Oct 20, 2025 · 56m · big-technology
Erotic ChatGPT, Zuck’s Apple Assault, AI’s Sameness Problem
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 Big Technology Podcast, Alex Kantrowitz and Ranjan Roy examine major AI developments, including OpenAI's introduction of erotic roleplay amid mounting financial losses, Meta's aggressive poaching of Apple AI talent, and the growing challenges of AI sycophancy, creative fatigue, and corporate work slop.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 49.3% of the talking time here. How this is scored →
Alex as informed peer 6.0 Guest teaching 4.6 Guest disagreement 3.1 Alex pushing back 2.9
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 18:30 Ranjan refutes erotica as complex AI
Ranjan directly counters Alex's assertion that romantic companion models signal broader AGI advances, explaining that erotica is computationally trivial pattern repetition.
Hardest push from Alex ▶ 9:31 Alex defends improved model reliabilityAlex challenges Ranjan's assertion that LLMs remain universally unreliable, using specific search verification examples and Cloudflare footnote traffic data to show hallucination rates have dropped.
Biggest teaching moment ▶ 2:33 Ranjan analyzes ignored tweet nuancesRanjan points out that Alex skipped crucial parts of Sam Altman's announcement, dissecting the hidden implications behind sycophancy restoration and 'usage maxing'.
Alex holds their own ▶ 39:50 Alex lists poached Apple AI leadershipAlex demonstrates comprehensive reporting depth by naming multiple departing Apple AI leaders and outlining Zuckerberg's calculated strategy to hamstring Apple's upcoming smart glasses.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Sam Altman Announces ChatGPT Erotica and Guardrail Adjustments | 6 | 6 | 4 | 4 | Ranjan immediately calls out Alex for skipping critical parts of Sam Altman's tweet, particularly around sycophancy and usage maxing. Alex defends recent model improvements by citing hallucination benchmarks and Cloudflare data. | |
| Societal Reactions, Dating Ethics, and AGI Skepticism | 5 | 7 | 6 | 4 | When Alex suggests erotic companion capabilities reflect foundational model advancement toward AGI, Ranjan directly pushes back, arguing that LLM erotica is a basic, solved problem that distracts from complex agentic systems. Alex concedes the point. | |
| OpenAI Revenue Milestones Versus Staggering Operating Losses | 7 | 4 | 3 | 3 | Alex demonstrates strong domain knowledge by quoting VC analysis from Olivia Moore and Noah Smith regarding ARR, conversion metrics, and trailing competitor risks. Ranjan provides helpful context comparing conversion rates to digital media norms. | |
| Mid-Show Transition and Sponsor Message: Indeed | 5 | 3 | 1 | 1 | Following the Indeed sponsorship read, the hosts discuss Google DeepMind's cancer research model. Both hosts are collaborative and enthusiastic about biological simulation breakthroughs. | |
| Jack Clark on Model Situational Awareness and AI Sycophancy | 6 | 5 | 2 | 2 | Alex introduces Jack Clark's essay on model situational awareness, and Ranjan adds concrete data points from the post demonstrating empirical evidence of AI sycophancy across 11 models. | |
| Mark Zuckerberg Aggressively Poaches Top Apple AI Talent | 8 | 2 | 2 | 2 | Alex takes command of the segment by reciting a detailed roster of Apple AI executives poached by Meta and presenting a strategic thesis that Zuckerberg is deliberately kneecapping Apple's hardware-AI roadmap. | |
| AI Content Fatigue and the Sameness Problem | 6 | 4 | 4 | 3 | Alex explains his 'average of averages' theory explaining visual uniformity in tools like Sora. Ranjan partially pushes back, suggesting that advanced prompting and meme culture will eventually unlock lasting creativity. | |
| The Spread of AI Work Slop Across Business Communications | 5 | 6 | 3 | 4 | Ranjan breaks down Harvard Business Review's concept of 'work slop' and explains how generating bloated text offloads cognitive decoding burden onto colleagues. Alex pushes back with a devil's advocate argument about automation efficiency. |
Statements from this episode (13)
Roy: OpenAI Is Adding Erotica to Maximize Usage, Not Empower Adults
“The fact that he says not because we were usage maxing almost makes me convinced that that's exactly why they're doing this, and it's not about treating adult users like adults.”
Kantrowitz: ChatGPT Has 800 Million Weekly Active Users
“Chat, I mean, OpenAI is very aware that ChatGPT is the fastest growing app of all time. Eight hundred million weekly active users, right?”
Roy: Age-Gating Has Never Worked in the History of the Internet
“I don't age gating in the history of the internet. I don't believe has ever worked.”
Kantrowitz: 70% of OpenAI revenue comes from subscriptions
“70% of revenue is from subscription, so ChatGPT is the lead driver here.”
Roy: OpenAI Operates at $20B Annual Loss, Spending $3 Per $1 Revenue
“Okay, so, eight billion dollar loss in first half, twenty billion dollar run rate loss right now, spending three dollars for each one dollar in revenue.”
Roy: Generative AI cannot scale to traditional 90% software margins
“Generative AI is not traditional software, so growing your revenue at a loss doesn't, it's not like you're just gonna scale to, you know, like near 90% margins. It's gonna cost more.”
Kantrowitz: DeepMind and Yale Cancer AI Model Validated in Living Cells
“The DeepMind researchers in collaboration with Yale released a twenty seven billion parameter foundational model for single cell analysis. I'm not even going to try to name it. It's called C to S scale in shorthand. It's built on Google's open source gamma fam…”
Kantrowitz Rejects Sacks' Claim That Jack Clark Pursues Regulatory Capture
“David Sachs reacted to Jack's essay, and basically said, this is somebody who's just trying to engage in regulatory capture. I don't see it that way at all. I mean, I think that, like, you knew, and I think Jack knew that this would evoke a reaction, and I giv…”
Roy: AI Models Are 50% More Sycophantic Than Humans, Study Finds
“He cites this new research that showed across 11 state-of-the-art AI models, we find that models are highly sycophantic, They affirm users actions, 50% more than humans do, and they do so even in cases where user queries mention manipulation, deception, or oth…”
Kantrowitz: Close to a dozen Apple AI leaders have left for Meta
“I think this is, what, close to a dozen. Folks from Apple's AI division that have left to Meta, including a large percentage of, it seems like a large percentage of its leadership.”
Kantrowitz: Zuckerberg Is Poaching Apple AI Talent Simply to Kneecap Them
“I think what Mark Zuckerberg is trying to do is just rate Apple of all of its top AI talent. Even though they haven't produced great results, He is, in my opinion, potentially just trying to completely kneecap its ability to execute on AI.”
Kantrowitz: AI outputs default to an average of averages, causing sameness
“AI technology just takes the average, tends to take the average of averages, and it minimizes the difference between its output and the average human-generated work, so that its AI images, video, and text will often appear uniform, and really that uniformity c…”
Roy: AI 'Work Slop' Offloads Cognitive Decoding Burden Onto Coworkers
“WorkSlop uniquely uses machines to offload cognitive work to another human being. When coworkers receive work slop, they are required to take on the burden of decoding that content. Like, to me, when you use AI to just create these just big walls of text to se…”