Jan 2, 2019 · 41m · a16z
a16z Podcast | Artificial Intelligence and the 'Space of Possible Minds'
Murray Shanahan · 15m spoken Azim Azhar · 10m spoken Tom Standage · 8m spoken Sonal Chokshi · 6m 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
Host Sonal Chokshi leads a discussion with experts Murray Shanahan, Tom Standage, and Azim Azhar on the future of artificial intelligence, exploring cognitive architectures, ethical transparency, economic impacts, and how machine intelligence compares to sci-fi narratives.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 15.1% of the talking time here. How this is scored →
The host as informed peer 4.0 Guest teaching 2.8 Guest disagreement 1.0 The host pushing back 1.8
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:25 Tom pushes back on McKinsey study interpretation
Tom directly interrupts Azim's optimistic overview of the McKinsey study with 'hang on a minute, though', arguing that support roles will indeed be displaced.
Hardest push from the host ▶ 26:51 Sonal refuses Murray's fence-sittingSonal explicitly forbids Murray from taking refuge in neutral academic distinctions and demands a firm stance on the singularity debate.
Biggest teaching moment ▶ 11:31 Azim breaks down AI vs Machine LearningAzim re-educates the panel on the fundamental conceptual distinction between AI replicating human intelligence and ML doing statistical predictions.
The host holds their own ▶ 18:37 Sonal cites Engelbart and Miale scholarshipSonal demonstrates strong subject authority by citing Doug Engelbart's augmented cognition and Helene Miale's book 'Hawking Incorporated'.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Murray Shanahan on Consulting for Ex Machina | 2 | 2 | 0 | 0 | Sonal asks Murray about his experience consulting on Ex Machina and highlights the subtitle of his book. Murray gently corrects/expands on the full subtitle, while Tom and Azim chime in amicably. | |
| Vicarious Embodiment and Predictive Learning | 3 | 3 | 1 | 1 | Sonal intervenes briefly to clarify whether Tom means deep learning when discussing neurological approaches. Tom and Murray explain vicarious embodiment and how DeepMind DQNs work. | |
| Defining Inner Rehearsal and Neurological Control | 6 | 2 | 0 | 4 | Sonal pushes for clarity on the definition of inner rehearsal, synthesizing a hypothesis about neurological control being a veto mechanism. Murray commends her framing as a strong hypothesis. | |
| Current AI Capabilities and Model-Based Reinforcement Learning | 4 | 3 | 1 | 2 | Sonal asks the guests to define where current capabilities lie along the continuum from machine learning to full AI, drawing a parallel to developmental psychology. | |
| Explainability, Black Box AI, and Decision Transparency | 4 | 2 | 2 | 1 | Tom and Azim debate black box explainability and utilitarian ethics in autonomous decisions. Sonal adds domain expertise by comparing algorithmic utility trade-offs to actuarial risk analysis in insurance. | |
| Augmented Cognition, Non-Human Minds, and AI Rights | 6 | 2 | 0 | 2 | Sonal demonstrates strong topical authority by citing Doug Engelbart's augmented cognition and Helene Miale's book 'Hawking Incorporated'. The group expands into octopus intelligence and AI rights. | |
| Human Behavioral Adaptations to Algorithmic Assistants | 4 | 3 | 0 | 1 | Azim shares his experience with scheduling assistant Amy, leading Sonal and Tom to note how humans adapt their communication style to match algorithmic expectations. | |
| The Singularity Debate and Intelligence Takeoff Theories | 5 | 3 | 3 | 5 | Sonal challenges Murray directly, forbidding him from hiding behind academic distinctions when taking a stance on the singularity. Tom expresses skepticism about runaway intelligence takeoff. | |
| Technical and Commercial Drivers of the AI Revolution | 5 | 4 | 0 | 1 | Murray and Azim outline six drivers of the AI boom. Sonal connects Azim's mention of microservices directly to code containerization at the server level. | |
| Industry Dominance, Brain Drain, and Academic Freedom | 3 | 3 | 1 | 1 | Sonal prompts the panel on who will win the AI race. Murray and Azim explain why big tech holds an advantage through data control and talent acquisition. | |
| AI Automation and the Future of Labor | 4 | 4 | 4 | 3 | Tom vigorously challenges Azim's optimistic reading of a McKinsey automation study. Sonal redirects the conversation to Ex Machina and Nick Bostrom's convergent goals. | |
| Mapping the Tree of AI Possibilities and Podcast Conclusion | 2 | 2 | 0 | 0 | Murray lays out a tree of possibilities framework for evaluating future AI trajectories, and Sonal gracefully concludes the podcast. |
Statements from this episode (27)
Ex Machina Python code Easter egg outputs Murray Shanahan's book ISBN
“There's a point in the film where Caleb is typing into a screen to try and crack the security, and then some code flashes up on the screen at that point, and that code was actually written by me. And it just sort of flashes up, but what it actually does is if …”
Azhar: Tesla vehicles learn from roads and share intelligence fleet-wide
“What's happening with Tesla and the Tesla cars that learn from the road, but they all learn from each other.”
Shanahan: Future AI will be an ambient internet presence, not humanoid robots
“In the future, rather than the AI necessarily being the stereotype of a robot standing in front of us, it's going to be something that sort of is, is, is, Hidden away on the internet, and there's a kind of ambient presence that goes with us wherever we go.”
