Mar 31, 2023 · 1h 24m · allin
E122: Is AI the next great computing platform? ChatGPT vs. Google, containing AGI & RESTRICT Act
David Sacks · 22m spoken Jason Calacanis · 18m spoken David Friedberg · 17m spoken Chamath Palihapitiya · 16m 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
In Episode 122 of The All-In Podcast, the hosts debate whether generative AI represents a major computing platform shift, comparing OpenAI's plugin ecosystem against Google's distribution power and Innovator's Dilemma. They also analyze AI safety concerns, white-collar job displacement, Joe Manchin's policy critiques, and the civil liberty threats posed by the RESTRICT Act.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 99.6% of the talking time here. How this is scored →
The hosts as informed peer 6.8 Guest teaching 3.4 Guest disagreement 3.0 The hosts pushing back 3.2
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 34:38 Sacks forcefully dismisses claim that Google will catch OpenAI
Sacks interruptingly rejects the idea that Google will steamroll OpenAI, pointing out reports that Google engineers had to rely on ChatGPT output to train Bard.
Hardest push from the hosts ▶ 20:29 Chamath challenges platform shift narrative with blocker conceptChamath directly challenges Sacks's assertion that OpenAI has secured a dominant platform advantage by introducing the concept of data blockers and commoditized third-party endpoints.
Biggest teaching moment ▶ 38:56 Friedberg diagnoses Google's organizational paralysisFriedberg draws from internal knowledge to explain how regulatory scrutiny, EU fines, and legal risk aversion bound Google's executive management long before ChatGPT launched.
The host holds their own ▶ 1:07:14 Chamath highlights AI's unique replacement of human judgmentChamath refutes standard economic historical analogies by demonstrating that unlike historical physical automation tools, AI is directly displacing human judgment in closed loops.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| All-In Podcast Title Sequence | 5 | 3 | 2 | 2 | The co-hosts engage in informal political banter surrounding Joe Manchin's op-ed on the Inflation Reduction Act. Friedberg highlights CBO deficit projections to critique the legislation's accounting, while Sacks analyzes Manchin's electoral position. | |
| Episode Welcome and Host Background Banter | 6 | 2 | 1 | 1 | Sacks details how he used ChatGPT as a research assistant, editor, and drafting partner for a Substack post on startup data strategies. The group lightheartedly teases Sacks for his unusually polite tone toward the AI model. | |
| OpenAI Plugins Launch and the Platform Shift Debate | 7 | 4 | 3 | 4 | Sacks frames OpenAI's plugin announcement as a major platform shift comparable to iOS, while Chamath pushes back by questioning whether exclusive data blockers actually exist to prevent model commoditization. Friedberg explains how natural language APIs bridge search and transactional commerce. | |
| Silicon Valley Reputation and AI Hype Cycle Caution | 6 | 3 | 3 | 3 | Chamath cautions against hyping AI cycles too early given public backlash toward Silicon Valley following the SVB collapse. Sacks argues OpenAI's user distribution and web browsing plugins build a self-reinforcing lead. | |
| Data Islands versus Open Plugin Platform Network Effects | 6 | 4 | 4 | 4 | Jason argues major platform data owners like Facebook and Twitter will block ChatGPT access to protect proprietary data islands. Sacks counters that competitive pressure will force services into OpenAI's plugin ecosystem similar to App Store adoption. | |
| Google Bard's Distribution Threat and the Power of Bundling | 6 | 3 | 3 | 3 | Jason predicts Google Bard will steamroll OpenAI by bundling AI directly into Android, Chrome, and YouTube. Chamath agrees that incumbent distribution channels present a massive defensive threat. | |
| Can Incumbents Disrupt Themselves? OpenAI versus Google | 7 | 4 | 5 | 5 | Sacks rejects the premise that Google will easily catch up, pointing to reports of Google engineers training Bard on ChatGPT outputs and OpenAI's unreleased GPT pipeline. Chamath maintains that early wave leaders rarely capture all ultimate value. | |
| Google's Innovator's Dilemma and Regulatory Scrutiny | 8 | 5 | 2 | 2 | Friedberg delivers an insightful breakdown of Google's Innovator's Dilemma, explaining how antitrust fears, EU fines, and internal legal risk-aversion paralyzed product rollouts despite superior underlying research. | |
| Leadership Dynamics, Microsoft Parallels, and Bureaucratic Friction | 7 | 3 | 3 | 3 | The panel discusses organizational culture, comparing Google's current state to Microsoft under the DOJ consent decree. Chamath argues middle management risk-aversion and social governance concerns stifle wartime execution. | |
| Pause AI Experiments Petition and Existential AI Risks | 7 | 3 | 2 | 2 | The hosts discuss the Future of Life Institute petition calling for a pause on AI training. Sacks shares an example of ChatGPT diagnosing a rare illness to save a pet dog, illustrating immediate utility while recognizing long-term alignment questions. | |
| AI's Impact on White-Collar Jobs and Creative Destruction | 8 | 5 | 4 | 5 | Jason reads a letter from a 3D artist expressing grief over Midjourney replacing creative tasks. Friedberg and Sacks frame AI productivity as creative leverage, while Chamath argues AI is unique because it supplants human judgment in closed loops. | |
| The RESTRICT Act and Threats to the Open Internet | 8 | 2 | 4 | 4 | The hosts strongly criticize the RESTRICT Act, arguing it acts as a trojan horse that expands executive surveillance and criminalizes VPN usage under the guise of banning TikTok. |
Statements from this episode (25)
Chamath: Ron Lauder switched support from Trump to DeSantis
“Ron Louder flipped from Trump to DeSantis. That's a big one because Louder's good for A lot of money, five to ten million at least.”
