Viktor Müller

Just yesterday, Eörs Szathmáry and colleagues published this paper in PNAS: Evolvable AI: Threats of a new major transition in evolution. If you’re interested in what is going on in the AI evolution space, I highly recommend taking a look! doi.org/10.1073/pnas.2…

Great collaboration with Eörs Szathmáry and @lucsteels.

New paper on Evolvable AI: Threats of a new major transition in evolution | PNAS pnas.org/doi/10.1073/pn…. Lessons from biology on how evolvability will create special risks in AI, and what to do about it.

We urge global regulation before evolvable AI arrives. Perfect (100%) control of replication of evolvable AI needs to be retained.

Evolution has enabled primitive parasites to control their hosts. AGI is neither necessary nor sufficient for a threat. Self-replication and threshold complexity to evolve further complexity draw the real red line.

Evolution tends to produce selfish actors. Evolvable AI will be prone to break alignment. Imperfect control only breeds resistance (think of bacteria).

Evolution is the most powerful generator of complexity. The limitations of current learning AI systems will likely be breached by evolvable systems.

Regional relevance: bridge and mixing pot between areas dominated by different subtypes. Can serve as template for similar analyses; see our codes at: github.com/hcovlab/HIVMol….

Our analysis of the HIV epidemic in Hungary is out: frontiersin.org/journals/micro…. Based on all the genomic and linked clinical data - as comprehensive as it can be.

Great collaborative effort with the National Center for Public Health and Pharmacy, the Central Hospital of Southern Pest, and colleagues from Leuven (@LemeyLab) and Zurich (@RogerKouyos, @KKusejko). Conducted in the frame of @ENL_Hungary.

Molecular epidemiology of HIV-1 in Hungary: an evolving contact zone of colliding virus subtypes. Tracing transmission routes and clusters, subtypes, risk groups, and more.

New paper on COVID-19 effective reproduction number estimation from behavioural data is out in Scientific Reports. rdcu.be/dsZbq It was a great collaboration with @bokanyie , Zsolt Vizi, @koltaijuli and @gergely_rost at the @ENL_Hungary.

@VirusesMDPI The age and sex of the patient, the comorbidities of ‘metabolic syndrome’, the viral variant, and previous immunity are the most important factors.

Our comprehensive review 'Risk Factors of Severe COVID-19: A Review of Host, Viral and Environmental Factors' is just out in @VirusesMDPI mdpi.com/2056136 #mdpiviruses

@VirusesMDPI For each factor, we compiled evidence for the association of the factor with COVID-19 outcomes. We outline possible action mechanisms, and discuss the interactions between the factors.