Databricks CEO: AGI Is Already Here, Why Companies Are Behind, and the Superintelligence Debate
Episode
62 min
Read time
3 min
Topics
Career Growth, Productivity, Remote Work
AI-Generated Summary
Key Takeaways
- ✓AGI Definition Shift: The original 2009 AMP Lab consensus definition of AGI was achieved in 2024, but researchers moved the goalposts toward Bostrom's 2014 superintelligence definition — a system capable of performing a millennium of human thought in minutes. Leaders should separate these two distinct concepts when evaluating AI risk timelines and making strategic technology investments, as conflating them distorts both urgency and planning.
- ✓Four Enterprise AI Blockers: Organizations face four sequential barriers to meaningful AI deployment: capturing full organizational context (including meeting recordings), securing that data against expanded attack surfaces, managing exponentially rising token costs against revenue that does not scale equivalently, and avoiding vendor lock-in to protect proprietary IP. Solving all four is required before agentic AI delivers measurable business value beyond chatbots and basic coding assistance.
- ✓Company Ontology as Competitive Moat: Databricks' Genie product builds a continuous knowledge graph of the entire organization — org charts, meeting transcripts, Slack, documents, metrics — enabling CEOs to query culture, performance, and decision patterns in real time. Ghodsi identifies a correlation versus causation error live: work-life balance scores did not cause low action scores, with data showing managers scoring 44 on work-life balance still achieving 77% action ratings.
- ✓RSI Requires Four Simultaneous Conditions: Recursive self-improvement leading to superintelligence requires training runs to become faster than linearly, resource costs to decrease simultaneously, accuracy to improve across all domains generically, and all three conditions to compound in a continuous loop. Currently, pre-training runs are taking longer, require exponentially more hardware per marginal gain, and rely on extensive human-curated data cleaning, meaning none of the four conditions are currently being met.
- ✓Developer Output L-Curve: Databricks' internal data on 2,000 engineers shows a sharp L-shaped distribution of pull requests shipped to production weekly. The top performers commit disproportionately more code at equivalent or lower token cost by routing simpler tasks to smaller open-source models. Organizations can replicate this by identifying which engineers run automated code factories and systematically pushing the remaining 90% up the productivity curve.
What It Covers
Databricks CEO Ali Ghodsi argues AGI has already arrived by the original 2009 AMP Lab definition, while enterprises remain stuck using AI as glorified search. He demonstrates Databricks' Genie platform live, explains four conditions required for true recursive self-improvement, and outlines why enterprise AI adoption will take decades, not months.
Key Questions Answered
- •AGI Definition Shift: The original 2009 AMP Lab consensus definition of AGI was achieved in 2024, but researchers moved the goalposts toward Bostrom's 2014 superintelligence definition — a system capable of performing a millennium of human thought in minutes. Leaders should separate these two distinct concepts when evaluating AI risk timelines and making strategic technology investments, as conflating them distorts both urgency and planning.
- •Four Enterprise AI Blockers: Organizations face four sequential barriers to meaningful AI deployment: capturing full organizational context (including meeting recordings), securing that data against expanded attack surfaces, managing exponentially rising token costs against revenue that does not scale equivalently, and avoiding vendor lock-in to protect proprietary IP. Solving all four is required before agentic AI delivers measurable business value beyond chatbots and basic coding assistance.
- •Company Ontology as Competitive Moat: Databricks' Genie product builds a continuous knowledge graph of the entire organization — org charts, meeting transcripts, Slack, documents, metrics — enabling CEOs to query culture, performance, and decision patterns in real time. Ghodsi identifies a correlation versus causation error live: work-life balance scores did not cause low action scores, with data showing managers scoring 44 on work-life balance still achieving 77% action ratings.
- •RSI Requires Four Simultaneous Conditions: Recursive self-improvement leading to superintelligence requires training runs to become faster than linearly, resource costs to decrease simultaneously, accuracy to improve across all domains generically, and all three conditions to compound in a continuous loop. Currently, pre-training runs are taking longer, require exponentially more hardware per marginal gain, and rely on extensive human-curated data cleaning, meaning none of the four conditions are currently being met.
- •Developer Output L-Curve: Databricks' internal data on 2,000 engineers shows a sharp L-shaped distribution of pull requests shipped to production weekly. The top performers commit disproportionately more code at equivalent or lower token cost by routing simpler tasks to smaller open-source models. Organizations can replicate this by identifying which engineers run automated code factories and systematically pushing the remaining 90% up the productivity curve.
- •GDP Impact Will Lag by Decades: Despite current AI capabilities, GDP movement will be slow because enterprise adoption requires recording meetings (which major bank CEOs refuse), securing expanded data surfaces, and restructuring legacy processes built over 50-plus years. Ghodsi estimates ten to fifteen years for meaningful enterprise transformation, directly contradicting lab timelines of six to twenty-four months, citing the bank CEO who refuses meeting recordings as a concrete illustration of the gap.
Notable Moment
Ghodsi pulls up live Databricks HR survey data mid-conversation and disproves his own CFO's theory in real time — showing that poor work-life balance scores do not causally drive low action scores, with a manager scoring 44 on work-life balance still achieving a 77% employee action rating.
