Real Alpha
Watch modeSubscribed

All-In · 3 min · Sep 10, 2025

Grok uses inference to correct and fill gaps in source material

Elon Musk describes using inference to assess source material, remove falsehoods and add missing context. The approach turns broad human knowledge into a more curated training corpus.

Elon Musk

Elon Musk is a Tesla CEO.

  • Chamath Palihapitiya, Host · Founder & CEO, Social Capital

    Led Facebook's growth team; his firm backed Slack and AI chipmaker Groq.

  • Jason Calacanis, Host · Founder & CEO, LAUNCH

    Early Uber angel; sold Weblogs Inc. to AOL; hosts This Week in Startups.

  • David Sacks, Host · Co-founder & partner, Craft Ventures

    PayPal's first COO, founded Yammer; White House AI & crypto czar 2025–26.

  • David Friedberg, Host · CEO, Ohalo

    Founded The Climate Corporation, sold to Monsanto for about $1.1B in 2013.

About this segment

Elon Musk describes using inference to assess source material, remove falsehoods and add missing context. The approach turns broad human knowledge into a more curated training corpus.

From “Elon Musk on DOGE, Optimus, Starlink Smartphones, Evolving with AI, Why the West is Imploding”, Sep 10, 2025. Watch the full episode

“the page to correct, remove the falsehoods, correct the half-truths, and add the missing context.”
Elon Muskat 27:55 · All-In27:55

Takeaways

  1. Grok assesses source material as true, partly true, false or missing.
  2. It rewrites entries to remove errors and add context.
  3. Musk says the process applies to books, PDFs and websites.

In this episode

Full episode
  1. 2:57Optimus hands are the hardest part to build at scale7:31
  2. 40:13Mars must build the means to sustain its own civilization4:30
  3. 28:52Musk expects AI to surpass human intelligence within five years2:35
  4. 12:16AI5 targets a 40x gain on an AI4 bottleneck4:37
  5. 16:46Starlink's direct-to-phone service is about two years away3:15
  6. 20:16Starship's heat shield remains a hurdle to full reuse6:00

  7. 26:23Grok uses inference to correct and fill gaps in source material2:35

Up next · All-InPalantir says its safeguards make data analytics harder to abuse6:30

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

  1. 26:28

    You want to just give us an update? I think you kind of talked about where the next-gen model is, and you said something incredible. I still don't think people really understand it, which is there's going to be a next training run where you expect not to start from the common web and common crawl, where you expected an enormous amount of synthetic data. Just tell us about how the evolution of Grok is going and this innovation and why it's so important.

  2. 26:58

    Yeah, so we're using a lot of inference compute and reasoning to look at all of the source data, which is really the corpus of human knowledge. And then thinking about each piece of information and then adding what's missing and correcting mistakes and removing falsehoods from that training data. So if you take, say, Wikipedia as an example, but this really applies to books, PDFs, uh the websites uh every form of information um the grok is using um heavy amounts of inference compute let's say to look at as an example a wikipedia page and say uh what is true partially true or false or missing in this page. Now rewrite.

  3. 28:00

    the page to correct, remove the falsehoods, correct the half-truths, and add the missing context. Elon, by the way, could you just publish that? Could we create like a Grokipedia? Yeah, especially for our bio pages, which are a disaster. Wikipedia is so biased, and it's a constant war. If something gets corrected five minutes later, there'll be an army of people trying to... I mean, it's become hyper-partisan, and there's activists all over it.

  4. 28:34

    So if you do fix, for example, Wikipedia as a source of truth, it'd be great to publish that just so the world has it. All right, I'll talk to you about that. So talk to the team about that, like Grokpedia or whatever. Here's the Grokpedia version. It'd be interesting, yeah. And then just have it out there for a few minutes.