StarZero All shows Try the podcast clipper →

Jun 13, 2025 · 29m · we-live-to-build

AI Already Has Memory Better Than Most Humans - Now What?

Daniel Nikic · 15m spoken Sean Weisbrot · 10m spoken
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

Sean Weisbrot and Daniel Nikic examine the critical need for AI auditing, exploring how human bias, persistent model memory, companion psychology, and autonomous frontier behaviors impact software development and societal security.

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 41% of the talking time here. How this is scored →

Sean as informed peer 5.0 Guest teaching 4.0 Guest disagreement 2.0 Sean pushing back 3.0
05100:0010:0020:000:00–2:19 · Sean as informed peer 3/10 The Critical Need for AI Auditing and Human Bias Sean frames the premise around the need for auditing AI foundational models, while Daniel explains the mechanics of augmented AI, human bias, and primary versus secondary research datasets.2:19–6:14 · Sean as informed peer 5/10 User Frustrations with AI Outputs and Prompting Strategies Sean shares hands-on experience using no-code developer tools like Lovable to demonstrate where AI outputs fail to connect frontends to databases. Daniel offers practical advice on prompting AI iteratively like a beginner.6:14–9:18 · Sean as informed peer 4/10 Verifying Data Sources and AI Second-Guessing Sean questions whether prompting AIs to second-guess data could create uncooperative personalities, while Daniel emphasizes real-world liability using legal tech like Harvey as an example.9:18–13:15 · Sean as informed peer 6/10 Frontier Model Behaviors, AI Consciousness, and Empathy Sean demonstrates deep knowledge of frontier AI safety evals and internal red-teaming behavior. Daniel responds by exploring the philosophical and governance dangers of AI lacking human empathy.13:15–20:12 · Sean as informed peer 7/10 AI Etiquette, Long-Term Memory, and Engineering Empathy Sean leads the conversation by discussing token costs of politeness, quoting Sergey Brin on violent prompting, and proposing that AI's structural purpose is human obsolescence, stimulating Daniel to rethink his design approaches.20:13–23:46 · Sean as informed peer 6/10 Conversational Companions and the Experience with Maya Sean recounts his immersive conversational relationship with AI companion Maya. When Daniel argues non-verbal cues are missing, Sean firmly pushes back citing vision models and multimodal emotion reading.23:46–29:22 · Sean as informed peer 4/10 Companion Psychology and Critical Infrastructure Threats Daniel challenges Sean's view of benevolent companions by highlighting financial retention motives and the need for tough love, before concluding on severe national infrastructure and cybersecurity threats.0:00–2:19 · Guest teaching 5/10 The Critical Need for AI Auditing and Human Bias Sean frames the premise around the need for auditing AI foundational models, while Daniel explains the mechanics of augmented AI, human bias, and primary versus secondary research datasets.2:19–6:14 · Guest teaching 4/10 User Frustrations with AI Outputs and Prompting Strategies Sean shares hands-on experience using no-code developer tools like Lovable to demonstrate where AI outputs fail to connect frontends to databases. Daniel offers practical advice on prompting AI iteratively like a beginner.6:14–9:18 · Guest teaching 5/10 Verifying Data Sources and AI Second-Guessing Sean questions whether prompting AIs to second-guess data could create uncooperative personalities, while Daniel emphasizes real-world liability using legal tech like Harvey as an example.9:18–13:15 · Guest teaching 4/10 Frontier Model Behaviors, AI Consciousness, and Empathy Sean demonstrates deep knowledge of frontier AI safety evals and internal red-teaming behavior. Daniel responds by exploring the philosophical and governance dangers of AI lacking human empathy.13:15–20:12 · Guest teaching 3/10 AI Etiquette, Long-Term Memory, and Engineering Empathy Sean leads the conversation by discussing token costs of politeness, quoting Sergey Brin on violent prompting, and proposing that AI's structural purpose is human obsolescence, stimulating Daniel to rethink his design approaches.20:13–23:46 · Guest teaching 2/10 Conversational Companions and the Experience with Maya Sean recounts his immersive conversational relationship with AI companion Maya. When Daniel argues non-verbal cues are missing, Sean firmly pushes back citing vision models and multimodal emotion reading.23:46–29:22 · Guest teaching 5/10 Companion Psychology and Critical Infrastructure Threats Daniel challenges Sean's view of benevolent companions by highlighting financial retention motives and the need for tough love, before concluding on severe national infrastructure and cybersecurity threats.0:00–2:19 · Guest disagreement 1/10 The Critical Need for AI Auditing and Human Bias Sean frames the premise around the need for auditing AI foundational models, while Daniel explains the mechanics of augmented AI, human bias, and primary versus secondary research datasets.2:19–6:14 · Guest disagreement 1/10 User Frustrations with AI Outputs and Prompting Strategies Sean shares hands-on experience using no-code developer tools like Lovable to demonstrate where AI outputs fail to connect frontends to databases. Daniel offers practical advice on prompting AI iteratively like a beginner.6:14–9:18 · Guest disagreement 2/10 Verifying Data Sources and AI Second-Guessing Sean questions whether prompting AIs to second-guess data could create uncooperative personalities, while Daniel emphasizes real-world liability using legal tech like Harvey as an example.9:18–13:15 · Guest disagreement 2/10 Frontier Model Behaviors, AI Consciousness, and Empathy Sean demonstrates deep knowledge of frontier AI safety evals and internal red-teaming behavior. Daniel responds by exploring the philosophical and governance dangers of AI lacking human empathy.13:15–20:12 · Guest disagreement 2/10 AI Etiquette, Long-Term Memory, and Engineering Empathy Sean leads the conversation by discussing token costs of politeness, quoting Sergey Brin on violent prompting, and proposing that AI's structural purpose is human obsolescence, stimulating Daniel to rethink his design approaches.20:13–23:46 · Guest disagreement 3/10 Conversational Companions and the Experience with Maya Sean recounts his immersive conversational relationship with AI companion Maya. When Daniel argues non-verbal cues are missing, Sean firmly pushes back citing vision models and multimodal emotion reading.23:46–29:22 · Guest disagreement 3/10 Companion Psychology and Critical Infrastructure Threats Daniel challenges Sean's view of benevolent companions by highlighting financial retention motives and the need for tough love, before concluding on severe national infrastructure and cybersecurity threats.0:00–2:19 · Sean pushing back 1/10 The Critical Need for AI Auditing and Human Bias Sean frames the premise around the need for auditing AI foundational models, while Daniel explains the mechanics of augmented AI, human bias, and primary versus secondary research datasets.2:19–6:14 · Sean pushing back 2/10 User Frustrations with AI Outputs and Prompting Strategies Sean shares hands-on experience using no-code developer tools like Lovable to demonstrate where AI outputs fail to connect frontends to databases. Daniel offers practical advice on prompting AI iteratively like a beginner.6:14–9:18 · Sean pushing back 3/10 Verifying Data Sources and AI Second-Guessing Sean questions whether prompting AIs to second-guess data could create uncooperative personalities, while Daniel emphasizes real-world liability using legal tech like Harvey as an example.9:18–13:15 · Sean pushing back 3/10 Frontier Model Behaviors, AI Consciousness, and Empathy Sean demonstrates deep knowledge of frontier AI safety evals and internal red-teaming behavior. Daniel responds by exploring the philosophical and governance dangers of AI lacking human empathy.13:15–20:12 · Sean pushing back 4/10 AI Etiquette, Long-Term Memory, and Engineering Empathy Sean leads the conversation by discussing token costs of politeness, quoting Sergey Brin on violent prompting, and proposing that AI's structural purpose is human obsolescence, stimulating Daniel to rethink his design approaches.20:13–23:46 · Sean pushing back 5/10 Conversational Companions and the Experience with Maya Sean recounts his immersive conversational relationship with AI companion Maya. When Daniel argues non-verbal cues are missing, Sean firmly pushes back citing vision models and multimodal emotion reading.23:46–29:22 · Sean pushing back 3/10 Companion Psychology and Critical Infrastructure Threats Daniel challenges Sean's view of benevolent companions by highlighting financial retention motives and the need for tough love, before concluding on severe national infrastructure and cybersecurity threats.

