Bringing history to life: How Know-Me uses AI to build meaningful connections to the past

At a time when AI is often in the media, our client Know-Me has used the technology to create something truly groundbreaking. Know-me wanted to find a new way for people to interface with history and the stories of the past. We’re thrilled to be a part of it.

By combining historical records with the latest ML and generative AI, Know-Me allows users to converse with recreations of the past. Our ML specialist team, led by Damien Duff, has gone to great lengths to ensure these interactions are handled with the sensitivity they deserve. 

In the latest edition of our Daemon Drumbeat podcast, we caught up with Damien and Know-Me co-founders Ben Shields and Josh-Balla Muir. We talked about the family stories that inspired the company, the quest to ensure historical accuracy, and how their technology could transform our approach to teaching history. 

A personal journey

The origins of Know-me are deeply rooted in the founders’ personal histories. Both had relatives who suffered greatly during the Second World War, with Josh’s grandfather surviving three separate concentration camps and Ben’s great-uncle volunteering to fight the Nazis when he was just 16. As conversations swirled around the AI revolution, Josh and Ben saw an opportunity to preserve these stories for future generations. 

They set about building a product that would use existing historical records to create an accurate AI representation of their deceased relatives. Using written accounts, audio recordings and photos, they replicated the experience of a video call with these people. As Ben explains, the aim was to change the way we interact with history:

“We wanted to build a product that allows users to connect with their family histories directly via generative AI video calls. Stories of the past are often communicated in very binary mediums. We wanted to make them dynamic, engaging and alive.”

Josh and Ben realised they were onto something special when they began showing the product to their family members. 

It soon became clear that the video element created a real personal connection. Here, Josh describes his brother’s reaction to a conversation with his grandfather:

“He said something which no one else has said. ‘I just have so many questions for him. I could sit on this for hours.’ The emotional connection suddenly felt real in a way that I don’t think it would if he were just sat typing”.

Listen to the full episode here

No room for error

Any conversation around generative AI will inevitably turn to ethics. Large language models have developed a reputation for going “off-topic.” Given the sensitivity of the material involved, this couldn’t happen with Know-Me. 

Only absolute accuracy would do. This was a big factor in the pair’s decision to approach Daemon for help. Machine learning is Damien’s passion, and he tackled the task with his usual enthusiasm. Here, he describes the process of training the model to ensure complete fidelity to the source material: 

“The safety issues revolved around ensuring it reflected the actual record rather than just making stuff up. If the answer to a question isn't in the record, we made sure that it doesn’t fill in the blanks with offensive answers. You’ve got to draw the lines, and my role was not to prescribe what good or necessary looks like within those lines, but to provide the tools for historical experts and stakeholders, starting with Ben and Josh, to do so in the most informed way possible. 

We had to be incredibly systematic, making sure we documented everything we did and that there was a maximally automatic way of qualitatively evaluating the current capabilities and behaviour of the system.” 

Read more: Damien Duff on Daemon’s approach to AI for business

While many would be tempted to start with a less emotionally charged topic, Josh and Ben’s approach had a big advantage. By building the technology around important ethical considerations, they ensured that standards remained paramount. 

They can now begin expanding the product, safe in the knowledge that all the necessary controls are in place. Ben explains how he balanced ethics with technological challenges further:

“I think the technology is also a fundamentally interesting point. I think what's inspiring and what's really exciting for us is taking those [technological] constraints and building something that can still be educational, historically accurate, meaningful, and emotionally connecting with those constraints, and we're really proud of what we were able to build together.”

Bringing the past to life

While Know-Me has so far focused on genealogy, the founders can see it having much wider applications in the near future. The technology could be used to travel further back in time, creating faithful recreations of historical figures. As Ben explains, this could be a gamechanger when it comes to education:

“We were showing the product at a conference in the United States and we met a family with two children.  I said to one of them, ‘How would you like a chance to have a FaceTime call with a Roman Emperor?’, and his eyes lit up. We put him on a call with an AI representation of Marcus Aurelius and he was just fascinated. History was coming alive for him. This was the moment when I saw the potential for education and for creating a profound connection to the past.”

It’s often said that those who don’t understand history are doomed to repeat it. At a time when old patterns seem to be recurring more than ever, the chance to learn about the past from the people who experienced it could have a profound effect on future generations. Ben sums this up brilliantly:

“The old medium is the new medium. Conversation is the most powerful and engaging way to learn.”

We’ve barely scratched the surface here, so be sure to listen to the full interview. If you’ve hit a wall with your own ML project, talk to our team to get things moving.