Patrick

Investor in frontier technology companies and advocate for AI-driven biological research.

Mentions

Appeared in

Discussed in

Key positions and views

The simulation of human biology via 'virtual cells' using foundation models is the most critical moonshot project for the future of medicine.

A new fundamental architecture for deep learning is overdue and likely to emerge around 2025, potentially supplanting the current dominance of Transformers.

The true economic potential of AI lies in the automation of the services economy through AI agents, which will be a far larger disruption than its impact on software.

Interdisciplinary research, as practiced by the ARC Institute, is the most effective organizational model for solving complex biological problems.

While AI's potential is immense, it's crucial to distinguish between proven applications (e.g., diagnostics, protein folding) and overhyped capabilities (e.g., current toxicity prediction).

Podcast consensus on Patrick

Points of consensus

▶Patrick consistently argues that AI, particularly foundation models, is poised to revolutionize biology and drug discovery, citing the ARC Institute's 'virtual cell' project as a primary vehicle for this transformation.Apr 2026

▶He views interdisciplinary collaboration as a critical component for scientific breakthroughs, highlighting the ARC Institute's intentional co-location of experts from fields like neuroscience, immunology, and machine learning.Apr 2026

▶Patrick believes the economic impact of AI will be most profound in the services sector, viewing AI agents as a disruptive force far exceeding the software economy's scale.Apr 2026

▶He emphasizes that while AI has overhyped areas, its capabilities in specific domains like protein structure prediction (AlphaFold), pathology, radiology, and protein binding are proven and significant.Apr 2026

Points of debate

▶Patrick presents a nuanced view on AI's current capabilities in drug development, championing its proven success in diagnostics and protein tasks while simultaneously asserting that its ability to predict drug toxicity is currently overhyped.Apr 2026

▶He highlights the monumental financial success of GLP-1 drugs, which created over a trillion dollars in market value, while also noting the systemic inefficiency of the broader industry where approximately 90% of drugs fail in clinical trials.Apr 2026

▶Patrick expresses high confidence in AI's potential to double human lifespans within a decade, citing Dario Amadei, yet he also characterizes the foundational technology for this (virtual cells) as being in a very early, 'GPT-1 to GPT-2' stage of development.Apr 2026

▶He speculates that a new, fundamental deep learning architecture is 'overdue' and likely to emerge around 2025, while also endorsing Sakana AI's research on improving existing architectures through techniques like model merging and evolutionary Mixture of Experts (MoE).Apr 2026

Key themes

▶AI as a Foundational Tool for BiologyApr 2026

Patrick posits that AI, specifically foundation models, will enable the simulation of human biology through 'virtual cells'. This technological leap aims to shift drug discovery from a trial-and-error process to an in silico, predictive science, allowing for the identification of novel drug targets and compositions computationally.

Investors should monitor progress in biological foundation models, as a 'GPT-3 moment'—such as rediscovering the Yamanaka factors—would signal a paradigm shift in biotech R&D, potentially de-risking early-stage drug development.

▶The Next Architectural Shift in AIApr 2026

Patrick speculates that the field of deep learning is due for a major architectural innovation around 2025, based on a historical eight-year cycle of paradigm shifts. He is also closely watching companies like Sakana AI that are exploring novel techniques beyond simple scaling, such as model merging and evolutionary selection within MoE frameworks.

Analysts should be wary of assuming that current architectures like the Transformer will dominate indefinitely; a breakthrough in efficiency or capability could rapidly reorder the competitive landscape and create new investment opportunities.

▶Economic Disruption via AI AgentsApr 2026

Patrick's primary economic excitement for AI is not in software but in its potential to automate and disrupt the much larger services economy. He predicts that AI agents for general computer use will follow the development path of coding agents, progressively increasing their capacity for autonomous, error-free work.

The true market for AI may not be in selling models but in deploying autonomous agents that perform services, suggesting that companies focused on agent orchestration and reliability will be key players.

▶Organizational Innovation in Scientific ResearchApr 2026

Patrick champions the ARC Institute as an 'organizational experiment' designed to break down silos between scientific disciplines. By co-locating researchers from genomics, machine learning, immunology, and other fields, the institute aims to accelerate discovery by fostering organic, interdisciplinary collaboration.

The structure and culture of a research organization can be as crucial as its technology; entities that successfully merge disparate fields of expertise may possess a significant competitive advantage in tackling complex problems like Alzheimer's disease.

Source episodes

Sentiment over time

Not enough data for timeline

Changes over time

Current State Analysis

Patrick outlines the current biotech landscape, characterized by massive but isolated successes like GLP-1s, contrasted with a systemic 90% failure rate in clinical trials.

Assessment of AI in Healthcare

He evaluates the tangible impact of AI to date, noting proven capabilities in pathology, radiology, and protein folding (AlphaFold), while cautioning that areas like toxicity prediction are overhyped.

Championing a New Research Model

Patrick details the mission of the ARC Institute, presenting its interdisciplinary structure and 'virtual cell' project as a necessary evolution in scientific research to overcome existing bottlenecks.

Defining a Key Benchmark

He establishes a clear benchmark for a breakthrough in virtual cell modeling: the ability to computationally rediscover the Nobel Prize-winning Yamanaka factors for cell reprogramming.

Future Projections

Patrick offers forward-looking predictions, including the imminent arrival of a new deep learning architecture (c. 2025) and the potential for AI to double human lifespans within the next decade.

Suggested prompts

How does Patrick reconcile his view of virtual cell modeling being in a 'GPT-1 to GPT-2 stage' with the prediction that AI could double lifespans in the next decade? &nearr;What specific evidence underpins Patrick's prediction of a new deep learning architecture emerging around 2025, beyond the eight-year cycle observation? &nearr;How does Patrick's investment portfolio in companies like New Limit and Nudge directly align with his stated thesis on virtual cells and AI-driven biology? &nearr;What are the primary obstacles Patrick identifies, implicitly or explicitly, to achieving the 'in silico target ID' goal of the ARC Institute? &nearr;

Key concepts

ARC Institute 5 ep Virtual Cells 5 ep AI in Drug Discovery 4 ep Deep Learning Architectures 3 ep Clinical Trials 2 ep AI Agents 2 ep Protein Folding & Design 2 ep Sakana AI 2 ep GLP-1 Drugs 1 ep Yamanaka Factors 1 ep

Notable quotes

“Our moonshot is really to make virtual cells at ARC and simulate human biology with foundation models.”

Patrick · Faster Science, Better Drugs

“the amount of market cap added to Lilly and Novo based on the, you know, development of GLP-1s is like over a trillion dollars is more... than the market cap of all biotech companies combined over the last 40 years have been started”

Patrick · Faster Science, Better Drugs

“could it predict that the four yamanaka factors would reprogram the fibroblast into a stem-like state right and essentially rediscover from the model something that won the nobel prize in 2009”

Patrick · Faster Science, Better Drugs

“if you look at the history of deep learning, it's like every eight years, there's something really different, right? And it feels like in 2025, we're really overdue for some net new architecture.”

Patrick · Faster Science, Better Drugs

Report last updated: Apr 21, 2026

Get started free

Back to Entities Intelligence Report