InstaText, Slovenian start:up of the year 2020
15. 07. 2020
InstaText is a web-based platform that uses artificial intelligence for text improvement. It behaves like your personal writing assistant who rewrites your texts like a native speaker. In a matter of seconds.
InstaText is very useful for improving English texts such as articles, projects, and marketing content. The tool provides very useful suggestions that help to bring the text to a professional level, so that no additional review by "native speakers" is required.
Matej Guid, the director and founder of InstaText, is a member of a tech community from Technology park Ljubljana. The company behind the web platform of the same name is the winner of this year's Startup of the Year Award in Slovenia.
In this interview, Ester Snedec from Association for the Development of Voluntary Work Novo Mesto spoke to Dr. Matej Guid about artificial intelligence and its role in informal learning and text improvement. Dr. Guid is also an assistant professor at the University of Ljubljana and works as a part-time researcher at the Faculty of Computer and Information Science. The interview was conducted within the Erasmus+ Youth Ambassadors project of Non-Formal Learning.
First of all, congratulations on the best Slovenian start-up company. What does the award mean to you?
Thank you very much. The Slovenian Start-up of the Year Award is an extraordinary encouragement and confirmation for us that our ideas have fallen on fertile ground. We have been recognized as "a company that has long-term ambitions, the potential for global growth, an irresistible competitive advantage in its niche, a company that stands firmly on its feet".
Please introduce your company briefly.
InstaText is a web platform which uses artificial intelligence to create instant recommendations and ideas about how to improve a text content-wise and make it more readable and understandable. It is based on the observation that clear and effective written communication is the key to personal success, and using artificial intelligence and language technologies we can take written communication to a whole new level.
Writing in English can be stressful for many people, as our vocabulary is not rich enough and we do not express ourselves as fluently and richly in writing as people who, for example, have spent their whole lives in England. Translators are out of the question, as we need improvements immediately, and existing online tools focus primarily on grammatical corrections. In the first few months since the launch of the platform alone, over 5,000 active users from more than 50 countries from around the globe have registered to the platform. The company is based in the stimulating environment of the Technology Park in Ljubljana.
You have a doctorate in computer science. What about the other employees in your start-up?
I have a doctorate in the field of artificial intelligence; one of the co-founders of the company Dr. Marcus Hassler from Klagenfurt holds a doctorate in language technology. The international team consists of experts in various fields, from software development, linguistics, user experience design to marketing, sales, and mentors and consultants for successful international growth. The majority of the team comes from Slovenia. Slovenia is a highly developed country with a long and respected tradition in the field of artificial intelligence. Of course, it is important to involve experts from different countries because we want InstaText to become a multilingual online platform.
As I mentioned, InstaText is a useful communication tool that I also use in my work. For all those who have not yet tried this tool (and which I warmly recommend op. cit.), How would you describe it? Who is it all for and what does InstaText actually “do” with the text we write? Is there any difference between your online tool and, for example, Grammarly and similar tools?
InstaText is based on the idea that we can rewrite the entire text as if it had been written by a "native speaker" or someone whose native language is the language in which we write. This means that the same text is rewritten, which can include corrections, including grammar, better vocabulary and word choice, and a thorough rewrite and optimization of the text. Unlike grammar checkers like Grammarly, which focus on grammatical improvement, InstaText not only corrects grammatical errors but also enriches the content, making it more readable and understandable. If necessary, the entire sentence is restructured so that it fits into the context.
Where did you get the idea to develop a tool that would focus on improving texts, not only to correct grammatical errors, as we do with Grammarly, but also to offer the author better more appropriate suggestions for the vocabulary used?
As a researcher in the field of artificial intelligence, which I have been engaged with since 2005, I have often been involved in writing scientific articles and other challenging texts. If we had a tool that would help me to write better quality texts in real-time, we could write much more and much better. Most people find it hard to admit that they would need some real-time support when writing in English. Tools like Grammarly and the like offer far too few ideas and helpful suggestions as they interfere with the author's text in a very conservative way. As a result, our cumbersome expressions in a foreign language often remain untouched. InstaText is based on a different approach: we prefer to offer too many ideas rather than too few, and then the author chooses what he or she wants to accept and what not. Here is an interesting thought from a university professor: "If I was going to a desert island with the intention of writing a scientific article or a book and could only take one tool with me, I would definitely choose InstaText because it provides far more ideas and recommendations than Grammarly. However, I am already competent enough to know which suggestions I can and cannot accept".
The development of InstaText would not be possible without artificial intelligence, which has been penetrating our lives more and more in the last 20, 25 years. From the Deep Blue supercomputer, which first defeated chess grandmaster Gary Kasparov in early 1996, to the AlphaZero program and robot Sophie, which received Saudi citizenship in 2017 - where are the limits of artificial intelligence? Can we even talk about limits?
Humanity is behaving quite irresponsibly in the face of technical progress. In his book Superintelligence: Paths, Dangers, Strategies, which has also become an international bestseller and was published by the renowned New York Times, Nick Bostrom argues about the problems we will face in this century, which may have a much greater impact on the lives of future generations than, for example, climate change. We might ask ourselves: "Will giant computer systems become smarter than us once they really understand what they are reading?" They read extremely fast and understand more and more. Just a little bit different than we might imagine. Humans are only slightly smarter than gorillas, for example, but this species is completely at the mercy of humans. While gorillas, due to their lack of information, probably live under the illusion that they have everything under control. Bostrom points out, among other things, that if intelligent computer systems are more intelligent than we are, this will only be the case for a short time: Because of the way they work, they will quickly become much smarter. Then it is pointless to talk about limits because we are not intelligent enough to foresee them. It amazes me that very few people are aware of this. But let's stick to the useful aspects of artificial intelligence, especially for improving written communication. In the phase of writing and editing our texts we often need professional help in real-time, and usually, there is no one to help us in this phase. We are proud and happy to work in a field that is ethically non-controversial and has great potential for improving people's lives.
