Accelerating Human Knowledge with AI

Jan 07, 2020

Accelerating Human Knowledge with AI

CERN Alumnus - Mait Müntel
At CERN – 2004 – 2012 Summer Student and User with CMS
Now – Founder and CEO of Lingvist

Transforming a hobby into a pioneering start-up

Imagine the scene: after an enjoyable day of hiking in the French Alps, you are homeward bound, waiting in a damp and cold railway station. You gesticulate to the stationmaster, using all your limbs to convey that you are looking for the Geneva train. As the last train pulls out of the station, you discover, to your horror that was your train and that you have missed it because you do not speak French. Some years ago, in a Haute Savoie station, it was at this very low point that Mait decided he needed to learn French, and more importantly, it sowed the seeds for an idea that is now accelerating human learning. 

Politics changes lives!

Text Box:  Mait close to Kamchatka 2005
 Kamchatka 2005, prior to Mait's arrival at CERN

Mait Müntel, native of Estonia and avid hiker, first came to CERN as a summer student in 2004. His coming to CERN was somewhat unexpected, but serendipitous nonetheless as it paved the way to developing his own company.
“I was doing my masters studies in high energy astronomy, happily calculating and modelling nuclear explosions of neutron stars, but then my supervisor became a politician and Vice-President of Estonia, so I had to change direction. It certainly was an upheaval, but it brought me to CERN!”  As in many cases, Mait’s experience of amazing lectures and extracurricular activities whilst a summer student put him firmly on track to pursue his studies in particle physics at CERN and more precisely within the CMS collaboration. In 2008, his obtained his PhD, completing his thesis on the double charged Higgs Boson, “I think it was the last thesis to be written using simulated data!”

Following on from his PhD, whilst affiliated with the National Institute of Chemical Physics and Biophysics (Estonia), Mait continued working at CMS. He recalls with strong emotion the atmosphere at CERN at the time,

“The environment was incredible; there were smart people everywhere! In addition, the machine had recently been switched on, I was one of those who did the first shifts when the machine was ramped up, and of course it was the most amazing time in 2012, with the discovery of the Higgs; such an emotional moment. The auditorium was so packed I had to participate via a conference room! “

Real data had taken over from the pre-LHC simulations Mait had worked and he realised very quickly that the real data looked nothing like the simulations; so many things had to be rectified. He admitted that he struggled with coding;

“I didn’t have any formal training, I thought I would easily ‘learn by doing’ however, with hindsight, I should probably have been more systematic in my approach to learning”

Did this experience subconsciously plant a seed in Mait’s brain, pushing him to reflect further, probe, and develop enterprising methods to improve learning?

Not Lost in Translation

Despite spending notable amounts of time in the Geneva region, Mait had held back from learning French. However, he became increasingly convinced that with the help of computers, theoretically, one could learn languages at an accelerated speed.

“I would describe what I was doing as a hobby. Over the course of one weekend, I wrote a crawler, which enabled me to download a collection of French subtitles from a film database. I wanted to analyse from a statistical point of view the language people were actually speaking, as after having spent several years learning both Russian and English, I was convinced that what is found in academic books and courses is not always so relevant.”

Mait admitted that in the subtitles, he discovered a bias towards crime and military vocabulary, peppered, of course, with a certain amount of swearing. However, scientists have confirmed that the direct speech Mait was analysing is the closest source to natural language. The next step was to study memory models to understand how one acquires new knowledge. Mait calculated that if a computer programme could intelligently make the decision of what would be optimal to learn in the next moment, it would be possible to learn a language in only 200 hours!

“This thought inspired me, so I started building some software using ROOT, the object-oriented program and library developed by CERN for particle physics data analysis. After only two weeks, I was able to read a proper book in French! I had included a huge book library in the software and as the computer knew my level of vocabulary, it could recommend books for me. This was immensely gratifying and pushed me to progress even further. ”

Two months after learning French with the software he had developed, Mait decided to test himself and take the national French language exam in Estonia, a test usually taken by students who have twelve years of studying French under their belts. He passed! In only two months, Mait had succeeded in attaining a nationally recognised French qualification, but was he able to converse in French?

“I didn’t have a component to support conversational French, but in the first month, I did try and speak French with French people. Before trying, I was under the impression that the French didn’t really speak English, but as soon as I made the effort to try and speak French, they would switch to English to help me! I realised that trying to speak their language was building bridges. It did have the downside that I couldn’t practice my French! After a couple of months, I improved and they stopped replying in English. For me this was a sign of quality, the software was working!” 

A CEO is born!

