Jan 13, 2025 · 28m · big-technology
The Future of AI and Entrepreneurship, According Two Leading Innovators
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
At LG InnoFest, host Alex Kantrowitz moderates a panel with Siri co-founder Norman Winarski and LG Nova's Sokwoo Rhee exploring multimodal AI interaction, viable startup strategies against foundation model giants, and emerging frontiers across digital healthcare and physical robotics.
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 30.7% of the talking time here. How this is scored →
Alex as informed peer 4.0 Guest teaching 4.7 Guest disagreement 1.6 Alex pushing back 1.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 17:14 Sokwoo rejects premise that apps will disappear
Sokwoo directly rejects Alex's hypothesis that standalone apps will vanish, arguing vendor incentives make umbrella unification unrealistic.
Hardest push from Alex ▶ 9:29 Alex presses foundation model existential threatAlex challenges the long-term viability of AI application startups by questioning why frontier model builders racing toward AGI will not simply subsume their products.
Biggest teaching moment ▶ 12:31 Norman explains Siri's intent-driven do engine visionNorman clarifies the conceptual foundation of Siri, educating Alex that it was originally created as an execution-oriented 'do engine' rather than a search tool.
Alex holds their own ▶ 19:51 Alex demonstrates multimodal prompt engineering in ClaudeAlex showcases practical technical expertise by detailing how he configured Claude's multimodal vision features to analyze food labels and score meals.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Debating Voice AI Versus Multimodal Interaction | 5 | 4 | 3 | 2 | Alex brings up Mark Zuckerberg's statement predicting voice will dominate AI interaction. Siri co-founder Norman unexpectedly pushes back against voice supremacy in favor of text and multimodal interaction, prompting Alex to cite reading versus speaking speed benchmarks. | |
| Key Ingredients for Successful AI Startups | 3 | 4 | 1 | 1 | Alex asks what separates successful AI ventures from the wave of acqui-hires and fizzled startups. Norman outlines fundamental startup criteria including acute pain points, large markets, and strong value propositions across vertical industries. | |
| Enterprise AI Opportunities and Infrastructure Dynamics | 4 | 4 | 1 | 1 | Alex argues enterprise software provides clearer near-term AI ROI than consumer products. Sokwoo builds on this with a historical analogy to the early internet fiber-optic infrastructure cycle preceding application deployment. | |
| Foundation Models Versus Vertical Market Dominance | 5 | 5 | 2 | 3 | Alex questions whether massive foundation model builders chasing AGI will inevitably subsume and destroy specialized AI startups. Norman and Sokwoo explain why vertical domain data, specialized workflows, and consumer segment focus protect application-layer companies. | |
| Siri's Original Do Engine Vision and Modern AI | 3 | 6 | 1 | 1 | When asked if AI is becoming an operating system, Norman explains Siri was originally architected as an intent-based 'do engine' rather than a search engine, a paradigm modern multimodal agents are now realizing. | |
| AI as an Enabler for Lean Teams and Workforce Evolution | 4 | 5 | 3 | 2 | Alex explores whether individual apps will become obsolete under unified AI operating layers. Sokwoo directly disagrees, arguing distinct corporate incentives will prevent full app unification, while Norman clarifies that apps will operate transparently as background services. | |
| Personalized Natural Language Apps and Privacy Risks | 6 | 3 | 1 | 1 | Alex explains how he personally programmed Claude with multimodal photo inputs to act as a custom diet evaluator. Norman validates the concept of personalized natural language software while warning about consumer privacy and ad-targeting risks. | |
| Promising Startup Frontiers: Beyond Language to Physical Waves | 3 | 6 | 1 | 1 | Alex asks about promising startup frontiers beyond standard language models. Norman explains that transformer architectures are advancing beyond text into physical kinematics, robotics, and biological signal interpretation like EEG brainwaves. | |
| Long-Term Predictions for AI Over 5, 10, and 15 Years | 3 | 5 | 1 | 1 | Sokwoo and Norman provide 5, 10, and 15-year forecasts, charting the progression from near-term vertical software tools to physical embodied robotics and revolutionary biotech breakthroughs. |
Statements from this episode (15)
Winarski: Voice will not be the primary interface for AI
“I disagree that voice is going to be the primary way that we're going to interact with AR.”
Winarski: Text will remain a dominant AI interaction form for a long time
“Text is going to be a dominant form, and for a long time.”
Rhee: Enabling technology usually does not make much money directly
“Usually, enabling technology does not make a lot of money.”
Rhee: Real market value will eventually be in vertical application AI startups
“Eventually, there will be a lot of the application-focused startups or vertical-focused startups that leverage, say, I think that's sort of really real market is, but I also understand, without infrastructure, we cannot get there.”
Winarski: Foundation AI giants will never sufficiently focus on market verticals
“So from my point of view, I believe that these big companies will never try to focus sufficiently on the market verticals that will really make a huge difference to that particular market, and so the industry will flourish in each of these individual markets.”
Winarski: Siri was created as a 'do engine', not a search engine
“When we first created Siri, our vision was very specific. It was a do engine. For, instead of a search engine. And what we meant by a do engine was Understand a person's intent, a request to do something. It always had to be a request to do something. Buy me a…”
Winarski: Apple repurposed Siri in 2010 to serve Apple Calendar and Mail
“Apple purchased Siri Steve Jobs, and they, in 2010, they took Siri as it was and reinvented it for the Apple purposes, so that would access Apple Calendar, Apple Mail, All of those things.”
Rhee: AI lets startups launch with one or two people, not ten
“The companies that new startups are coming in previously maybe took 10 people to start a company, decent you know, ramp up. Now they can do it, you know, one person or two. Basically that's because it makes easier for startups to do what they did using enterpr…”
Rhee: Mobile apps will not disappear because companies will not unify
“Actually, I don't agree that apps will disappear, and here's why. If you have a four, let's say you have a 30 different apps, in theory, you don't need 30 apps. You need, maybe, if all these vendors and companies work together for one single app, that can do e…”
Winarski: Users will stop interacting directly with standalone mobile app interfaces
“So I do think that everybody will have their own Application business, but I don't think we'll see it the same way as we do today with your iPhone or Android phone.”
Winarski: Real-time visual AI assistants are likely, but privacy risks loom
“So more and more this, your vision, what you just said is not only possible, but likely unless people have access to your information and start sending you ads every which way.”
Winarski: Working with Google Spin-Out NexSense on Brainwave Sensing
“I'm working with a spin-out out of Google on this called NexSense.”
Winarski: Next AI Wave Will Move Beyond Language to Brainwaves and Robotics
“So you, we're going to start seeing generations beyond text and language going into other new breakthroughs, and we're going to see the same impact in those ways.”
Rhee: In 10 to 15 years, home-assisting robots may become a reality
“10 years, 15 years, I think it will happen. So then, all the sci-fi movies that we are seeing hopefully, maybe robots helping us at home, it may not be a stretch.”
Winarski: In five years, AI will yield vertical monopolies and novel cancer therapies
“Five years from now, we will, first of all, see major industries that have formed in each of these market verticals that are dominating those verticals. And we'll also have seen some incredible new science come out. Biotech with cancer treatments and discovery…”