Episode Transcript
So it's kind of out of despair that we started Databricks because no one else is listening to it. At that time, there was zero interest in AI. Now I think we have all the intelligence we need to completely transform the planet. It's not possible to have a thousand agents doing stuff realistically. Am I correct? They're just using AI in very rudimentary ways. They're just using it as a glorified Google search. We're actually getting into, like, management money ball, aren't we? So if the token cost is going up exponentially, is your revenue at your company going up exponentially? But I actually think GDP will move very slowly, and it's gonna take decades. Are they a little drunk on their own success? Just because 98% of the code is being written by AI, we don't have RSI. If it automates the last 2%, who cares? When are you going public? I'm doing all of you. If you were building a global financial system from first principles today, you wouldn't build it on €50 legacy rails. You'd build Airwallex, one AI native platform for global accounts, cards, and payments. It's designed to make the entire world feel like a local market. Others are bolting AI onto broken infrastructure, but Airwallex was built for the intelligent era from day one. Stop paying the legacy tax and start building the future at airwallex.com/ all in. Airwallex, built for the future. Alright, everybody. Welcome back to the all in interview. We're so lucky because all of these amazing leaders in the world, CEOs, politicians, scientists, they they watch the show. They're part of the community. And Ali Gottsey is one of those individuals. He, of course, is the CEO of Databricks, a private company that has been doing incredible. I don't know why it's not public yet. I think a lot of people are asking that question. We'll talk about it today. In 2015, they were doing 1,500,000 in revenue. It's August... Or the the rumor I heard, August 2026, Ali, $7,000,000,000 run rate growing 80% year over year. And, obviously, you got a lot of competition, the the snowflakes of the world, etcetera. But this is a tremendous business, and we're gonna talk a lot about how you're using AI to run this incredible business. We're gonna talk about recursive, self improvement. But welcome to the program, and then maybe you could tell the audience a little bit about what you do at Databricks and how it became so successful. Yeah. Happy to do that. Thanks for having me on. Excited for this. Yeah. So we started the business in 2013, but if we rewind a little a little bit before that, we were all at UC Berkeley. There were about 20 of us, and we were actually part of a lab called AMP Lab, which was, I think, the largest AI research lab in the world at that time. And, you know, we were trying to figure out how …
Get the full transcript (13,390 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
Browse all All-In with Chamath, Jason, Sacks & Friedberg transcripts →
You just read a 3-minute summary of a 59-minute episode.
Get All-In with Chamath, Jason, Sacks & Friedberg summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from All-In with Chamath, Jason, Sacks & Friedberg
Jensen Ousted as Nvidia CEO: "Spy" Scandal, Export Restrictions, Market Impact
Nov 14 · 84 min
a16z Podcast
Databricks CEO on AI Pacing, Cyber Risk, and the Enterprise
Sep 18
More from All-In with Chamath, Jason, Sacks & Friedberg
Freedom vs. Safety: Axon CEO Rick Smith on the Future of Policing
Nov 10 · 55 min
The Prof G Pod
Schools Are Teaching Kids the Wrong Things — with Ted Dintersmith
Apr 2
Books, tools, and gear mentioned in this episode
SignalCast may earn commission on purchases via these links. As an Amazon Associate, SignalCast earns from qualifying purchases.
Books

by Nick Bostrom
“researchers moved the goalposts toward Bostrom's 2014 superintelligence definition — a system capable of performing a millennium of human thought in minutes”
Tools
- GenieBy guest
by Databricks
“He demonstrates Databricks' Genie platform live, explains four conditions required for true recursive self-improvement”
More from All-In with Chamath, Jason, Sacks & Friedberg
We summarize every new episode. Want them in your inbox?
Jensen Ousted as Nvidia CEO: "Spy" Scandal, Export Restrictions, Market Impact
Freedom vs. Safety: Axon CEO Rick Smith on the Future of Policing
Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What's Next for Trump
Anthropic S-1 Shocker, Greatest Business Ever? Google's $500B Wipeout, Midterm Predictions
WHOOP CEO, Will Ahmed: 50B Hours of Health Data, a 4 Million Member Data Moat, & the FDA Fight
Similar Episodes
Related episodes from other podcasts
a16z Podcast
Sep 18
Databricks CEO on AI Pacing, Cyber Risk, and the Enterprise
The Prof G Pod
Apr 2
Schools Are Teaching Kids the Wrong Things — with Ted Dintersmith
Invest Like the Best with Patrick O'Shaughnessy
Nov 11
Zach Reitano - Building Ro and Why Healthcare Should Be 50% of GDP - [Invest Like the Best, EP.500]
The Diary of a CEO
Nov 2
James Talarico: Is America's Political System Designed To Keep You Poor?
a16z Podcast
Oct 28
Mark Cuban’s Plan to Make Healthcare Affordable
Explore Related Topics
This podcast is featured in Best Tech Podcasts (2026) — ranked and reviewed with AI summaries.
You're clearly into All-In with Chamath, Jason, Sacks & Friedberg.
Every Monday, we deliver AI summaries of the latest episodes from All-In with Chamath, Jason, Sacks & Friedberg and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime