Nov 15, 2024 · 56m · big-technology
Is Generative AI Plateauing?, Booming Bluesky, Apple’s Smart Glasses Play
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
Alex Kantrowitz and Ranjan Roy analyze reports of large language model performance plateaus, the rise of reasoning-focused AI architectures, and practical enterprise market disruptions. They also evaluate Bluesky's post-election growth against Twitter's network dominance and explore Apple's exploratory moves into smart glasses hardware.
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 50.6% of the talking time here. How this is scored →
Alex as informed peer 5.6 Guest teaching 4.2 Guest disagreement 2.2 Alex pushing back 2.6
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:53 Ranjan rejects the premise that an R&D slowdown damages AI
Ranjan forcefully dismisses Alex's concern over diminishing returns in model training, arguing that over-raising capital for pure R&D actively harms the industry compared to building usable software.
Hardest push from Alex ▶ 8:26 Alex challenges guest for ignoring the business stakes of plateauing modelsAlex refuses Ranjan's relaxed framing on model stagnation, aggressively pressing him on the massive VC capital, Anthropic rounds, and nuclear power contracts predicated on continued scaling.
Biggest teaching moment ▶ 35:02 Ranjan breaks down the macroeconomic shift to outcome-based pricingRanjan explains to Alex how generative AI forces a structural transition away from seat-based and billable-hour pricing models across advertising, law, healthcare, and enterprise software.
Alex holds their own ▶ 5:14 Alex marshals industry heavyweights to counter guest's optimismAlex demonstrates subject-matter authority by citing Reuters reporting on Ilya Sutskever, Andreessen Horowitz commentary, and Information reporting to substantiate the reality of LLM pre-training plateaus.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Mike Tyson vs. Jake Paul Fight Banter | 7 | 3 | 5 | 7 | Alex aggressively challenges Ranjan's optimism regarding the LLM scaling plateau, citing Ilya Sutskever's Reuters comments, Marc Andreessen, and OpenAI's massive $6B funding round. Ranjan firmly resists Alex's framing, arguing that slowing foundational model development is beneficial for practical productization. | |
| The Shift Toward Reasoning Models and Inference Compute | 7 | 5 | 2 | 2 | Alex explains the transition to test-time reasoning compute and cites a TEDAI poker statistic demonstrating efficiency gains over pre-training scale. Ranjan expands on how targeted synthetic data enables startups like Writer to build focused models at fractionally lower costs. | |
| Deconstructing AGI Definitions and Industry Expectations | 6 | 4 | 3 | 3 | The pair debate definitions of AGI, with Alex citing Yann LeCun's human-level generalist benchmark while Ranjan questions whether AGI is simply a nebulous marketing and contractual term that distracts from narrower utility. | |
| Gemini's Benchmark Surge and Real-World AI Workflows | 5 | 4 | 1 | 1 | Alex highlights Google's experimental Gemini model topping the Chatbot Arena leaderboard, while both agree that superior benchmarks do not immediately overcome interface and workflow preferences. | |
| Practical AI Impact: Writer, Chegg, and Agency Billing Models | 6 | 6 | 2 | 2 | Alex discusses Chegg's 99% valuation collapse and shifts in ad agency billing models due to AI efficiency. Ranjan explains the broader transition across SaaS, legal, and healthcare sectors toward outcome-based pricing rather than seat or hourly billing. | |
| Spotify's Recommendation Pitfalls and AI-Generated Music | 5 | 5 | 1 | 1 | Ranjan describes how Spotify's recommendation algorithms break under household edge cases like kids' music, then explains his positive tests with Spotify prompt-based playlists and Suno-generated folk tracks. | |
| Podcast Format Updates, Listener Feedback, and Starshield | 3 | 2 | 0 | 0 | Alex conducts housekeeping, addresses listener feedback, and shares a listener correction regarding the US Department of Defense's Starshield satellite initiative with SpaceX. | |
| Bluesky's Growth Momentum vs. Threads and Twitter Network Effects | 6 | 4 | 4 | 6 | Ranjan argues Bluesky has achieved genuine mainstream momentum post-election, while Alex pushes back citing Max Read's analysis that it remains a specialized echo chamber unable to dislodge Twitter's core network effects. | |
| Apple Smart Glasses Study and Mark Zuckerberg's Z-Pain Track | 5 | 5 | 2 | 1 | Ranjan evaluates Apple's exploratory Project Atlas smart glasses against his hands-on experience with Snap Spectacles, and the hosts banter over Mark Zuckerberg's acoustic Z-Pain music release. |
Statements from this episode (19)
Roy: Netflix is genius for streaming Tyson-Paul fight with polarizing figures
“I think Netflix is genius in promoting their live programming by just bringing out two people that no one wants to see win”
Kantrowitz: OpenAI's $6 billion funding is the largest VC round ever
“Opening
, I just raised the largest VC round in history, six billion dollars.”
