Hacking Aging using Big Data and AI - Peter Fedichev Summary

Video

Watch presentation at StreamETH

Summary

• Aging research has reached an interesting technological transition point where solving aging and enabling radical life extension is a key frontier, akin to a “phase transition” defining the future order.
• The speaker’s AI company models aging through a physics/statistical mechanics lens, viewing it as differing degrees of stability/instability in regulatory dynamics across species. Mice exhibit an unstable, exponentially accelerating aging process that can be reversed/rejuvenated by reducing molecular error/damage levels. Humans instead show a more linear, irreversible accumulation of stochastic entropic damage that degrades recovery/regulation.
• Using machine learning on longitudinal data, they extract key markers representing the reversible “dynamic aging” component that drugs can target (as shown in mice) vs the irreversible “entropic aging” that most current treatments cannot sufficiently impact.
• In human data, functional measures like VO2 max and cognition decline linearly until hitting zero around 120 years old, while the variance of other factors increases hyperbolically, suggesting a loss of recoverability/regulation. This contrasts the exponential failure mode seen in mice.
• While life-extending for mice, drugs optimized in those unstable systems produce only minor, temporary benefits in humans by targeting reversible pathways akin to “counter-smoking.” A new approach is needed to impact the root, irreversible entropic aging that limits human healthspan and lifespan.

Key Takeaways

• Aging may represent a “phase transition” to a future where solving it is key, requiring a shift from disease treatment to directly targeting the root aging processes.
• Modeling aging through physics/statistical mechanics provides a unifying framework to understand differences across species in aging dynamics (stable vs unstable, reversible vs irreversible).
• Using AI/ML on longitudinal data allows extracting biomarkers representing the reversible “dynamic aging” that current drugs target vs the irreversible “entropic aging” that limits healthspan/lifespan.
• Mice exhibit an unstable, exponential aging process that can be reversed/rejuvenated by reducing molecular errors, while human aging is more linear and irreversible, with functional declines until a loss of recoverability around 120 years old.
• Drugs optimized in short-lived, unstable model systems have limited benefits in humans, so a new approach directly targeting the root irreversible/entropic aging processes is needed for radical human life extension.
• Novel interventions impacting the linear accumulation of stochastic entropic damage hold potential for multifold human life extension if the key drivers can be identified and addressed.

Speakers

• Peter Fedichev
• Founder of AI company focused on aging research
• Demonstrated expertise in physics, statistical mechanics, machine learning, and aging biology
• Key contributions:
• Modeling aging through physics/criticality principles
• Using AI/ML to extract aging markers from longitudinal data
• Differentiating reversible vs irreversible aging processes
• Experimental results on rejuvenation/life extension in mice
• Analysis of human aging as more entropic/irreversible