Taking the plunge

Mait had stumbled across something very powerful and was convinced that the underlying technology could be applied to other fields of education. Impressed by the results, his friends wanted to use the new method he had developed. Mait realised that he had to do something with his idea. He went on holiday, hired two software developers to take his code and develop it so it would work on the Web.

“Whilst on holiday, I happened to meet a friend of a friend, later to become my Lingvist co-founder, who was keen to work with me on my idea. During our conversation, he registered the company and that was that, we were in business. In Estonia, it is relatively simple to register companies; furthermore, Estonia has a fantastic start-up ecosystem. Thanks to Skype, we have a great software development culture. Neither of us had any experience in running a company, but we went ahead and hired a couple of people and started raising funds for the company.”

Before he knew it, Mait had become CEO. He extended his holiday and announced to his boss at CERN that he was leaving to run his own company.  Fortuitously, Mait met the technical co-founder of Skype at a conference, who coincidentally had been working on building some software to accelerate human learning. He dropped his attempts as soon as he saw Mait’s software and became their first investor. Things were falling into place, but Mait explains that launching a start-up is not a bed of roses;

“You can use the analogy of sitting in a nice warm office at CERN, surrounded by beautiful mountains. In the office, you are safe and protected, but if you go outside and climb the mountains, you encounter rain and hail, it is an uphill struggle and very uncomfortable, but immensely satisfying when you reach the summit. Even if you work more than 100 hours per week.”

An investment in knowledge pays the best interest,
                                                                            B Franklin

Constantly curious, Mait wondered, what if people could learn twice as fast? What is the value of accelerated learning?

“If people could learn something in 1000 hours when it usually takes 2000 hours, what is the value of the free 1000 hours? For a start, you can do something productive with the 1000 hours. Secondly, if you take the average salary, you will find that the cost of 1000 working hours for a population (per person) is worth more than the defence expenditure for most countries.” 

The language learning software developed by Mait, Lingvist, currently has 3 million users and counting. University studies have demonstrated that it accelerates learning, significantly. Indeed certain schools in the USA have started using the software and classes have seen overall improvement, not just in the most talented, but also in those less capable students. Teachers are seeing incredible progress, so much so that they are covering topics in the 1st year, usually taught in the 3rd year. At a recent State level language learning competition, Lingvist students earned 1st, 2nd and 3rd place!

Boundless opportunities to accelerate learning

If the technology works in language learning, Mait is convinced it can be applied to all types of education. Lingvist utilises the brain’s natural way of learning using AI and machine learning. Redolent of scientific work carried out at CERN, Mait draws parallels to his company’s activities and CERN’s,

“At CERN you collide particles and collect data, at Lingivist, we also collect big data, and like physicist, we build models, more precisely, cognitive models about how people learn. What our data have demonstrated is that levels of learning in people are very different. Short term memory capabilities can differ between 5 minutes and 2 seconds! Currently based on our data, the older generation has much better memory characteristics. The benefit of our software is that it measures memory, and no matter one’s retention capabilities, the software will help improve retention rates.”

Faced with a future where AI will make numerous jobs extinct, and many people will need to retrain, competitiveness will be derived from the speed at which people can learn. Computers can be harnessed to help human learning. Mait enthuses that with the growth in his company and its entire infrastructure, coupled with the influx of massive amounts of data, he is now building his data science research team.

Hiring from the CERN talent pool

Unsurprisingly, Mait wishes to hire from within the CERN talent pool. Traditionally, physicists have excellent modelling, data analysis and machine learning skills, even though they might not be aware of it. Furthermore, they know how to run experiments and have great coding skills.  Additionally, there are other advantages to joining the Lingvist team,

“The environment is great. Working at Lingvist means you can continue working in a scientific research milieu, you can work with the same legendary technical team that built Skype, and you benefit from a fast feedback loop if you come up with an idea you want testing. It’s really dynamic! If you are interested, find out more on” 

Mait Muntel at CERN's first LinkedIn Live

As a member of the CERN Alumni Network, Mait is delighted to witness the growth of such a valuable community. In October, he featured in CERN’s first ever LinkedIn live, broadcast during the Moving out of Academia to Entrepreneurship event and was amazed at the positive impact and numerous contacts made.  When asked about his hopes for the future development of the Network Mait quotes mathematician, economist and physicist Eric Weinstein *), “The theoretical physics community is bar none, the most profound intellectual community we have created.”* Mait adds, “I am convinced that the CERN Alumni Network will give rise to a multitude of ideas and companies.”


Author: Rachel Bray

Other news

Cookies help us deliver our services. By using our services, you agree to our use of cookies.