Introduction
Ryan: Welcome, everyone, to another edition of the AiFounders podcast show. Our podcast is dedicated to celebrating the remarkable accomplishments of AI innovators, entrepreneurs, and visionary founders and the captivating stories behind the movements they have built. I’m your host, Ryan Davies, and I have the honor of hosting today’s episode, the business impact of rapidly changing AI regulations worldwide, with our special guest, Kanstantsin Vaitsakhouski.
Kanstantsin Vaitsakhouski is the head of AI and technical product management at A.P.R.O Software, with over 12 years of experience in AI startups and product discovery. He leads a team of AI engineers and researchers focusing on healthcare, smart mobility, and more. He’s passionate about deep analysis and technical excellence, and his expertise extends across multiple roles in AI software development, project management, and product discovery. This is going to be a great topic. What a perfect guest for us to have. Thank you for being here. I can’t wait to dive into this topic. It is an incredibly interesting one.
Kanstantsin: Yeah, great. So I’m happy to be here.
Ryan: We’re focusing on AI regulations, how rapidly they’re changing, the challenges because it’s different from country to country, region to region. But there are a lot of concerns about this. So, let’s start there. What are some of the concerns that you’re hearing today about AI regulations from the business stakeholder standpoint?
AI Regulations Worldwide
Kanstantsin: Honestly, there is a lot of concern about this point because the regulations have become slightly unpredictable. For example, here in the European Union, they try to regulate everything, all aspects of AI technology. They are not case-specific, but they try to create a unified document, unified regulation, and unified compliance, covering everything, starting from data collection to interaction with models and end-users. The main concern is that it needs to be simplified, and it can also ban many nice projects and startups from the market. Unfortunately, that’s true. We may not have a dark future for AI. Still, there is a significant risk that the European region will face a small AI winter because the regulations will bring more technical difficulties and unforeseen expenses, and many companies have decided to move to the USA or Asia.
Ryan: Interesting. With that in mind, I know we had this for later in our discussion, but I want to pivot to that right away. Let’s talk about AI regulations and how they’re different from place to place and the cause and effect. You’re in the EU, I’m in North America, and we have different regulations. It’s a bunch of legislators trying to keep up with some of the fears. But you just mentioned in Asia, it’s more like the Wild West, a little more open. Tell our listeners about that and the challenges depending on where you are.
Kanstantsin: Currently, we have a lot of differences, which is good. For example, in the US, AI is case-specific, so they create special rules and compliances to cover specific cases. This approach is more suitable for businesses. In the European Union, they are trying to regulate everything and create a document about everything, which can even ban technologies like facial recognition and emotional recognition from the market. Face detection and face recognition technologies, which are now a part of many smart home applications, are subject to this regulation. Right now, this act says that only security agencies can use these technologies in very important cases. They also divide technologies and domains by the degree of risks. For businesses from the US trying to work in the EU, they must be compliant with EU regulations, even if their technology was developed abroad.
Ryan: That’s a great point. For AI founders looking to scale or start up, how do they ensure they have someone up to date on these regulations, and how do they collect this information since it’s changing fast?
Kanstantsin: We are in a grey zone because these regulations were somewhat unexpected. They are still in the process of development, but there is a high probability that this act will be introduced next year. Companies working with AI, even small teams, spend much time on data collection, preparation, and validation. This regulation makes a significant impact on data. It’s not entirely negative; it focuses on copyrights and private self-information, which is good. However, companies already in the product development process must revise their data. We need real use cases on how to comply with this regulation, so companies will have to invest more money in preparing documentation and transparency procedures. It will create a new domain of companies and legal agencies specializing in AI regulation and compliance documentation.
Regulations and Compliance
Ryan: It’s unbelievable to think about all these regulations worldwide, from compliance issues to legal aspects and the fines and consequences. We also talked about the myths behind AI. What are some of these myths causing things to slow down?
Kanstantsin: The main problem is that the people creating these regulations must understand how AI works. For example, what a model is. It’s a huge black box for them. They may think these models only perform simple tasks, like predicting the next word in a response based on the input. The myth and prejudice are based on old science fiction from movies and books, where AI is portrayed as an extremely emotional and intelligent entity that’s always trying to harm someone. Unfortunately, in reality, it doesn’t work that way. AI is just a tool, and it only does something once it receives a specific command. Even a simple object like a piece of paper can be dangerous if used improperly, and AI is no different.
The main point here is that people don’t understand how AI works, and they want to introduce additional transparency procedures to make sure AI models can be explained. However, describing how one AI model works is extremely challenging. To do so, you would need to create another AI model that can provide explanations because it’s very complex for humans. Artificial neural networks mimic how our brains work, although it’s a simplified model. But here’s the hypocrisy: when a person commits a crime, nobody understands why they did it deeply. There needs to be a thorough analysis of their biochemical processes, motivations, environmental factors, or genetic predispositions. Yet, when it comes to AI, we demand much more detailed explanations for these models than we do for humans, which is quite absurd.
Ryan: I’ve heard several times that the problem is that legislators need an understanding of how AI works. They often rely on myths, conjectures, or theories that don’t reflect the reality of AI technology. Moreover, the pace of technological advancement often outstrips the time it takes to draft and pass legislation, rendering the regulations outdated by the time they become effective. It’s a situation where regulations try to solve problems that either still need to exist or have already been solved through technological evolution.
