Dembrane‘s AI transforms hours of talk into actionable insights. A simple tool records, transcribes, structures, and analyzes your conversations with stakeholders in minutes. Dembrane has found a way to survey large groups of citizens on a given topic and, with the help of artificial intelligence, distill from them the main conclusions and nuances of what is really going on within the group, including the corresponding sentiments. Bram Delisse, process and team lead, answers our questions.
The AI Pitch Competition
The AI Pitch Competition is a Brabant initiative aimed at showcasing the most innovative AI solutions and allowing startups to present their ideas, connect with industry leaders, and accelerate their growth. The competition is taking place again this year. Ten ambitious AI startups have been selected. The finale will be held on November 7, 2024.
1. What specific AI technology or algorithm is at the core of your solution, and how does it differentiate from existing approaches in the industry?
“At Dembrane, we are using state-of-the-art embedding, clustering, and transformer models to power our in-house developed large-scale conversation data analysis tool, which is based on much democracy and deliberation research (like cross-pollination or sortition). This way, we uniquely merge AI innovations with democratic innovations to build the infrastructure for deliberation at scale.”
2. How scalable is your AI solution, and what are the primary challenges you anticipate in bringing it to a global or large-scale market?
“The technology is extremely scalable by design. What we anticipate will be more challenging to scale is the thinking required for using Dembrane: a strong belief that people know-how and that listening to the whole community will lead to a better future for all.”
3. How does your startup address ethical concerns about bias, fairness, or transparency in AI decision-making? Can you provide an example of how you mitigate such risks?
“When claiming to facilitate the whole community to be heard, one has to operate with great care. Bias, fairness, and transparency are all significant topics for Dembrane. Currently, we are already tackling these in multiple ways. Firstly, transparency is strongly boosted by our data analysis tool, which only uses actual quotes from the conversations to generate insights. This way, it’s not the AI talking; it’s the community talking. Next, we are striving to minimally use AI in our tooling – only where required to ensure every voice is heard. We elaborate further on this topic in this blog post. Shortly, we plan to set up our ethical research division within Dembrane to research the ethics of our impact.
4. In what ways do you believe your AI solution can positively impact society? Do you foresee any unintended societal risks or challenges, and do you prepare to handle them?
“We need to talk. Communication is essential for a community to collaborate constructively. We’re pretty bad at inclusive conversation at scale, and we believe we see this back in our struggling collaboration. By building the infrastructure for large-scale conversation, we hope to contribute to a world where communities get larger, wiser, more connected, and more fair because everyone can join the conversation. Unintended societal risks consist of people in power using the infrastructure to learn much about the community without actually acting upon the ideas of the community. We plan to handle these risks through various methods, such as automatically full transparency of the results for every participant or only being able to use the digital infrastructure through licensed facilitators. But these are still ideas and need yet to be worked out.”
5. What has been the biggest hurdle in launching your AI startup, and how did you overcome it? How has the current AI landscape shaped your business model and growth strategy?
“The biggest hurdle in launching Dembrane has been balancing our consultancy-based process with product development. As a young team of now five full-timers and one part-timer, we’ve had to juggle serving customers while simultaneously building our internal tooling. We overcame this challenge by adopting an iterative approach and continuously improving our tools based on real-world applications. The current AI landscape has enabled us to move quickly.”
6. How do you prepare for the increasing regulatory frameworks around AI, such as GDPR, AI Act, or other data privacy laws? How does this impact your innovation and development process?
“We currently address privacy by not tracking who is saying what. This way, only the ideas and critiques are stored, not who stated them. We are GDPR compliant by handling all personal data on self-hosted models (no personal data is used for training) or on servers hosted in the EU. We’ve examined the AI Act, which does not affect us significantly. We’re convinced we can build our infrastructure while building ‘minimal risk’ AI. In the short-mid term, we expect to hire someone fully dedicated to privacy and cybersecurity.”
7. How do you see your technology evolving over the next 5 to 10 years, and what role do you expect your company to play in shaping the future of AI?
“We envision Dembrane to become a democratic membrane within any community that enables conversations where every member can join and be heard. This way, we hope to contribute to a world where, as communities get larger, they get wiser, more connected, and more fair. In the longer term, Dembrane’s role will, next to developing and maintaining the infrastructure, significantly consist of research on what it means to have a conversation at scale and to educate communities on the correct usage of Dembrane through some academy.
Then, we will discuss how Dembrane could play a role in shaping the future of AI. One of the central challenges in AI development is the problem of AI alignment: we would like AI systems that think and act according to how a community would like them to think and act. Next to the technical challenge of incorporating the voice of the community, a significant challenge is finding out how a community would like an AI system to think and act in the first place (this problem is nicely described in this blog post). We envision Dembrane taking a central place in this process. We’ve explored this before in collaborative research with OpenAI (the report).”
8. Why are you going to win the AI Pitch Competition?
“We believe that by empowering conversation and democracy with Dembrane, we are tackling a difficult and fundamental challenge that aids with many of humanity’s other challenges. The next generation of AI models and chatbots will come relatively easily (their discovery aided by AI). The real difficult but critical challenge is involving people in the process. By creating AI-powered infrastructure that is built upon democratic values and grass-roots approaches, we ensure human society and AI develop symbiotically.”
Collaboration
Related Posts