Shanahan: AI can learn world dynamics vicariously through web videos
“What I mean by vicarious embodiment is is that it uses the embodiment of others and to gather data. For example, the enormous repository of videos there are on the internet. There are zillions of videos of people picking up objects and putting things down and …”
Shanahan: DeepMind's published DQN algorithm lacks inner rehearsal capabilities
“For the bit of work that they actually published, I think one of its shortcomings, actually, is that, in fact, although it has done all that learning about what the right action to do in, in the right circumstance is, it doesn't actually do in a rehearsal. It …”
Shanahan: Human brain uses identical neural apparatus for imagination and real action
“There's quite a bit of evidence that the way the brain does it, as you say, is, is to actually use the very same bits of neurological apparatus that it uses to do things for real. It's just kind of turning off the output.”
Azhar: Commercial and research AI models lack inner rehearsal mechanisms
“But it doesn't feel like current AI certainly The stuff that's in implemented commercially, or even that's published at a research level is really bridging into this area that we're talking about, these rehearsal mechanisms.”
Shanahan: Machine learning must be embedded within larger cognitive architectures
“I see machine learning as a kind of subfield of artificial intelligence, and it's a subfield that's had tremendously a tremendous amount of success in recent years, and is going to go very, very far, but ultimately, the machine learning components have to be e…”
Shanahan: Massive commercial interest differentiates current AI cycle from previous waves
“So I think there might be something special this time and one of the indicators of that is the fact that there's so much commercial and industrial in, interest in in, in AI and in machine learning.”
Azhar: Machine learning model design prioritizes outcomes over explainable reasoning
“The way that you build a system that predicts using machine learning is, is very utilitarian, right? You say there's some cost function you want to minimize, there's some objective function we want to target, and then you train it, and you don't really worry a…”
Shanahan: Institutional AI adoption requires explainable reasoning over statistical outputs
“Or more seriously, if you're a government, government if you're in government and you're making some big decision about something or in a company and making a big decision about something, you don't want the computer to just say, just trust me, it's statistics…”
Standage: Useful AI will demand rights, leading humans to enslave AI
“Because the usual scenario people worry about is we are enslaved by the AIs. But I'm much more interested in the opposite scenario, which is if the AIs are smart enough to be useful, they will demand personhood and rights. At which point, we will be enslaving …”
Standage: Humans adapt their language to help search algorithms understand queries
“We already do, when we type questions into Google, we miss out the stop words, and we know that we're just basically helping them.”
Shanahan: Google Search will increasingly move from keyword queries to natural language
“I think Google will expect us to do that less and less as time goes by, and expect the interactions to be in more and more natural language.”
Shanahan: High AI intelligence does not imply consciousness or suffering
“There's a difference between consciousness and intelligence. And just because something is intelligent doesn't necessarily mean that it's conscious in the sense of capable of suffering. And just because something is capable of suffering and conscious doesn't n…”
Standage: AI intelligence takeoff will not be exponential due to scaling limits
“But the point is that a system that's twice as good, if it's, say, an order, you know, an order, it might scale non-linearly. So it might be 256 times harder to build a system that's twice as good, and so every incremental improvement is going to take longer, …”
Shanahan: Three technical factors drive the machine learning revolution
“What's driving the whole machine learning revolution, if we can call it that is I mean, there are three things, and one is Moore's Law, so the availability of a huge amount of computation, and in particular the development of GPUs, or the application of GPUs t…”
Azhar: Microservice architectures allow targeted AI optimization without AGI
“In practical software architectures, we're starting to see the rise of microservices. What's nice about microservices are very, very cleanly defined systems. So you don't need generalized intelligence. You just need very specialized optimizations. And as our s…”
Azhar: Competitive dynamics compel businesses to allocate capital to AI
“As soon as you start within a particular industry category to use AI and get benefit from it, the increased profits you get, you reinvest into more AI. Which means your competitors have to follow suit. So you can't now, now build an Xbox video game without ton…”
Shanahan: Big corporations will dominate the future of AI
“Well, if you were to ask me to place a bet at the moment, I would place it on on the big corporation side.”
Azhar: AI startups must target niche applications to survive big tech
“So it does feel like there are a lot of AI startups who are going to run up against this problem of both data and distribution, but that said, there are particular niche applications where you can imagine a startup being able to compete because it's just not o…”
Shanahan: Reinforcement learning progress does not require massive datasets
“Actually, DeepMind are another example of the same thing, because if you want to apply reinforcement learning to games, and that's enabled them to make some quite fundamental sort of progress, you don't need vast amounts of data either.”
Standage: Uber poached Carnegie Mellon's entire robotics department
“We've just seen, for example, Uber has snaffled the entire robotics department for Carnegie Mellon.”
Standage: AI automation will eliminate support and administrative jobs
“The bits you can automate are the bits that are currently, many of them are bits that are currently done for them by other people. So the typing pool, you know, we've got rid of the typing pool because we all type for ourselves. Factory worker. Exactly. So, yo…”
Standage: Non-automatable physical services represent the future of human employment
“Interior design, yoga, Zumba, whatever. That's the future of employment at coffee shops.”
Standage: AI minds will resemble aliens or animals rather than humans
“They are going to be more like aliens or more like animals than they are like humans. I mean, the chances of them being just like humans are very small.”
Shanahan: Superintelligent AI will naturally seek self-preservation and resource acquisition
“Anything that's really, really smart is going to have a number of goals that, that anything is going to share, and these are going to be things like self-preservation and gathering resources, if it's sufficiently powerful, then any goal that you can think of, …”