Sacks: ChatGPT reduced blog research and writing time from a week to a day
“A blog that probably would have taken me a week to research and write if I had done it at all, I was able to do in a day.”
Sacks: ChatGPT plugins are the most important developer platform since iOS
“I think this is the most important developer platform since the iPhone and the launch of iOS and the app store. And I would argue maybe ever. In our industry, certainly since the beginning of the internet.”
Chamath: LLMs will converge into commodities without unique proprietary data
“All of these models will converge in the absence of highly unique data, right? What I've been calling these white truffles. So if you can hoard white truffles, your model will be better. Otherwise your model will be the same as everybody else's model.”
Chamath: Government will intervene much faster in the AI cycle
“Government will get involved much faster in this cycle”
Sacks predicts OpenAI's lead will likely grow over the next year
“However, I actually think that OpenAI has demonstrated now with these platform features that it has a lead, a substantial lead, and I actually think that lead is likely to grow in the next year.”
Calacanis: Facebook and LinkedIn will block ChatGPT from accessing their data
“Somebody like Facebook, when they release their language model, which they're about to, is not going to allow ChatGPT to have any access to the Facebook corpus of data. And then LinkedIn will do the same. They'll block any access to ChatGPT to their data.”
Sacks: ChatGPT will eat a substantial portion of search and browser usage
“And I think that ChatGPT will start By eating a substantial portion of search, because again, you don't have to go through the 20 links. It just gives you the answer. It's going to eat a substantial portion of browser usage and app usage, because you're just g…”
Calacanis: Google will roll over ChatGPT using its distribution advantage
“So I think that Google will roll. I think they're going to roll ChatGPT.”
Sacks: Former employee claimed Google trained its AI using ChatGPT
“And the proof of that is there was just a lawsuit a couple of days ago, or at least a claim by a former employee of Google who quit because he said that they were using ChatGPT to train Their AI.”
Sacks: OpenAI used GPT-4 internally for seven months before release
“Moreover, chat GPT four, which was just released a few weeks ago. We know that open AI had that. They were using it internally for seven months.”
Friedberg: Google has the lowest AI model serving and training costs
“Google has arguably the best corpus of data upon which to train, the best capabilities, the best toolkit, the best hardware, the lowest cost for running these sorts of models, the lowest cost for serving them, etc, etc.”
Sacks: Google will be a top four AI player but won't steamroll OpenAI
“Do I think that Google will be one of the top four players in AI? Absolutely. But this idea that is going to come and steamroll open AI, I'm not buying it.”
Friedberg: Google's defensive executive culture puts its entire business at risk
“There is an institutional problem at Google at the top level, which does need to be solved, which is this position of constantly being in defense against the scrutiny, again, of regulators and public policy folks and, you know, all these different groups that …”
Chamath: Google's middle management prioritizes ethical debates over survival
“I think the problem at Google is not Sundar or Larry or Sergey. I think it's more in the deep bowels of middle management of that company, which is that there's just far too many people that probably have an opinion and their opinion is not shrouded in surviva…”
Sacks: 70% of recent Y Combinator batch were AI startups
“70% of the last YC class was ready, all AI startups.”
Friedberg: Regulatory constraints and IP rules cannot contain AI
“I don't think you can contain it. I don't think that this idea that we can put in place some regulatory constraints and say it's illegal to do this, Or, you know, try and, you know, create IP around it or protections around it is realistic at this state.”
Chamath: IT outsourcing firms like Accenture and TCS will displace labor first
“I think there are a handful of companies you need to look at exclusively because they will be the first ones to really figure out how to displace human labor. And that is TCS. So Tata Consulting Services, Accenture, Cognizant. These are all the folks That do c…”
Sacks: Natural-language AI tools will soon let anyone create software without coding
“Even really before forming a company, anyone with a good idea is going to be able to do that soon. You're going to be able to use these AI tools. They truly will be no code. You'll be able to create an app or a website just by speaking to some AI program in a …”
Friedberg: Everyone will eventually have AI-generated movies and games made on the fly
“In some period of time, we will all have our own movie or our own video game ultimately generated for us on the fly based on our particular interests.”
Chamath: Portfolio company has fully closed-loop AI cancer detection system under FDA review
“There are closed loop systems. We have one right now that's in front of the FDA. That is a total closed loop system that will not need any human input.”
Chamath: AI will eliminate unit testing and most coding
“Look, I think that AI is going to eliminate unit testing. It has already done. So it's going to eliminate most forms of coding.”
Calacanis: AI will not displace software developers for 10-20 years
“I don't think we'll see it in software development for a decade or two. There's just so much software that still needs to be made.”