Podcast Summary: AI in 2025 – The Great Normalisation, with Nathan Benaich
Podcast: Azeem Azhar's Exponential View
Host: Azeem Azhar
Guest: Nathan Benaich (Founding General Partner, Air Street Capital)
Date: December 26, 2024
Overview
In this engaging, forward-looking conversation, Azeem Azhar and Nathan Benaich dissect the future trajectory of artificial intelligence, reflecting on surprises and missed predictions from 2024 and looking ahead to normalization and impact in 2025. They cover everything from business models and the power dynamics among AI companies to consumer and enterprise adoption, breakthroughs in biotech, the future of robotics, and bold forecasts for the coming year. The dialogue is grounded in tangible examples, candid self-evaluation, and a keen sense of both optimism and realism about the disruptive power of AI.
Key Discussion Points & Insights
1. Reflecting on AI Predictions: Hits, Misses, and Uncertainties
- Self-grading of predictions: Both Azeem and Nathan admit that their predictions for the AI industry were largely accurate but perhaps too conservative, leading to an internal debate about ambition and insider knowledge.
“We both graded ourselves pretty well, so almost too well. So are we being too unambitious with our predictions of the state of the AI world or is it just more predictable than we let on?” — Azeem [01:16]
- Semiconductor sector stumbles: Nathan candidly discusses his misplaced optimism regarding competition with Nvidia in AI semiconductors; investments in startups have lagged dramatically behind Nvidia's explosive growth.
“I was originally excited about this as a prospect to compete with Nvidia, but it just looks like it’s a losing battle over and over again.” — Nathan [01:46]
2. 2024’s AI Boom: Adoption, Revenue, and Business Models
- Unexpected scale of consumer AI:
- ChatGPT’s ascent to over 305 million weekly users and the enormity of enterprise spending in generative AI were surprises—even to industry insiders.
- “If you’d asked me at the start of this year, 8th busiest website in the world…” — Azeem [03:35]
- Revenue realties clash with earlier skepticism: OpenAI and Anthropic’s rapid revenue growth has outpaced expectations.
“I don’t know anybody who predicts they can go from zero to billions like that quickly… but like, they’ve done it.” — Nathan [04:13]
- Open source vs. proprietary platforms:
- Idealism around open-source models has yet to fully materialize in the market—while enterprise clients with strong privacy needs may care, most consumers stick to polished, reliable services.
- “On the consumer side, it seems to really be like ChatGPT equals AI and that’s just it.” — Nathan [05:12]
3. Market Structure and Product Differentiation
- Enterprise vs. consumer decisions: For enterprises, factors like support, reliability, and ecosystem matter; for consumers, it’s about default, frictionless solutions.
“It feels like it’s much deeper than ‘you don’t get fired for buying IBM.’ It’s deeper… account manager, roadmap, developer community…” — Azeem [06:28]
- Emergence of ‘neutral’ cloud platforms: Companies like Databricks and Snowflake may serve as neutral ground for open source AI models, though incentives in the cloud ecosystem remain misaligned [07:19].
- Notable user workflow shifts: Experimenting with new releases—such as OpenAI’s 01 Pro and Google’s Gemini Deep Research—has yielded major productivity gains, altering daily professional routines [07:46–08:58].
4. Differentiation Among Top AI Model Providers
- Distinct capabilities and experiences:
- Coding: Developers are embracing Anthropic's Claude for code; Google's Gemini may soon rival it.
- OpenAI's strength is in consumer-facing features like voice mode; Google’s developer experience is criticized as confusing.
“Each of them feel to me to be distinctly different qualitatively and quantitatively than GPT, which was the best we had at the start of the year.” — Azeem [09:57]
- Voice interaction—still early, but promising: Both use AI voice assistants for mundane and creative tasks; Azeem describes a daily workflow hack using Claude’s voice mode to capture and summarize thoughts during his commute [11:13–13:51].
5. The Consumerization of AI and Unlocking New Use Cases
- The ‘Uncanny Valley’ is being crossed: Voice cloning technology and intelligent agents are shifting from novelty to genuine utility.
“Once you get like the quality of audio cloning that if you were to share it, 98% of people would think it’s you … it is now usable and I can deploy it.” — Nathan [18:33]
- Opportunities in consumer & small business markets: AI is making previously-impractical tasks affordable and routine, such as taking detailed notes for every meeting or providing 24/7 customer support with accountability and reminders [20:13–21:35].
- Cautionary tales: Poorly-implemented AI customer service (e.g., Oura, SoundCloud) highlights the risks of rushing technology to market before it’s ready [22:30].