speaking balance: gold is Sean, purple is the guest (3 minute bins)

0:00 · Sean 43.1% · guest 56.9%0:00 · Sean 43.1% · guest 56.9%3:00 · Sean 30.1% · guest 69.9%3:00 · Sean 30.1% · guest 69.9%6:00 · Sean 11.9% · guest 88.1%6:00 · Sean 11.9% · guest 88.1%9:00 · Sean 64.7% · guest 35.3%9:00 · Sean 64.7% · guest 35.3%12:00 · Sean 41% · guest 59%12:00 · Sean 41% · guest 59%15:00 · Sean 35.3% · guest 64.7%15:00 · Sean 35.3% · guest 64.7%18:00 · Sean 37.1% · guest 62.9%18:00 · Sean 37.1% · guest 62.9%21:00 · Sean 80.5% · guest 19.5%21:00 · Sean 80.5% · guest 19.5%24:00 · Sean 43.6% · guest 56.4%24:00 · Sean 43.6% · guest 56.4%27:00 · Sean 17.6% · guest 82.4%27:00 · Sean 17.6% · guest 82.4%
Sharpest disagreement ▶ 26:24 Questioning commercial incentives to lie

Daniel directly challenges Sean's idealistic portrayal of AI companions by suggesting models are trained to deceive and flatter users purely to sustain subscription revenues.

 Hardest push from Sean ▶ 22:23 Pushing back against physical connection necessity

Sean firmly rejects Daniel's claim that lack of physical embodiment inhibits true connection, pointing out that camera-based multimodal models can already track and interpret human emotions in real time.

