An in-depth discussion about an entrepreneurial AI journey
Killian Vermersch, CEO and Founder of Golem.ai, has given us an insightful look into his personal and professional journey. His company is at the forefront of natural language understanding, driven by AI to interpret and process human language. In an illuminating conversation with Killian, we explore the ascent of AI, its multifaceted applications, and speculate on its potential evolution.
Killian, could you introduce yourself and shed light on your entrepreneurial path?
Certainly! My journey as an entrepreneur began quite young, as I started freelancing in web development at 18, learning to code through Open Classrooms. After some time working in the insurance sector, I began my studies at Epitech, an engineering school known for its project-based learning approach.
There, in 2014, I co-founded Idequanet, a web agency offering tailor-made solutions. Two years later, in 2016, my co-founders Guillaume, Thomas and I decided to launch Golem.ai, an AI company specializing in Natural Language Processing (NLP). After a challenging first year, we secured €5 million in late 2020 and early 2021 in Series A funds.
At Golem.ai, our solutions aim to simplify the work of our clients and to help them reduce time spent on low-valued added tasks. Our solutions are geared towards automating tasks, improving customer service, and enhancing applications through a deep understanding of human language.
How did you decide to integrate AI in your product? What was your first AI experience?
Our identity as a deep tech company was clear from the very beginning. We started with AI as our core technology, and we designed our products around it. Our initial vision was always centered around the idea of understanding human language. However, we soon came to realize that possessing cutting-edge technology was not enough in itself: we needed to develop tangible products!
Collaborating with clients, we developed up to five products, all based on our technological core, AI and natural language processing. But by 2020, it became evident that our product range was overextended.
We thus decided to streamline our strategy around two main products, InboxCare, designed for incoming messages and document attachment processing, and DocuChecker, an advanced document finalizer. Ultimately, we narrowed it down to just InboxCare, to be able to develop a fully coherent strategy around one product.
For a few months now, AI has emerged as a popular topic of discussion. What’s your take on its rising popularity, and do you see it as a game changer?
Recent models have experienced a rise in popularity for two main reasons: accessibility and similarity. Tools such as ChatGPT are groundbreaking because they’re easy to use and generate results previously thought to be solely within human capability.
However, the view from within the tech industry is different. Professionals who’ve tracked the evolution of these models over the past five years, and observed a spike in internal investment since 2020, may find this progress less spectacular than the general public does.
Why the divergence in views? From an industry insider’s point of view, generative AIs are still lacking maturity.
AI tools, such as ChatGPT, have now reached mainstream audiences, becoming topics of discussion from tech influencer platforms to casual dinner chats among family and friends. All these AI solutions share a common trait: they are generative, which means that they are designed to produce or “create” content. Currently, we make a distinction between generative AI and analytical AI. The latter produces structured outcomes like categories or extracted data.
For example, generative AI might generate a persuasive email response, but if the aim is extracting email content to update a CRM system, analytical AI would be the tool of choice.
But here’s the thing: analytical AI is not as spectacular as ChatGPT.
However, the true achievement is that AI has now gained widespread recognition and popularity. From a technical point of view, generative AI is not yet mature. ChatGPT still needs quite a long and precise prompt to operate properly. But from a broader perspective, it is paving the way for AI to become more and more mature.
Regarding the impact of AI, I’m convinced that it has the potential to be the next industrial revolution! It is definitely going to change the world and the way we work.Killian Vermersch
Based on your insights, it seems like AI still has a long journey ahead. How efficient is today’s AI, and how would you characterize it?
My take is that AI should always be seen as a tool, an ally, something we can rely on, but definitely not as a replacement for human beings! I see it as a very powerful tool, that can sometimes even exceed our expectations and impress us, but nothing more.
At its core, it is designed to help us – and it’s designed by us. Generative AIs such as ChatGPT are already very helpful, but they’re not yet fully mature.
My vision of AI is that it should be imbued with values. Its construction should not be disconnected from essential values, such as sustainability for example. At Golem.ai, we’ve always prioritized an eco-friendly approach, emphasizing frugality and lightweight technological infrastructure. We’re transparent about our green initiatives.
Lastly, I think that we need to collectively reflect on the concept of AI. A frequent criticism of AI is that it lacks “true intelligence”, since it largely relies on repurposing previously acquired knowledge. But isn’t that what humans always do as well?
In Luc Julia’s book “L’Intelligence Artificielle n’existe pas”, he explains that machines do not really think or understand like we do. Instead, they process information and follow instructions, which can give the illusion of intelligence.
The reality is that AI excels at very narrow tasks and will outperform humans in speed and accuracy on these specific tasks. But in the end, it is still just a tool to amplify our capacities, not replace them, and humans remain central in technology’s evolution. It seems that Mantu’s motto is not wrong, as even within an AI-driven world, “tomorrow will be human” after all!