Thinking On Paper: Technology, Considered

The Vij brothers join Thinking on Paper to discuss Neo, an autonomous machine learning engineer designed to automate parts of the AI development process.

As demand for AI systems grows, companies and governments are competing for a limited pool of experienced machine learning engineers. The challenge isn’t only access to data or computing power. Many organisations also lack the technical expertise required to build, test and deploy effective models.

Neo uses a multi-agent system to perform tasks normally handled by machine learning engineers, including analysing datasets, selecting modelling approaches, running experiments and evaluating results. The aim is to automate repetitive technical work while allowing human engineers to concentrate on higher-level decisions and more creative problems.

In this episode, we discuss:

What an autonomous machine learning engineer is

How Neo’s multi-agent AI system works

Why skilled machine learning engineers are in such high demand

Which parts of AI development can be automated

How autonomous agents compare with traditional machine learning workflows

Why Kaggle Grandmasters are considered leading practitioners in applied machine learning

Whether AI agents can match expert human performance

How automation could affect machine learning jobs and salaries

The evolution of GPUs from graphics hardware to AI infrastructure

What the Vij brothers learned from working at CERN

How autonomous AI systems could change business, creativity and technical work

Neo is intended to expand access to machine learning expertise rather than simply generate code. Its development raises a wider question: what happens when AI systems can perform the specialised work required to build other AI systems?

This conversation examines the technical capabilities of autonomous machine learning agents, the shortage of experienced AI talent and how automation could reshape the role of engineers
--
Timestamps
(00:00) Why Are There So Few Machine Learning Engineers?(01:54) Meet Gaurav Vij and Saurabh Vij(02:57) Lessons Learned from Working at CERN(04:45) How to Explain The Importance Of A.I. to Your Parents(07:24) The World’s First Autonomous Machine Learning Engineer: What AI Problem Does NEO Solve?(08:17) AI Competitions and Kaggle Grandmasters(11:06) How Many A.I./ML Engineers Do We Need?(17:30) Fixing The A.I. Hallucination Problem(18:09) Hot Buttons: 5 AI Questions In 30 Seconds(18:46) Hollywood: Doomed by A.I, or Reborn?(20:26) AI News: Nvidia Digits Explained(21:51) Moore's Law And Could AI Models Be Motivated by Rewards?(25:42) AI And Quantum Computing(29:45) The Thinking on Paper Carry-Over Question(30:16) After Hours: Backstage Extra
--
Check out NEO: https://heyneo.so/Learn more about the show: www.thinkingonpaper.xyzFollow Thinking On Paper On Instagram: https://www.instagram.com/thinkingonpaperpodcast/

We read NASA’s Moon Base User’s Guide and ask what it would take to establish a sustained human presence on the Moon.

A permanent lunar base requires far more than rockets, landers and astronauts. NASA and its partners would need to build an integrated infrastructure system covering power generation, communications, navigation, habitats, transportation, logistics, robotics and resource extraction.

In this episode, we discuss:

How NASA plans to build a permanent Moon base

Why reliable power is essential for long-term lunar operations

Whether nuclear power will be required on the Moon

How astronauts, vehicles and robots would communicate and navigate

What lunar habitats need to protect crews from radiation and extreme temperatures

How autonomous robots could prepare sites and maintain infrastructure

Why lunar dust creates serious engineering problems

How equipment from different companies and countries could work together

Whether water, oxygen and construction materials can be extracted from lunar resources

What infrastructure must exist before humans can live and work on the Moon continuously

The discussion also examines the gap between NASA’s long-term ambitions and the systems currently available. Many of the technologies exist individually, but they haven’t yet been combined into a reliable, scalable lunar operating environment.

This episode asks whether a permanent Moon base is a realistic extension of human spaceflight or a programme whose infrastructure requirements remain badly underestimated.

--
Chapters

00:00 Executive Summary and Vision
01:17 Phased Approach to Moon Base Development
07:21 Challenges of Lunar Environment
09:06 Interoperability and Coordination in Space
15:13 Economic Incentives and Future of Space Development
17:03 Identifying Gaps in Space Technology
20:23 Functional Gaps and Their Implications
24:01 Dust Challenges and Solutions
29:10 The Moon as a Launchpad for Mars
31:08 Human Factors in Lunar Missions

--
Thinking on Paper is a technology podcast about AI, Space, quantum computing, science, and the systems shaping the future. 

🏠 Buy us a beer on Substack
🎧 Take us with you on YouTube
🎧 Remember steve jobs on APPLE
📺 Get the clips and outtakes on Instagram

Sanjay Vijendran of TerraSpark joins Thinking on Paper to explain how space-based solar power could become a practical source of clean energy.

TerraSpark is developing wireless power-transmission systems that could eventually collect solar energy in orbit and beam it to receivers on Earth. The company plans to demonstrate the concept by powering a live music event in Portugal and by testing radio-frequency power transfer aboard Dcube’s Arrakis mission.