Roy: Over-funding foundational AI models instead of business applications is a mistake
“Over-raising for the R&D side of things rather than the actual, like, operationalization and building out businesses on top of the existing technology. I mean, again, we've debated this plenty. I think that is a huge mistake in that it actually, you know, pote…”
Roy: AI video generation models like OpenAI's Sora are nowhere near interesting
“Video generation is the one area that I do think we are severely, we're not even close to anything interesting, and we've been promised things that are interesting, i.e. Sora but we're very, very far away.”
Kantrowitz: AI research is pivoting entirely toward reasoning models over sheer scaling
“So I think what we're about to see is a pivot in the AI research field or yes, they might be applying practically some of the models that exist today. But it seems to me like everybody is going to go completely in on this reasoning format. And that is going to…”
Roy: Smaller, domain-specific AI models will outperform massive general foundation models
“And I think in the coming months and years, we're going to start to see some awkward headlines around size matters because smaller will be better in terms of actual models being used.”
Roy: AGI's lack of a clear definition distracts from practical technological progress
“There's not one like clear accepted definition or one clear vision that's communicated by the biggest people in the field, the Sam Altman's and everything else. So it remains this murky kind of like dystopian robots taking over who knows what it is, will be a …”
Kantrowitz: Superintelligence will not arrive through traditional scaling of large language models
“I think that that has led a lot of the investment and a lot of the hype around this that will eventually get there. But it just doesn't seem like it's gonna be through the traditional scaling of LLMs.”
Kantrowitz: Google's experimental Gemini 1114 model currently tops the Chatbot Arena leaderboard
“They have a new experimental Gemini model. It's called Gemini exp one, four. And Do you know about, ah, Chatbot Arena where they test which is the best LLMs? It's currently sitting at the top of Chatbot Arena and kicking the butt of ChatGPT for O, O-one Previe…”
Roy: Perplexity AI's clear format makes it the biggest traditional search competitor
“Perplexity is when I'm watching a movie, a sports game of any sort, like, it just is so good in just giving you quick information in a really nice format with Additional links to keep exploring in questions that I think for that, like, and which makes me think…”
Kantrowitz: Chegg stock is down 99 percent, erasing $14.5 billion in value
“Chegg stock is down 99% from early twenty-twenty-one, erasing some 14.5 billion of market value.”
Roy: AI automation will force SaaS to abandon seat-based pricing for outcomes
“I think for so many of these industries, the way the entire pricing structure changes, it's going to change. And I even think in SAS, that's going to be the case. And there's been a lot of talk around this with even Salesforce's AI agents or in many others is …”
Kantrowitz: Bluesky user base soared to 15 million following the US election
“So blue sky is now up to fifteen million users, and it is it's really soaring in the wake of the election.”
Kantrowitz: Meta's Threads added 15 million new users in November 2024 alone
“Threads has added fifteen million users since the start of November”
Roy: Bluesky's latest growth surge has staying power and threatens Twitter directly
“I think it does have staying power this time.”
Kantrowitz: Meta's Threads feels empty and lacks engagement despite massive user growth
“One thing we can tell you, say for sure is it doesn't look like threads is working. I mean, threads added fifteen million people since the start of the month. And Michael Learmonth, who I work with as an editor pointed out to me, he's like, does it feel like t…”
Kantrowitz: Twitter's massive network effects make it incredibly difficult to actually replace
“Twitter is going to be the one with staying power. It's the network effects. It's very, very, it might be the most difficult to replace social network.”