- New speculative horizons:
- Real-time, speech-to-speech translation at zero latency could debut in 2025 [24:17–24:41].
- The concept of a lifelong AI-generated autobiography based on passive data collection is floated as both outlandish and plausible [26:07–27:09].
6. AI for Science & Biotech: Promise & Pace
- Cautious progress, not miracles: While AI tools are making scientific data analysis, reproducibility, and routine tasks vastly more efficient, the ultimate test—an FDA-approved blockbuster drug undeniably enabled by AI—remains in the future [29:26–35:03].
“I think if you start automating experimentation, automating data analysis, having it be far more robust, …you can make quite a big dent into the irreproducibility of science.” — Nathan [31:31]
- Accelerating grant writing: Early, pragmatic uses such as grant brainstorming and drafting are already shaving months off scientific workflows [32:17–33:08].
7. Robotics Renaissance and Humanoid Robots
- Warehousing and logistics see rapid gains: Robotics has shifted from folly to practical renaissance thanks to leaps in general-purpose AI and labor shortages. Warehousing and logistics are set for rapid automation, while humanoid robots may first find footholds in homes or as luxury products solving simple, tedious tasks [36:40–39:49].
- Skepticism and realism on humanoid robots: Most economically useful tasks don’t require a humanoid form factor, but consumer desire—such as automating household chores—could drive early adoption [39:49].
8. Industry Predictions for AI in 2025
-
Three predictions from Nathan Benaich:
- Fully generative world games: Expect a video game hit that’s entirely generated by AI, unlocking new forms of play (see: Google DeepMind’s Genie 2) [40:25].
“I think there’s going to be a pretty cool video game that’s based around interacting with gen elements. Like the whole thing is generative.” — Nathan [40:25]
- No-code viral apps: 2025 could see an app or website created by someone with zero coding ability going viral—hitting the Apple Store top 100.
- Geopolitics and National Security: The boundary between major labs and government/nation-state involvement will become a live issue, particularly around funding and strategic control [41:12–42:15].
- Fully generative world games: Expect a video game hit that’s entirely generated by AI, unlocking new forms of play (see: Google DeepMind’s Genie 2) [40:25].
-
Singular determinant for 2025:
“If you’re not a frontier model, then nothing else matters.” — Nathan [42:52]
-
Growth & normalisation:
- Nathan predicts that growth in AI capability and usage will remain steady, as breakthroughs scale up, but the “wow” moments will become rarer as society adapts [43:34–44:19].
“The bar is so high now, like, oh, cool, you have a better voice model… Yeah, I’ve seen that already.” — Nathan [44:07]
- Nathan predicts that growth in AI capability and usage will remain steady, as breakthroughs scale up, but the “wow” moments will become rarer as society adapts [43:34–44:19].
Notable Quotes & Memorable Moments
- “Models are not products because people need help and they need guidance and they need some affordances in the product and the UI to get things done and prompts and things.” — Nathan [15:28]
- “Before, when you wanted to generate a narration for text … But once you have like a voice clone, I mean, this is game over. Like, why would you ever read a piece of text?” — Nathan [16:18]
- “The most exciting technologies are ones that unlock new user experiences and new product experiences that you couldn’t do before.” — Nathan [17:51]
- “It is magic. And I agree with you. … I sometimes I barely believe what I have today.” — Azeem [44:32]
Timestamps for Key Segments
- AI predictions & semiconductor landscape: 01:01–03:00
- Explosion of AI adoption and revenue in 2024: 03:00–05:30
- Open source vs. proprietary AI models: 05:31–07:45
- Product differentiation in model providers: 09:27–12:33
- Voice mode & daily workflow hacks: 11:13–14:12
- Unlocking consumer AI use cases & ‘uncanny valley’: 17:05–19:22
- Consumer AI’s impact on productivity & new ideas: 20:13–27:09
- AI in science, biotech, and grant writing: 29:26–35:03
- Robotics and humanoid robots in 2025: 36:40–39:49
- Major AI predictions for 2025: 40:25–42:15
- Frontier models & future determinants: 42:52–43:34
- Capability & usage growth, normalization: 43:34–45:10
Tone and Style
The conversation is candid, insightful, and often self-reflective with a balance of technical acumen and everyday pragmatism. Both speakers blend optimism about AI’s possibilities with a healthy skepticism and willingness to revisit and revise views as the field rapidly evolves.
This summary captures the essential content and flavor of the episode, preserving speaker insights, humor, and thought-provoking forecasts for AI in 2025.
