this is how we financed the technology at the beginning.

SG: How did you get these clients? How did you get any income?

MW: We were creating some commercial solutions for example for insurance companies or banking. First in Poland and later in the US, where MasterCard and Procter&Gamble were among our clients. It was an atypical hybrid of a lab and a company, where the company is making money for the lab to make research. And I think it’s very cool that the research is not about PhDs or getting a medal from a President.

SG: In sum, you started with the government program that lasted two crazy weeks, then came back to Poland to grow and came again to the US to raise money.

MW: Yes, in the meantime we were flying back and forth to learn, to understand the market approach, to network with great people like the founders of Siri and researchers from Stanford and other universities. Also, we came back to learn how to convert a lab into a startup.

“One of the investors told me: ‘The most important thing in Silicon Valley is to understand who is your friend, and who is not’. — This is what we learned the hard way.”

SG: It must have been a pretty intense course?

MW: Yes, it was a crash course and we made all the possible mistakes. We had a lot of material to learn.

SG: What is the most distinctive mistake that you made?

MW: We came from a culture, where people are more skeptic towards innovations. Suddenly, everyone sounded so excited about what we do. At the beginning, we thought that everyone who says they are our friend is a real friend. But we crossed paths with some shady people who want to take advantage of you being naive and new. I think this is something people who come from Europe are not used to. This is actually a great lesson that can be learned in startup accelerators. One of the investors told me: “Michael, the most important thing in SV is to understand who is your friend, and who is not”. This is what we learned the hard way. I would really encourage everyone to really check who they are talking to, ask for references, talk with other startups, etc.

SG: Since you did not have any network here at first, how did you find first ways to arrange meetings with investors?

MW: We managed to meet with angel investors and this is pretty simple, but it’s very hard to get to VCs. Our great IP lawyer Barbara Courtney, who was working with us on a patent, was impressed with what we were doing. She believed that this is something completely different and new. She introduced us to her friend, Forest from NEA.

SG: It’s pretty common that lawyers help here in Silicon Valley and that’s why it is very important to make sure who we are hiring. Did you raise money pretty quickly?

MW: Not at the beginning. Here, everyone is used to investing in startups that take existing technology and create a product. No one understood us and when we said that we are better than Stanford University, no one took us seriously. It takes time to gain credibility.

Guy Kawasaki said that when an entrepreneur opens his mind, he starts lying. I can truly understand investors who deal with all this bullshit and they hear: “We have this new AI, which is better than others”. It took us a while to learn fundraising. We approached angels, who are usually less likely to take such a long term risk, but then we understood how to pitch, how to communicate and we were lucky enough to find investors who understood us.

SG: Where do you spend more time, SV or Poland?

MW: This year I was mostly here in California. I was also able to skip three winters in a row which is amazing!

SG: Where is the main headquarters of your company?

MW: The main HQ is here and we are a Delaware corporation that owns a Polish subsidiary.

SG: How about visas? This is a challenge for every foreigner. How do you deal with that?

MW: Usually startups are applying for investment visas, where founders are treated as investors in their own startup. We made this funny mistake that we raised too much money, because the vast majority of investment was from the US. Right now we are on E1 visas that is called a treaty trader visa. I, as a CEO, can be a CEO of the mother company in the US, because there is special treatment for a treaty country like Poland. But visas are a huge issue for startups and I have friends that lost their startups over this issue even though they were great developers. That’s why I strongly encourage everyone to hire a good immigration lawyer and do it as fast as you can.

SG: How soon did you start preparing the visa stuff?

MW: It was much too late. We already did one round of financing. I think that after you are seed founded and are incorporated here, this is the next step. It’s important to secure your visa early. It can take a lot of time. I came to Poland to grab my visa within two weeks and it turned out they put me on hold since they wanted to do a background check on me. I ended up staying in Poland for 4 months. That can be deadly for the business.

SG: What was the biggest struggle in your startup?

MW: Currently we are in the middle of fundraising and for me, as a CEO, this is special, beautiful time.

SG: Are you raising a new round now?

MW: Yes, I am leaving and breathing fundraising right now, this is all I do at the moment. It’s everything I am focused on. It’s also very exciting.

SG: You said that this year you’ve been mostly in Silicon Valley. When you come here, where do you stay? Where do you live?

MW: We always rent a house. Right now we live in Foster City. People from Fido and lots of friends often stay with us. Some people call it a Fido embassy. This is a great option for now, to live together, work together and have some fun together as well.

SG: How many co-founders do you have?

MW: I have two co-founders and three late co-founders who do magic with us.

SG: Do they also travel with you?

MW: We are so lucky the whole team visited us here in the US. I wanted everyone to experience the spirit of SV. But we also have employees here. It is extremely amazing when you pitch a product to someone and they leave their corporate world to work with you on a new, disruptive idea.

SG: What is the next step that you are working on in Fido Labs aside from the investment?

MW: We are working on a platform for product managers that will read and understand the Internet of Humans, that will read all kind of feedback about their products from reviews, blogs, and let them ask questions. This is a way to get revenue.

“Structured data, which everyone is dealing with right now, can tell us what is happening, for example how our conversion works or how many people bought our product. Unstructured data tells us why it happens. And without this knowledge we are all half-blind.”

SG: With your product, they will be able to do research with all the data that is in the internet pretty easily then?

MW: Yes. I like what Marc Benioff, the CEO of Sales Force, said about structured and unstructured data. Structured data, which everyone is dealing with right now, can tell us what is happening, for example how our conversion works or how many people bought our product. Unstructured data tells us why it happens. And without this knowledge we are all half-blind. When you ask a question you don’t want to get more data out of data. You want to get precise and relevant answers.

SG: In how many years will it be a very common thing in our world?

MW: I think it’s a great marathon. It takes a lot of time and patience. But already now app developers can benefit from the system, for example they can analyze app reviews. This is just a milestone, I think in a short time, couple of years, anyone will be able to ask any type of question and get their own, personalized answers.

SG: Thank you so much!


Enjoyed the interview? Sign up to the newsletter and be notified about new inspiring startup stories and articles. You will get only the best, curated content.

Subscribe