Artificial intelligence is a challenge for society, which was already discussed by the European Economic and Social Committee (EESC) in spring 2017 and adopted the opinion that there are currently 11 such areas, in addition to ethics, security, respect for privacy, legislation, etc., also the field of education and competencies. Also in Youthpass, the certificate that participants for all activities in projects within the Erasmus+ Youth in Action and European Solidarity Corps can receive, one of the basic competences is digital competence. Do you think that artificial intelligence has a great impact on non-formal education or is artificial intelligence useful in non-formal education? In what way?
Certainly, artificial intelligence can be useful in many areas of education, both formal and non-formal. One interesting area, for example, is intelligent tutoring systems. It is generally accepted that individual teaching is significantly more effective than teaching in the classroom but too expensive in most situations. Intelligent tools based on the ability to tailor teaching to each individual can come to the rescue. In the USA, such tools are already widely used in mathematics and problem-solving subjects in general.
InstaText is, in my opinion, a great tool for non-formal language education. Currently, users can use it to improve texts in English and at the same time learn or improve their vocabulary. In the future, do you plan to add other languages to English?
I often hear from InstaText users how much they have improved their English after only a few months of using the tool. We have also noticed that we have started using English terms and various expressions that we did not know before. Above all, InstaText teaches us not to complicate things and to express our thoughts and ideas as effectively as possible. We intend to support at least the German language already this year. The aim is to support all the major world languages.
What about Slovene, which is considered a grammatically very demanding language?
Slovene is an extremely demanding language from the point of view of artificial intelligence and language technologies. Not only dual (besides singular and plural), inclinations, and declensions - there are many exceptions that we are not even aware of. But technology will eventually develop to the point where programs will be able to write excellent Slovenian texts.
Do you intend to add dictionaries to InstaText or adjust the use of vocabulary for a specific user or professional profile?
That is right; at the moment we are working on the introduction of personalized user dictionaries, which will allow each of us to adapt the tool to our needs and professional vocabulary. In this way, we will be able to ensure that the program will never change certain expressions. We will also be able to ensure that, for example, "credit scores" are never corrected to "credit ratings", although this may be useful in a particular context.
How do you improve InstaText vocabulary? Is it learning on its own?
Just as Google's algorithms breakthrough endless online content, InstaText can access a wide variety of publicly available texts from many different areas. Scientific articles, for example, are usually freely available via Google Scholar. A strong inspiration was the performance of the program AlphaZero, which learned to play chess, i.e. a very challenging game for a human, at the level of a world champion and even better. If a computer who has no idea about chess could learn to play it better than anyone else by playing against himself, could this approach also be used in computer-based learning of written communication? But how would the program know which texts are good and which are not so good? Such and similar questions had to be answered. It turned out that there is surprisingly much room for improvement. The artificial intelligence behind InstaText is progressing on a monthly basis, so to speak, and the quality of the suggestions the program offers surprises even editors and translators.
In 2017, the EESC issued an opinion on data protection in relation to artificial intelligence. How does InstaText take care of the protection of personal data, given that your program is used by companies and employees of various profiles who, with your help, correct their texts, whether for private or business' reasons?
We undertake not to store any user texts. There is also a text next to the web interface for text enhancement, which always lets us know that our privacy is not at risk: "Your texts are never stored". Users enter their text in the web interface on the left. On the right, they receive a text suggestion for text improvement in a few seconds. This text is immediately deleted from the server, and the user then edits the text locally, in a web browser, and can do so without an internet connection – in short, no one has access to their work. The advanced user interface in the browser allows users to easily accept and reject the suggestions offered so that they quickly receive a clear, high-quality text. The final result is then copied with a simple click of a button and can already be used in any other application on our computer.
In your opinion, what is the future of non-formal education or education in general in connection with artificial intelligence?
In any case, the role and mode of education are changing. Some online courses have already reached an enviable level of quality, while they also cover areas that traditional universities have not even been able to focus on. The newer generations will increasingly choose to acquire higher education and useful skills, particularly through online education giants, i.e. outside the usual framework. However, it will also be necessary to ensure the supervision and proximity of the human teacher. As there will never be enough high-quality mentors for everyone, artificial intelligence will play a very important role in education.
Finally, if I go back to the development of artificial intelligence - for example, Sophia: it is visible that she is a robot, but in terms of expression, sparkling and witty answers in which she even jokes, I could say that she is a very intelligent woman; or AlphaZero: a computer program that has learned to play chess, walk, and shogi (Japanese versions of chess, op. cit.) - should we be afraid of the development of artificial intelligence?
I recommend reading the aforementioned book by Nick Bostrom or at least the summaries of the book. It will be best for everyone to form their own opinion on this subject. The AlphaZero program has clearly shown that intelligent programs can overtake us in unexpected ways. Moreover, they can do this much faster than we expected.
We certainly do not have to fear the many tools that can help us to be happier and more successful if we use them properly. When improving texts – with or without the help of artificial intelligence – it is of course important what content we write about and in what way and for what purpose we try to get our message across. Clear and effective written communication can be extremely powerful. Similar to artificial intelligence. Ultimately, the way in which this power is used depends primarily on people.