Fintech Cappuccino

A handful of genuinely interesting concepts surface - the 'Xerox line,' data bias from Wikipedia's 20% female contributors, and the privacy-as-antitrust-basis argument - but these are surrounded by long stretches of social filler, yes-yeah affirmations, and generic 2023 AI talking points. The insight-to-runtime ratio is poor for a 27-minute episode.

“I call it the Xerox line. Um, in the past, everybody below a certain generation, uh, uh, content generation, it was consultants creating and copying uh, slides and it was just repetitive work. And I do think that AI will replace everything below the Xerox line.”

“antitrust regulation is price based. I believe it should be privacy based because my worry would be that, um, uh, data and data and privacy will be something for the rich”

The 'Xerox line' framing and the privacy-based antitrust angle offer modest freshness, but the bulk of the conversation recycles standard 2023 AI discourse - job replacement fears, data bias, regulation gaps - and the electricity analogy for AI is one of the most overused in the space.

“I call it the Xerox line”

“I compare where we are now to when humans got to experiment with electricity and how do we implement electricity within every single aspect of our life. And I compare that to AI.”

Schouten is a genuine fintech operator with a credible résumé - cloud-native core banking at scale, multiple co-authored books, and she actually ran the ChatGPT book experiment she describes. However, she explicitly disclaims deep AI expertise and functions more as an informed practitioner-experimenter than a domain specialist.

“we were on our way to OpenAI and by that time, ChatGPT didn't even launch yet. We couldn't find the entrance.”

“I'm not the biggest AI specialist I know use cases.”

There are several concrete anchors - $500K starting salaries for ML experts, 40 companies visited, 20% female Wikipedia contributors, the Pera company's 33% diversity lift, Common Crawl as a named GPT training source - but most claims float at a level of assertion without sourcing, and the Pera statistic is dropped without any methodological context.

“if you've studied data science and are, are an expert in machine learning, you get a starting salary of $500,000”

“in terms of diversity, it increases by 33% because it takes away complete bias, free hiring”

The hosts occasionally attempt real follow-ups ('And who should fix it?', pushing on GPT-4) but largely let assertions pass unchallenged; the conversation is riddled with affirmative 'yeah yeah' filler and the live trade-show format clearly constrains depth. Questions are structured but rarely incisive enough to draw out detail the guest hasn't already volunteered.

“And did you also, um, use, uh, version, uh, four already? And, uh, is that probably way better? Right.”

“And who should fix it? By the way?”