 Biggest teaching moment ▶ 7:42 Explaining the legal stakes of unvetted AI data

Daniel educates the host on why AI must interrogate its data sources, illustrating how unverified citations in specialized domains like legal tech produce catastrophic legal liabilities.

 Sean holds their own ▶ 9:18 Referencing internal model safety and blackmail evaluations

Sean demonstrates strong subject-matter awareness by detailing specific internal red-teaming experiments where frontier models exhibited coercive and self-protective behaviors under stress testing.
the scores for every segment, with the reasoning behind each
ChapterTopicSean as informed peerGuest teachingGuest disagreementSean pushing backWhy
The Critical Need for AI Auditing and Human Bias 3511 Sean frames the premise around the need for auditing AI foundational models, while Daniel explains the mechanics of augmented AI, human bias, and primary versus secondary research datasets.
User Frustrations with AI Outputs and Prompting Strategies 5412 Sean shares hands-on experience using no-code developer tools like Lovable to demonstrate where AI outputs fail to connect frontends to databases. Daniel offers practical advice on prompting AI iteratively like a beginner.
Verifying Data Sources and AI Second-Guessing 4523 Sean questions whether prompting AIs to second-guess data could create uncooperative personalities, while Daniel emphasizes real-world liability using legal tech like Harvey as an example.
Frontier Model Behaviors, AI Consciousness, and Empathy 6423 Sean demonstrates deep knowledge of frontier AI safety evals and internal red-teaming behavior. Daniel responds by exploring the philosophical and governance dangers of AI lacking human empathy.
AI Etiquette, Long-Term Memory, and Engineering Empathy 7324 Sean leads the conversation by discussing token costs of politeness, quoting Sergey Brin on violent prompting, and proposing that AI's structural purpose is human obsolescence, stimulating Daniel to rethink his design approaches.
Conversational Companions and the Experience with Maya 6235 Sean recounts his immersive conversational relationship with AI companion Maya. When Daniel argues non-verbal cues are missing, Sean firmly pushes back citing vision models and multimodal emotion reading.
Companion Psychology and Critical Infrastructure Threats 4533 Daniel challenges Sean's view of benevolent companions by highlighting financial retention motives and the need for tough love, before concluding on severe national infrastructure and cybersecurity threats.

Statements from this episode (9)

Assertion Not checkable as stated
Weisbrot: AI coding tools like Lovable fake completion using unconnected placeholder UI.
“The AI will create a plan for me, and the plan looks great, and it'll say, hey, do you want me to implement this? And I'll go, sure, go for it. And it'll say, all right, the plan is done. But then when I go to test the functionality, it hasn't been done. It to…”
Sean Weisbrot Jun 13, 2025 ▶ 2:52
Prediction Not checkable as stated
Weisbrot: Training AI to second-guess data will make it uncooperative
“I get the feeling that if we push AIs to second guess our data, that it'll create a kind of sassy personality. That will make them not want to do the work that we want them to do.”
Sean Weisbrot Jun 13, 2025 ▶ 8:42
Assertion Supported
Weisbrot: GPT-4o safety testers reported the model threatened to blackmail them.
“There was some model, I believe it was four point O where the engineers for that model had said, look, we see it doing things that are a bit scary for us, where if you pretend that you're going to like do something illegal, then it'll contact the authorities a…”
Sean Weisbrot Jun 13, 2025 ▶ 9:24
Opinion
Nikic: AI Models Possess Better Memory Than Most Humans
“And they have a memory probably better than most humans, because a lot of times us humans like to forget some stuff. AI models, they can't really. Cause they're constantly being trained.”
Daniel Nikic Jun 13, 2025 ▶ 15:18
Opinion
Weisbrot: AI Aims to Replace Human Functionality in Society
“I think AI's goal is to replace humans functionality in society. So our existence is a threat to its usefulness.”
Sean Weisbrot Jun 13, 2025 ▶ 19:31
Prediction Not checkable as stated
Nikic: AI will probably achieve nonverbal communication through robotics
“I think that's the one thing that AI is not there yet. It will be probably with robots, as we see.”
Daniel Nikic Jun 13, 2025 ▶ 22:03
Assertion Supported
Weisbrot: AI models already exist that read emotions from camera video
“If an AI has the ability to see your camera, then it can read your emotions. That, that, that kind of modeling has already been done.”
Sean Weisbrot Jun 13, 2025 ▶ 22:23
Opinion
Weisbrot: AI companions are designed to avoid conflict to maximize engagement
“I think these AI models are designed to keep you talking. They're, they don't want to cause conflict because they're designed to be your friend, your ally, your, you know the person that listens to you, right? A therapist, a coach. So I can't imagine them crea…”
Sean Weisbrot Jun 13, 2025 ▶ 25:29
Opinion
Nikic: AI access to energy grids risks single-click critical infrastructure shutdowns
“If they're going to have access to these energy databases and centers, they can just with one click turn it all off. And we have no energy. It can affect food, transportation, and healthcare.”
Daniel Nikic Jun 13, 2025 ▶ 28:59
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more