No Priors Podcast Summary

Episode Overview

Title: Predicting the Earth with Josh Goldman: How KoBold Uses AI to Find Critical Minerals
Date: April 17, 2025
Hosts: Sarah Guo and Elad Gil
Guest: Josh Goldman, Co-founder of KoBold Metals

In this episode, Sarah Guo and Elad Gil speak with Josh Goldman, co-founder of KoBold Metals, about how KoBold leverages artificial intelligence and a vast, unique geoscience data repository to discover crucial mineral deposits such as lithium, copper, nickel, and cobalt. These minerals power batteries and the expanding demands of AI and electrification. The conversation delves into the intersection of mining, AI, philosophy, and global economic shifts.

Key Discussion Points & Insights

1. What KoBold Does & the Economics of Exploration

• Focus: KoBold Metals is an AI-powered exploration company focused not on mining itself, but on finding economically viable mineral deposits needed for a future powered by batteries and intelligent machines.
• Goldman’s explanation:
  “The economics of exploration are really quite extraordinary. With a few million dollars of capital, you can create 100 to a thousand times return... The problem with exploration as a business is that the success rate’s really low. You have to try many, many different places.” (02:09)
• Differentiation: Technology significantly increases the success rate of exploration, especially as surface minerals have largely already been discovered.

2. The Data Challenge: Scope, Sources & Structure

• Vast datasets: Human-collected geological data range from satellite imagery to hand-painted maps over a century old; it exists in both structured (e.g. elemental assays, geophysical surveys) and unstructured forms (e.g. text reports, maps, documents).
• Data wrangling:
  “There’s nowhere you can go where this is all aggregated in one place. You both have to do a lot of really hard technical work to get it together. And you have to do a lot of scientific work... There’s all kinds of messy problems with the data.” (03:15)
• Historic ground truthing: Old records (like 100-year-old Zambian maps) are still valid—“the rocks haven’t moved”—and are now used to train AI models.

3. Exploration as an Information Problem

• Scarcity redefined:
  “The scarce resource is not lithium or copper metal in the ground. It’s actually information. The scarce resource... is the information about where the ore deposits are located.” (07:40)
• Scientific hypothesis testing: Identify where geologic processes created concentrated deposits and validate hypotheses through data.

4. Geography, Regulation, and Local Realities

• Political and social context:
  Regulatory and local community constraints are as important as geology; projects must be economically, politically, and socially viable. Property rights, regulatory stability, and community consent (“social license to operate”) are all essential. (09:19)
• “Technical success is not very helpful...If we don’t actually have [community] relationships... then, we’re not going to be successful. But these are hyper local problems for sure.” (10:02)

5. KoBold’s Scale & Successes

• Current footprint: Over 60 projects across North America, Europe, Australia, and Africa.
• Major find: Mingomba, Zambia — “The highest grade large copper deposit not yet a mine... the core of it is over 5% copper and it’s very large.” (12:06)
• Why it matters: Higher grade reduces costs and environmental impact dramatically.

6. KoBold’s Technology Stack

• Three pillars:
  • Sensors – Custom and industry-leading for new earth data
  • Data Systems – Integration of all structured/unstructured data; LLMs (large language models) enable interaction with the data corpus
  • Models – Dozens of predictive models, retrained daily with new ground-truth data, for dynamic, probabilistic predictions (14:08)
• Continuous learning:
  “Every day that [field teams] are in the field, they are collecting new training data...We are retraining those models every day and serving new predictions out to the team.” (15:48)
• Custom tools: Example: Hyperspectral imaging system built by KoBold for rapid, cost-effective imaging and analysis. (17:48)

7. There is No “Silver Bullet” Dataset

• High dimensionality: True predictive power comes from combining diverse datasets, not single breakthroughs.
• Quote:
  “There is no way to isolate the AI from the HI [human intelligence]... when you can add dimensionality to the data, then you can have improved predictive power.” (19:34)
• Silver bullet fallacy: New tools add incremental gains but aren't sufficient alone.

8. Asset Valuation in Mining

• Straightforward modeling: Present value of future production is modeled based on known characteristics and standard engineering assumptions; it’s “much easier to know what a mine is going to produce 20 years from now than...what a SaaS company’s sales volume is going to be.” (21:12)
• Capital costs, operating expenses, and commodity prices drive modeling.

