Interview
Manolis Kellis: Human Genome and Evolutionary Dynamics | Lex Fridman Podcast #113
- Manolis Kellis anticipates the podcast episode will be repeated soon.
- Future observers within 200 years will likely view current computer science and his work as "infantile" and "ignorant."
- Research plans include publishing on diet and exercise effects using single-cell profiling to analyze tissue reconfiguration.
- Humans will be trained to generate specific mind patterns to control machines before machines can fully comprehend human thoughts.
- High-resolution output interfaces reading thousands of neuronal connections are expected in the near future, whereas reading complex abstract thoughts remains unlikely for the near term.
- Language will remain the primary human-to-machine input method rather than complex direct neural impulses.
- The human brain is expected to control an artificial "third limb" at natural limb speeds if the machine maintains consistent responses.
- As of 2020, fully developed science for personalized nutrition based on genetics and microbiome matching was not yet available.
- The pandemic could leave a "tremendous imprint" on human genetic makeup if uncontained, similar to the Black Death, with genetic associations determining survival or death.
- The S1 protein of SARS-CoV-2 is expected to evolve rapidly, with the D614G mutation continuing to dominate outbreaks due to superior transmission in the human population.
- Future evolutionary dynamics will be predicted by analyzing current and likely future mutation regions.
- Human immune system diversity is expected to provide resilience against pandemics, ensuring the species is not wiped out.
- The pandemic is expected to teach the scientific community that knowledge is not fixed and that embracing mistakes is part of the process.
- Within 20 years (by 2040), new senses may be discovered to perceive currently "blind" physics or intangible aspects like "love."
- The democratization of knowledge faces the risk of abolishing respect for expertise, necessitating that educational systems teach epistemology alongside specific knowledge.
- Knowledge transfer speed during the pandemic was described as "phenomenal," citing the timeline from December 29, 2019, to the first genome sequence.
- Current understanding of brain function is "infantile," and the brain will continue to develop in an extremely malleable form late in life to facilitate education.
- Future AI systems must prioritize "resilience" over "perfection," learning from the messiness and fault tolerance of biological systems.
- Deep learning systems are advancing resilience by allowing natural parameter evolution to escape local optima.
- The human brain is viewed as a digital computer surrounded by vast analog processes, containing trillions of connections encoded by a single genome.
- Engineering can learn from biology by embracing "breaking things" to reach new optima, as nothing will ever be perfect.
- Neural networks are expected to perform well on human-like images but poorly on noise or random signals unmatching the physical world.
- The human genome is expected to be fully modeled to understand the entire universe of variables required to predict human steps.
- The FTO locus allele, once advantageous for obesity during ice ages, is now detrimental in the current environment.
- Future AI and human brains will make similar mistakes because both evolved for the physical world we inhabit.
- Ambiguity in language is a feature that drives creativity through misinterpretation and emergent communication network properties.
- The human genome is expected to be shaped by pandemics, with diverse immune repertoires ensuring species survival.
- Exercise and nutrition are expected to be the most effective ways to boost health aspects including Alzheimer's, cancer, and coronavirus outcomes.
- The microbiome is expected to be a more significant contributor to personalized nutrition than the genome itself.
- The placebo effect can be harnessed to improve healing by tricking the brain into initiating self-healing processes.
- Biological systems are extremely fault-tolerant, capable of withstanding significant genetic letter flips without notice.
- Life might be a distinct entity in the universe acting as a force beyond traditional physical rules.
- The quest for meaning itself is expected to be the meaning of life.
- Raising children is expected to complete the human journey by creating new components of the heart through vulnerability and love.
- The human brain continues to form new neuronal connections and neurons as knowledge is acquired throughout development.
- Conveying knowledge is essential, as understanding research is necessary to validate its existence.
- The path to engineering a deadly virus would likely start with a strain never having infected humans, whereas natural recombination of mixing and matching coronavirus strands is the probable source of hybrid viruses.
- Future biology will involve understanding non-immune functions co-opted from immune genes, such as microglia roles in cleaving neuronal connections.
- Non-immune functions of immune genes, such as microglia in neuronal connection cleavage, will be a focus of future biological research.