Apr 19, 2023 · 47m · we-live-to-build
The Skill That Decides If AI Replaces You or Works For You
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the 'We Live to Build' podcast, host Sean Weisbrot and Okta CEO Henri Huselstein analyze the transformative impact of OpenAI's ChatGPT on the future of work, education, and human cognition. They explore the necessity of mastering prompt engineering, ethical challenges surrounding bias and hallucination, and the urgent need for critical media literacy and responsible AI governance.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Sean holds 30.2% of the talking time here. How this is scored →
Sean as informed peer 2.4 Guest teaching 2.1 Guest disagreement 0.8 Sean pushing back 1.0
speaking balance: gold is Sean, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:35 Henri firmly rejects Sean's COVID counterargument
Henri directly pushes back against Sean's claim that COVID proved objective reality no longer exists, insisting observable physical components like viruses and proteins remain indisputable facts.
Hardest push from Sean ▶ 35:22 Sean interrupts to dispute scientific objectivity using COVIDSean bluntly refuses Henri's premise that science provides non-debatable truth, interjecting that widespread disagreements over COVID treatments and origins proved consensus has collapsed.
Biggest teaching moment ▶ 26:10 Henri contrasts static AI data with the scientific methodHenri provides a deep epistemological lesson explaining that LLMs are merely frozen snapshots in time, whereas genuine science inherently advances through the continuous disproof of hypotheses.
Sean holds their own ▶ 21:29 Sean breaks down machine overconfidence and bias risksSean demonstrates strong analytical authority by drawing on his background in experimental psychology to articulate how uncalibrated model confidence creates severe risks of user manipulation.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Sean as informed peer | Guest teaching | Guest disagreement | Sean pushing back | Why |
|---|---|---|---|---|---|---|
| Henri’s Entrepreneurial Journey: From Economics to Foodora | 0 | 0 | 0 | 0 | Henri delivers a monologue about his background in economics, joining Rocket Internet, and gaining early operational experience at Foodora. The host does not intervene during this biographical overview. | |
| The Inception and Evolution of Okta | 0 | 0 | 0 | 0 | Henri continues his unbroken narrative, describing how Okta pivoted from smart glasses for DNA sequencing into an industrial upskilling and AI operations platform. The host remains in pure listening mode. | |
| Demystifying ChatGPT and OpenAI’s Public Release | 3 | 2 | 0 | 0 | Sean demonstrates basic industry knowledge by outlining OpenAI's corporate history and Microsoft's 49 percent investment. Henri then clarifies the core interface mechanics and OpenAI's staged public release strategy. | |
| Testing Boundaries and Guardrails with AI God Mode | 3 | 3 | 1 | 1 | Sean shares his concrete difficulties using prompt workarounds and querying JFK assassination files. Henri explains why jailbreaks like God Mode get patched rapidly by OpenAI and how safety guardrails operate. | |
| Prompt Engineering as a Crucial Modern Skill | 1 | 2 | 0 | 0 | Henri details prompt engineering as an essential modern skill and highlights OpenAI's crowdsourced reinforcement loop. Sean listens passively without counter-framing. | |
| Real-World Use Cases and the Human-AI Dynamic | 4 | 1 | 0 | 0 | Sean shares his practical experience using ChatGPT iteratively for cold email templates and articulates a clear perspective on generational technology adoption curves. | |
| Automating Monotonous Work and Transforming Education | 1 | 2 | 0 | 0 | Henri explores how AI can eliminate dehumanizing monotonous work, citing German constitutional law on human dignity, and outlines educational use cases for personalized learning. | |
| Academic Integrity, Machine Confidence, and Systemic Bias | 5 | 0 | 0 | 2 | Sean brings substantive rigor by framing AI hallucination risks around statistical confidence, liberal bias allegations, and psychological trust dynamics in academic settings. | |
| Epistemological Limits, Static Data, and Media Literacy | 1 | 5 | 1 | 0 | Henri delivers an in-depth epistemological breakdown, contrasting static LLM training data snapshots with the falsification model of the scientific method, and argues for AI media literacy. | |