Kanstantsin: Exactly, they’re addressing problems that are sometimes more theoretical than practical. Take the example of fighting fake news. A year ago, everyone was concerned about fake news being generated by AI. However, today, we see a manageable surge in fake news generated by AI. Humans have always been quite proficient at developing fake information without AI. Yet, there’s a disproportionate focus on AI’s potential for creating fake content, which aligns differently from reality.
AI and Its Practical Use
Ryan: That’s a valid point. Often, the heat of the debate is centered on creating a diversion and attributing problems to AI when they may have more to do with other human factors. AI is indeed a convenient scapegoat. The agility to adapt to rapidly changing regulations is crucial for businesses because regulations are evolving at an unprecedented pace. Many companies want to avoid investing in advanced AI technologies, worrying they won’t comply with stringent regulations. For example, even tasks such as detecting and classifying certain images or content raise concerns about legality due to these regulations. Companies are contemplating hiring humans to manually validate AI results, effectively turning AI into a producer of routine tasks, contradicting the original purpose of AI to simplify and streamline processes. It’s a fascinating paradox where AI, intended to alleviate burdensome and repetitive tasks, is subjected to regulations that complicate its implementation and shift the burden back to humans.
Kanstantsin: Exactly, it’s a counterproductive loop. AI was created to make our lives easier and reduce mundane tasks. However, due to regulations, AI is now completing routine tasks for humans.
Ryan: This full-circle irony highlights the need for a more balanced approach to AI regulation.
Kanstantsin: It’s a perspective that should be considered.
Ryan: And it’s not just the regulations; the challenge also lies in the fact that regulators might need to be equipped to understand the complexities of AI systems.
Kanstantsin: That’s true. The rapidly evolving AI landscape often leaves regulators needing help to keep up. The lack of a deep understanding of AI further complicates the regulatory process.
Ryan: Right, and it’s an evolving field. AI keeps advancing, and the regulations need to strike a balance between oversight and enabling innovation. Businesses must stay agile and adaptable to navigate this complex regulatory landscape effectively.
Kanstantsin: That’s the key. Staying agile is crucial to success.
Adapting to Changing Regulations
Ryan: Before regulations are introduced, businesses should focus on staying informed about the changing AI landscape, understanding the potential impact of upcoming regulations, and aligning their strategies accordingly.
Kanstantsin: Being proactive and anticipating regulatory changes is a smart approach. It allows businesses to prepare and adapt effectively.
Ryan: Absolutely. It’s about being ready to pivot when needed and making informed decisions to ensure compliance with future regulations.
Kanstantsin: Precisely, adaptability and strategic planning are essential.
Ryan: Well said. Thank you for sharing your insights on this complex and evolving topic.
Regulations and Their Consequences
Kanstantsin: My pleasure. It’s crucial to shed light on these issues and promote a balanced understanding of AI regulations. I believe regulations will be introduced in some form, and you must prepare now. You need to be cautious and vigilant about your data sources. Ensure all data is copyright-free and can be traced back to the sources. Understand the properties of all data in your datasets. Additionally, you must adapt your development process and roadmaps. For AI solutions in critical areas, every change during the product’s life cycle must be documented. Drafts for documentation can be found online, thanks to those who have shared them. Your architecture should be extremely detailed, and you must consider transparency in explaining how the model works. Several methodologies already exist, although they may be complex and require additional tools and special dashboards. Having this information now is beneficial, but remember that the more information you can document, the better, as the regulatory landscape is still evolving. Preparing for regulation is essential, focusing on data transparency, process transparency, algorithm transparency, and proactive documentation.
Ryan: This advice is invaluable for our listeners. It’s crucial to be proactive and start immediately before the regulations affect your business directly. Thank you for sharing these insights. Our discussion has been fantastic, and I look forward to more conversations in the future. I’d like you to tell us more about A.P.R.O software and where people can connect with you.
Kanstantsin: You can find me on LinkedIn, where I consolidate my professional activities. Regarding A.P.R.O software, our headquarters are in Prague, and our main development office is in Warsaw, Poland. I’m based in Warsaw as well, so if you’re in the area, we can have a meaningful conversation and exchange our insights. I truly appreciate the opportunity and look forward to connecting.
Ryan: That’s wonderful, and I encourage our listeners to take advantage of such an invitation from someone like Kanstantsin, who possesses valuable knowledge and expertise in AI. It’s an opportunity to engage in discussions. Thank you for joining us on this enlightening journey through AI innovation, Kanstantsin Vaitsakhouski. We hope you’ve found inspiration in his stories and advice. The future is shaped by pioneers like Kanstantsin, and AI holds limitless possibilities. Stay curious, innovative, and explore the boundless horizons of technology. If you’ve enjoyed our podcast, please leave a review and subscribe on your favorite platform. Your feedback and support help us bring you more incredible content and guests like Kanstantsin Vaitsakhouski. Until next time, thank you so much for tuning in.
Kanstantsin: Thank you. It was a pleasure to participate in this podcast. It was great. Thanks.
Ryan: Wonderful. Thank you so much. This is Ryan Davies signing off. Until next time, everyone, take care.