In this episode, we discuss:

How space-based solar power works

How energy can be transmitted wirelessly

TerraSpark’s plan to power a concert in Portugal

What its in-orbit power-beaming experiment will test

The differences between radio-frequency and laser power transmission

How near-infrared power beaming works

How much energy is lost during wireless transmission

Whether orbital data centres could use the same infrastructure

How space-based solar could improve energy security

Why spectrum regulation and interference testing matter

What investors and regulators need to see before the technology can scale

Sanjay explains the engineering, regulatory and commercial challenges behind power beaming, including transmission efficiency, safety, spectrum allocation and the cost of placing energy infrastructure in orbit.

This conversation examines whether space-based solar power can move beyond demonstration projects and become a credible alternative to terrestrial energy generation and fossil fuels.

--

Thinking on Paper is a technology podcast about AI, Space, quantum computing, science, and the systems shaping the future. 

🏠 Buy us a beer on Substack
🎧Be With Us On YouTube
🎧 Remember steve jobs on APPLE
📺 Get the clips and outtakes on Instagram

--

Chapters

(00:00) Introduction to Space-Based Solar Power
(01:37) Market Trends and Projections
(03:52) Energy Crisis and Global Dependencies
(06:26) The Threat to Power Structures
(07:39) Innovative Demonstrations of Wireless Power
(10:31) Future Plans and Space Missions
(20:41) Scaling Power Transmission from Space
(22:35) Technologies for Space-Based Solar Power
(31:22) Governance and Regulation of Space-Based Solar Power
(49:57) The Future of Space-Based Solar Power

Jennifer Dunn, professor of chemical engineering at Northwestern University, joins Thinking on Paper to explain how lithium and copper mining affect water, ecosystems, local communities and the wider energy transition.

Lithium and copper are essential to electric vehicles, grid storage, renewable energy, drones and data centres. But the environmental consequences of extracting these minerals vary sharply depending on the mine, location, technology and supply chain.

Life cycle assessment offers a way to compare those impacts across different forms of production, from lithium brines and hard-rock mining to copper extraction, refining and recycling.

In this episode, we discuss:

The environmental impact of lithium mining

How lithium brine mining compares with hard-rock lithium mining

Why copper demand is rising

How mining affects water use and local water stress

The risks of pollution, biodiversity loss and mining waste

How life cycle assessment compares mines and supply chains

Why local conditions matter more than global averages

The role of mine permitting in the energy transition

Whether recycling can reduce demand for new mining

How battery supply chains shift environmental costs between regions

What responsible critical-mineral production should look like

Jennifer explains why no single measure can capture the full impact of a mine. Carbon emissions matter, but so do water availability, land use, waste, local ecology and the distribution of costs and benefits.

This conversation examines whether clean energy can scale without transferring environmental harm from fossil-fuel systems to the communities that supply lithium, copper and other critical minerals.

--

Thinking on Paper is a technology podcast about AI, Space, quantum computing, science, and the systems shaping the future. 

🏠 Buy us a beer on Substack
🎧 Take us with you on Spotify
🎧 Remember steve jobs on APPLE
📺 Get the clips and outtakes on Instagram --

Chapters

(00:00) Disruptors & Curious Minds
(02:10) The Demand for Copper and Lithium
(02:57) Environmental Impact of Mining
(05:59) Water Consumption and Mining Methods
(08:30) Community Concerns and Local Impact
(11:29) Recycling and Wastewater Mining
(14:04) Life Cycle Assessments in Mining
(27:06) Understanding Emissions in Mining
(29:45) Life Cycle Assessment: A Comparative Approach
(34:05) Stakeholder Perspectives on Mining Impacts
(37:42) Technology and Transparency in Mining
(42:42) Consumer Awareness and Ethical Sourcing
(48:55) Challenges in Quantifying Social Impacts

Anders Sandberg examines whether artificial general intelligence could manage the global economy more effectively than human institutions.

A sufficiently capable AI system might coordinate markets, allocate resources, interpret legal rules and respond to complex global problems faster than governments or companies. Greater efficiency, however, wouldn’t necessarily mean greater freedom.

In this short excerpt from a longer Thinking on Paper conversation, Anders discusses:

Whether AGI could manage the global economy

How superintelligence might improve global coordination

Why markets and legal systems are difficult to optimise

Whether AI could make better decisions than human institutions

How highly efficient systems could concentrate power

The challenge of keeping advanced AI under human control

How evolutionary pressures could shape competing software systems

Whether humans could become wealthier while losing political agency

What role people would retain in an AI-managed economy

The central question isn’t simply whether AGI could run economic systems better. It’s whether humans would still control the goals, rules and trade-offs behind those systems.

This is a short from a much longer conversation with Anders Sandberg about superintelligence, governance and the future of human decision-making.

--
Thinking on Paper is a technology podcast about AI, quantum, space and their impacts on society, business and culture. It's very good.

🏠 Buy us a beer on Substack
🎧 Watch on YouTube
🎧 Remember Steve Jobs on APPLE
📺 Get the clips and outtakes on Instagram