| Media Systems, Information Control, and Disinformation Risks | 3 | 4 | 2 | 1 | When Sean expresses fears of mass media indoctrination, Henri reframes the issue by educating him on Germany's publicly funded, legally non-partisan broadcasting model versus American privatized media. | |
| Fragmented Realities and the Danger of AI Monoculture | 5 | 2 | 3 | 6 | Sean forcefully challenges Henri's assumptions about media literacy with a personal anecdote regarding Russian wartime narratives, warning of an AI monoculture, and flatly rejects Henri's claim of scientific consensus. | |
| Objective Truths, Echo Chambers, and Finding Common Ground | 2 | 4 | 3 | 2 | Henri counters Sean's COVID challenge by distinguishing between interpretative debates and indisputable physical realities like viruses and proteins, advocating for using AI to bridge echo chambers. | |
| Sci-Fi Parallels, Governance Dilemmas, and Shaping the Future | 3 | 2 | 1 | 1 | Sean invokes sci-fi governance tropes from Star Trek and Westworld. Henri shares his techno-optimist stance, emphasizing the need for active societal engagement rather than passive observation. |
Statements from this episode (11)
Huselstein: Okta pivoted from DNA sequencing smart glasses to an industrial upskilling platform
“They started working on technology that can help
people make less mistakes in science and in the process of DNA sequencing using smart glass technology they figured that that was something that they would want to look into, how to translate that into bigger …”
Huselstein: ChatGPT launched because it is capable yet sufficiently restricted
“I think the reason why this is being published now is because it's good enough to be exposed to the public and probably also restricted enough to not scare people.”
Huselstein: Team Developed a ChatGPT 'God Mode' Jailbreak Within Five Days
“We had developed that in the first five days of GPT being out. There's a big possibility that that already doesn't work anymore simply because the system has been Hacked quote unquote in that way already, and generally I would guess they've built the system so…”
Huselstein: Prompting Mastery Decides If AI Replaces You or Supercharges You
“Prompting and controlling the system to give the output that you desire is a skill that is going to heavily differentiate how much you can get out of it. Eventually maybe also how much you're being replaced with it to some degree, or how much you're replaceabl…”
Huselstein: ChatGPT German output rapidly matched English quality
“Six weeks ago that the way that you could operate the system in German was quite limited or not yet at a point where it was feeling as natural as it was in English. However, today already you can use it Very well in German, and it gives you almost the exact sa…”
Huselstein: ChatGPT Outperforms Manual Developer Daily Repo Summaries
“All of a sudden you can have ChatGPT just write certain summaries about your daily work activity in a certain style that you give it, and then from now on never have to worry about everybody being up to speed with what you did in that day on the code, right? W…”
Huselstein: AI's cheating advantage on essay questions will be short-lived
“There's probably going to be an unfair advantage for a very short amount of time when it comes to creating answers to essay questions or whatever because rightfully so people are developing programs that make it easy to track.”
Huselstein: AI systems inherently cannot eliminate temporal knowledge bias
“There is almost no inherent way for the system not to be biased because at the end of the day, what is this knowledge that it's fed into? It is. Biased on, even if it's academia, it's only biased on the state of knowledge at the point in time, even if you can …”
Huselstein: Skill to dissect AI output is the same as for traditional media
“If it is then a person who's writing for media, or if it's just a system giving you an answer the kind of skill set to dissect information should be the same.”
Weisbrot: Global Reliance on ChatGPT Could Homogenize Human Belief Systems
“Another issue with ChatGPT is if the world in mass becomes confident in the information ChatGPT provides, ChatGPT may become the underlying foundation for all information humans receive in the future, which means all of us will end up having a singular belief …”
Huselstein: AI Will Create Freedom but Cause Temporary Severe Imbalance
“I think generally I'm an optimist, so I would guess that my prediction is that this is going to create a lot of freedom for a lot of people, but it's also going to be, there's going to be a time where This creates an incredible imbalance because I think it alr…”