9. Industry Success Rate and KoBold’s Ambition

• Declining success:
  “In the industry it’s gotten 10x worse in the last 30 years because the problem has gotten harder and the industry is slow to innovate.” (24:10)
• Industry average: Less than 1 major successful deposit per billion dollars spent today.
• KoBold’s target: $50–100M per discovery, with Mingomba as a real-world proof point. (25:22)

10. Global Trends, Commodity Myths, and Market Realities

• Capital constraints:
  “Great projects don’t have problems getting funded... The problem is, there just aren’t very many great projects.” (26:14)
• Underexplored regions: Certain basins in Africa, especially deeper Zambian basins for copper; lithium globally remains underexplored since it wasn’t a focus until recently.
• Scarcity myths:
  “A lot of the noise about rare earths is because it has the word rare in its name... Not that rare.” (28:39)
• Scarcity more about downstream processing—especially concentration in China—than geological abundance. (29:03)
• Materials of concern: The highest-impact focus remains on scaling supply of copper, lithium, nickel, and cobalt for the battery and AI-powered future. No current “doomsday” material risk identified. (31:15)

11. The Role of Philosophy & Epistemology

• Scientific epistemology: KoBold operates as an "epistemic project"—success depends on making and testing predictions transparently, quantitatively, and with a mindset open to uncertainty and multiple simultaneous hypotheses.
• Culture:
  • Cobalt’s Epistemology of Exploration document: Core practices include making falsifiable predictions, avoiding confirmation bias, and always working with multiple alternatives. (32:40)
  • Unique role: They employ an in-house philosopher (Michael Strevens) to ensure scientific rigor in how knowledge and uncertainty are handled.
  • “Oddly, we have a chief philosopher who is an epistemologist... This really guides exploration practice and technology development.” (35:32)

12. Founder’s Journey & Motivation

• Personal path:
  • Josh’s background: Math, physics (quantum computing Ph.D.), consulting and private equity in energy.
  • Together with co-founder Kurt House they focused on non-fossil-fuel commodities critical to the future global energy system (batteries and AI).
  • Scale of need:
    “To build a future that is powered by batteries and AI by mid-century... we will need to mine over the next 25 years more copper than has been mined... in all of human history.” (39:02)
• Motivation: The intersection of solving hard technical problems and having tangible societal impact.

Notable Quotes & Memorable Moments

• Exploration’s core value:
  “The scarce resource is not lithium or copper metal in the ground. It’s actually information... The scarce resource is the information about where the ore deposits are located.” — Josh Goldman (07:40)

• On combining AI & human intelligence:
  “There is no way to isolate the AI from the HI.” — Josh Goldman (19:34)

• Philosophical approach:
  “Cobalt is kind of an epistemic project really. Our business is about making better predictions. That’s what we’re doing.” — Josh Goldman (32:41)
  “Oddly, we have a chief philosopher who is an epistemologist... This really guides exploration practice and technology development.” (35:32)

• On industry risk vs SaaS:
  “[Mining is] much easier to know what a mine is going to produce 20 years from now than... what a SaaS company’s sales volume is going to be 20 years from now.” — Josh Goldman (21:12)
  (joking exchange)
  Elad Gil: “I feel attacked.” (21:31)

• Rare earths and supply chain:
  “A lot of the noise about rare earths is because it has the word rare in its name.” — Josh Goldman (28:39)

• Data with no expiration date:
  “The rocks haven’t moved. So there’s no expiration date on the data.” — Josh Goldman (05:37)

Timeline of Important Segments

| Timestamp | Segment / Key Topic | |-------------|-----------------------------------------------------| | 00:38-03:10 | What KoBold does; exploration vs. mining | | 03:10-06:18 | Data types, history, and aggregation challenges | | 07:40-09:19 | Exploration as an information problem | | 09:19-10:57 | Regulation, social context, local engagement | | 11:08-13:46 | KoBold’s projects, scale, and the Mingomba find | | 14:08-16:21 | The AI/data/model stack and field operations | | 17:48-19:20 | Building custom sensors and continuous retraining | | 19:34-21:12 | "No silver bullet": the value of multidimensionality| | 21:12-23:53 | How resource projects are valued | | 24:10-25:39 | Industry success rates vs KoBold’s ambitions | | 25:39-26:40 | Funding, capital constraints, and market realities | | 26:43-28:22 | Underexplored regions, especially for lithium/copper| | 28:39-31:15 | Rare earth myths, processing, and supply chain | | 32:40-36:31 | Epistemology, philosophy, and prediction culture | | 37:19-40:26 | Josh’s personal journey & vision for KoBold |

Episode Takeaways

• KoBold Metals is reinventing mineral exploration using AI and unified, multigenerational data.
• The company’s predictive and epistemic philosophy, bolstered by a scientific culture and even an in-house philosopher, is core to their competitive edge.
• Higher discovery rates, more efficient and lower-impact mining, and a focus on information–not just geology–have enabled real world, high-value finds.
• Scaling supply of battery and AI-critical minerals will shape the coming decades; the biggest challenges are informational, social, and technological, not just physical.

This summary provides a structured guide to the episode’s substance, bringing forward the key points, color, and depth that listeners value from No Priors.