#37: Generative AI and financial operations

How can Generative AI be applied to financial operations and accounting? What is the link connecting language and finance and what does this mean for using language models? Kasper Junge , a skilled and seasoned AI expert, kicks of a brilliant episode by going back to Mesopotamia with me, an educated historian, in a surprising role shift. But don't be fooled - this episode is very much about how to apply and think generative AI into accounting and financial operations. And while we explore this, we uncover a lot of truths about how we work, how we structure our enterprises, and what novelty AI brings to the toolbox.

Read more about Kasper at the bottom of this newsletter.

As always, I am very happy that we once again managed to make an episode that explores more than it explains 💡

Listen to the episode here:

Apple: The Only Constant on Apple Podcasts

Spotify: The Only Constant | Podcast on Spotify

Spreaker: The Only Constant (spreaker.com)

What made this episode special to me?:

As I mentioned above, Kasper actually just starts off by talking ancient history. Well - how cool is that? I am not a classical or ancient historian, but still, this simply has to take the prize for most deep rooted, diachrone perspective to open any podcast episode that focuses primarily on AI.

Kasper is not only a great AI engineer, he is also a founding member of the Danish Data Science Community and host his own tech-podcast (in Danish), Verbos. He is a community builder, literally, and his insights on the podcast echo of so many conversations past. I love this. Also because it unveils a subtle truth: that only trough human interaction, evaluation, pondering does the value of AI stand out. And I hope and trust that this is also what the current episode will bring to the listeners.

My key post-recording deliberations are:

• Generative AI as a witness. I never thought about it this way, but the ability for Generative AI to quickly add relevant, contextual metadata to any transaction actually means it becomes a witness. Evidence, after all, is what finance in essence is all about?...
• Can AI assist those that do not know what good looks like? Kasper raises this as a clear dilemma in his work - how to enable business owners to do their accounting assisted by an AI, when they do not know if the AI is even advising them correctly?
• And this, seriously: Kasper, who had spent years working in NLP, was actually absolutely surprised at how good GPT 3.5 actually was. That it worked. It had been kind of a funny idea from OpenAI that they thought they could achieve this simply by scaling the models. But then it worked. I absolutely love how Kasper admittet to this flabbergasted position, even as an AI engineer. Because, honestly - it still feels like magic when you think about it...

And so much more. This episode is a classic, and I hope people will enjoy listening to it as much as we enjoyed recording it.

About Kasper Junge :

Kasper Junge is the Lead AI Engineer at Dinero, where he leads Dinero's work with AI and Natural Language Processing (NLP), including generative language models. He is the co-founder and chairman of the Danish Data Science Community and is an enthusiastic advocate for AI technology, engaging both engineers and general audiences. He is known for his contributions to the AI community in Denmark, including co-hosting the "Verbos Podcast," for which he received the Data and AI Influencer DAIR Award in 2023.