At Softlandia, we specialize in applied artificial intelligence and our goal is to make AI accessible to every one of our partners. Applied AI is a combination of modern software expertise and a deep understanding of contemporary AI methods. This means that AI and various AI methods and models are increasingly a part of both new and existing software solutions. It could even be said that an understanding and utilization of applied AI is a prerequisite for implementing competitive digital solutions in the future – in industries, SaaS products, and the public sector alike. Wherever code is needed, AI is likely needed as well.

An example of applied AI was witnessed by over half a million Finns during a live special broadcast of the Arto Nyberg talk show just over a week ago: we had the opportunity to implement a real-time talking and moving AI avatar for the Arto Nyberg program. We owe thanks to Production House and Liisa Akimof for this. Mikko Lehtimäki from Softlandia was also interviewed by Arto Nyberg about AI.

We have been asked how the AI seen in the program was implemented and some have even doubted the authenticity of the conversation. The conversation is authentic: Arto chatted with the AI as he would with any person. Depending on the models used and their quality, responses may take a second or more. The conversation itself is already very natural. You can even ask the AI to tell you what you've already discussed, and the AI will summarize the entire conversation.

Technically, this involves seamlessly gluing together several different AI models and various optimization tricks to minimize response times. In the case of the avatar, there are also challenges related to real-time video playback and synchronization of facial/lip movements, and modifying the avatar character is extremely labor-intensive. Additionally, displaying the avatar, especially in real-time, is still relatively expensive – one euro per minute is a good cost estimate. We will publish a more detailed description of the technical implementation in a separate blog post later.

We have been working with applied AI since the founding of Softlandia and have learned a lot along the way, which we were able to utilize in this implementation. Starting from scratch would have been difficult. We also learned something new: avatar technologies are still in the early stages of development, and most of the services we went through are mere promises without a real, functioning, quality implementation. However, many different models for converting speech to text and text to speech are already quite usable for various applications.

AI is making huge leaps and developing so quickly that it's hard to keep up. Our task is to help track this change while implementing the best possible AI-assisted software solutions. Finally, we can offer one tip about AI: chat- and conversation-based implementations are just a small scratch on the surface of the possibilities offered by AI, or more precisely, language models. Real savings and benefits come from integrating AI seamlessly into software solutions.