Synthetic content is becoming increasingly more important

Synthetic content is becoming increasingly important and will help prepare teaching content in the future. At CodeBrainer, we research synthetic content primarily because it gives us the ability to produce rapidly changing content in STEM teaching (especially coding) faster. For example, when the layout of the code editors changes, the content needs to be re-recorded. Therefore, we have developed a way to make such content with the help of artificial intelligence, which prepares speech, mouse movement around the screen and typing texts. This reduces input errors and speeds up re-recording.

What is a synthetic course and what is synthetic content?

First, we would like to clarify what we mean by the term synthetic course. The term synthetic course is used to describe educational content prepared using artificial intelligence (AII). The preparation of this content could be classified in the segment of synthetic media , where we used synthetic speech (text-to-speech (TTS)) and technology supported by AI for the preparation of video and audio content, including mouse movements and keyboard emulation.

Types of synthetic content

The main reason for the growth of popularity of synthetic content has mainly been with the progress of deep learning especially GPT2 and GPT3, where the creation of text and also creation of images has become very useful. As a result algorithms for synthetic text creation have become much more creative and more natural. This can be useful displacing monotonous repetition e.g. when writing multiple ads, UI can offer us many options than we can choose from.

What are the types of synthetic content:

  • speech
  • text (natural-language generation)
  • sound
  • image synthesis
  • interactive multimedia
  • ...
Synthetic content - Images - e-Dale - 1
Synthetic content - Images - e-Dale - 2
Synthetic content - Images - e-Dale - 3

How do we use synthetic content at CodeBrainer?

At CodeBrainer, we are developing an AI Teacher named Cobie AI (artificial intelligence) who is able to explain the content, follow the students and help them with their work. Our teacher's name is Cobie and we are very proud of her. At the same time, we can rely on it more and more. From the beginning of the journey, we decided to actively develop tools that can create synthetic content. Synthetic content will help us a lot in the long run, as we can adapt the content faster. One more benefit is that the content can be adapted to students' needs, enabling personalised learning. Cobie AI already helps us create video material, check the code our students write and create full-blown synthetic lectures. As we gather more data, her knowledge base will grow, and our AI models will get better, which will open doors to possibilities that we can't imagine today.

Why use synthetic content in education at all?

The trend of education in all segments is in a personalized approach. If we want personalized content at a scalable level, we need to use all the tools we can. We all know we have a limited number of teachers, professors, mentors… and we have to boost their knowledge. Make no mistake, we are not saying that artificial intelligence will replace teachers. We are saying that teachers could make great use of tools that would enable them to teach students better in a more personalized way and on a broader scale. This is true at all levels of education, from elementary school to adult learning. And here synthetic content can help us a lot, because we can use AI to quickly adapt, upgrade, change, supplement the content…

Synthetic voices are getting better

"This voice still feels robotic!" Alexa, Siri and Nadja, don’t be offended, we don’t mean anything personal with that. Many times there is a concern that the synthetic voices are too robotic and impersonal. Partly this is true, but technology is changing so fast that we can expect a very good human voice simulation in the near future… Much progress was made a few years ago when Google introduced WaveNet and offered artificial voices that are already very close to the human. Now all competitors are making a lot of progress in more interesting perspectives of synthetic speech e.g. Microsoft with cartoon voices, Google with voice from John Legend.... Just imagine what kind of possibilities that synthetic voice opens with this field of using voices from famous actors, characters, presenters

The biggest advantage of synthetic voice is that it never gets tired and it has an infinite amount of time available. Imagine a teacher who has to tell each student the content in a slightly different way, repeat it several times, use as many practical examples as possible, in all European languages. Unfortunately we humans are limited to 24 hours a day and one teacher can deal with a limited number of students, despite the desire to help as many students as possible.

Our AI teacher also speaks Slovenian!

This is a unique feature for our students. You can use this feature to compare your code and see if you have any bugs.

There are several text-to-speech services available in the cloud and those services support most popular languages already. But we come from Slovenia, Slovenia is a small country with a population around 2 mio people. And having the slovenian voice for speech synthesis is not available on most services. We are proud to say that our AI teacher also speaks Slovenian. We decided to use the eBralec speech synthesizer, developed by the Slovenian company Alpineon. With their solution and implementation in our platform, our teacher now speaks also in Slovene. This opens a new horizon for the preparation of educational content in the Slovenian language. At the same time, we want to include Slovenian knowledge in our development, and we are on the right track with such connections.

Our AI teacher - Cobie AI writes the code herself

Synthetic content is becoming more and more diverse. Speech is just one of the aspects of synthetic content prepared by Cobie AI. Cobie can go a bit further. She enables us to manipulate various applications on the screen using AI, she can write code, control the mouse, this allows us to prepare a completely synthetic lecture (audio and video). Content can be also prepared for other domains, not just for learning programming. The result is a synthetic lecture, which is a video of the presentation on the screen, recorded commands, writing code in the code editor,…

When teaching programming, it is important to have control over the editor and to be able to record examples of code being written as easily as possible. Cobie AI can automatically convert the code into instructions for the video where the writing of the code is shown and supported by spoken explanation. This is especially important so that we can quickly correct, upgrade the video for code examples. This is also a great basis for the next challenge, personalized learning, where each student will have a slightly different content.

Can synthetic video content be personalized?

Content, tools and, of course, time are needed to make a quality educational video. Each adjustment takes extra time as the video needs to be re-recorded. Let alone that each student would have a video adapted to him/her. We also have to make it simultaneously in multiple languages.

Ability to change video content in real time is already here. Together with the use of translation services and a speech synthesizer, we can have a lecture in multiple languages simultaneously. Development of content synthesis and UI goes in the direction where we will be able to support the individual student's learning process with the content that he/she need the most at any given time. We are now just a year or two away from full personalization and automation, but the existing systems allow us to improve efficiency in the learning process a lot.

A good example can be a math tutorial, where we are able to offer students a practically unlimited number of exercises. Our teacher can persistently explain each case step by step and many times as needed. Compared to a human teacher who can only showcase one or two examples for the whole class.

We invite you to view a part of our synthetic lecture.

Will synthetic content replace teachers and their lectures?

At CodeBrainer, we see the progress of artificial intelligence and with it the ability to prepare synthetic content as a good tool for teachers, lecturers, writers, ... Synthetic content and tools offer flexibility like we did not know before. In education we see this progress as something that in the end it offers even more creativity than before. It offers opportunities to prepare more examples, more detailed examples, and therefore spend less time preparing content, more time for mentoring and teaching. This way we give more to the student, and teachers are relieved of repetitive tasks and so they have more time to add a personal note, motivate and help those who are the best and those who need a little more help to make progress.