The breathtaking advances of the modern computer age have been built, in part, on the idea of miniaturization — that every two years, twice as many transistors could be embedded on a computer chip, doubling its speed — a principal called Moore’s Law. In the next decade, the 60-year run of this venerable law will likely come to an end, as the electronics of ever tinier transistors begin to enter the strange, vanishingly small world of quantum physics, where the familiar rules of our classical reality don't seem to apply.

Rather than seeing quantum behavior as an impediment, theoretical physicists like Del Maestro are exploiting its bizarre qualities to create a new paradigm: quantum computers that could be up to 100 million times faster than today’s supercomputers — revolutionizing research in the basic sciences and in fields ranging from drug development and financial modeling to cyber-security and weather forecasting.

Quantum computing is far from a futuristic dream and simple quantum computers have already been built in advanced university labs and a few companies, such as Google, IBM and Microsoft.  However, just as the first mainframes took up entire rooms, these machines are unwieldy, require extremely low temperatures to operate, and can utilize only a handful of quantum bits.

Professor Del Maestro uses the extreme classical computing power of the Vermont Advanced Computing Center as a sort of laboratory to study the entanglement in quantum liquids at the atomic scale.  “Like a super-powered microscope, the VACC lets us directly observe this strange quantum handshake between atoms in a manner that wouldn’t be possible in a brick-and-mortar lab.”  Del Maestro believes these uniquely quantum correlations will be used to fuel the next generation of practical quantum computers that could operate outside of specialized facilities and allow us to collaboratively tackle the challenging problems of our age.

Adrian Del Maestro,
Associate Professor, Physics
Director, VACC
www. delmaestro.org/adrian/
Tweets by ‎@agdelma

In less than two decades since the first human genome was sequenced, rapid technological advancement has led to an explosion of research in the field of epigenetics, or the biological mechanisms responsible for gene expression. At UVM, researchers are using epigenetics to answer a vast range of research questions, including how to produce better beef.

For producers in the multi-billion dollar beef industry, cattle health, efficiency and profitability are top priorities. Stephanie McKay, professor in UVM’s Department of Animal and Veterinary Sciences, is applying epigenetics to help understand which genes are responsible for specific, desirable traits in beef cattle, such as having a docile temperament.

Aggressive cattle pose a continuing problem to animal welfare, production and profitability for herd managers. Less aggressive, or docile cattle, have been found to be more feed efficient, more disease resistant and produce less methane – the second most abundant greenhouse gas. By identifying which genes are associated with docile temperament, McKay aims to help beef producers make selective breeding decisions to produce better beef, improve animal welfare and lower their environmental impact.

Thanks to modern next-generation DNA sequencing, it’s become relatively easy to generate huge volumes of sequencing data, but having access to computing technologies to make sense of that data presents a bottleneck for many in the field, says McKay. Thanks to the computing power of the Vermont Advanced Computing Center, McKay is able to process whole genome sequencing data at record speeds and has been able to identify which genes are associated with docility in cattle.

Having computing capabilities to process this data within UVM also enables a rich learning experience as McKay’s students can access high-performance computing programs at the tips of their fingertips, and has also helped secure highly competitive grant funds. With no end in sight for utilizing the capabilities of the VACC, McKay is exploring new research opportunities using genomic sequencing to help improve conservation efforts of moose.

Stephanie McKay,
Associate Professor, Animal and Veterinary Science
http://blog.uvm.edu/sdmckay/

Neuroimaging has great potential to shed light on the brain structures and functions that underlie depression, psychosis, addiction and other psychopathologies. But to date the technique has been limited by a lack of longitudinal data. The brain scans of cannabis users might look different from non-users, for instance, but the variation may well have been pre-existing. 

A $300 million study funded by the National Institute of Health aims to address that issue by following a cohort of 12,000 adolescents for 10 years, from roughly age 10 to 20. Researchers at 21 sites, including the University of Vermont, will gather information from questionnaires, bio samples, and both functional and structural MRI’s in marathon meetings every two years, with briefer information gathering sessions in between.

At the end of the decade, says UVM lead Hugh Garavan, researchers hope to gain a much clearer understanding of one of society’s most pressing problem: what mechanisms cause psychopathology to develop in a growing number of teens – a life period when the vast majority of mental illnesses develop – and what interventions could prevent the onset of disease.

Standing in the way of that inspirational goal?  A colossal amount of data – tens of millions of data points for every child in the study – that researchers need to crunch and draw meaning from.

The UVM site will play a leadership role in that effort. The university has expertise not just in neuroimaging, thanks to Garavan and his team, but also – via the Complex Systems Center – in machine learning, which can tease meaning out of just the kind of giant data sets the project is amassing. 

Without the Vermont Advanced Computing Center, the work would be impossible. Using machine learning to identify telltale patterns in the brains of the 12,000 study participants requires tremendous computational power. “The VACC is absolutely indispensable,” Garavan says.

NIH is relying on UVM’s unique combination of leading edge expertise and infrastructure. To help make sure all the data the project gathers isn’t for naught, it recently awarded the university a grant to train graduate students in the Complex Systems Center to work with large neuroimaging genetic data sets using the VACC.

“This is an exciting time to be at UVM,” Garavan says.

Hugh Garavan,
Professor, Psychiatry
https://www.researchgate.net/lab/Hugh-Garavan-Lab

According to Hollywood and an army of op ed writers, artificial intelligence will be either humanity’s salvation or its doom. But creating machines that have the capacity for good or evil is still a far distant prospect – the stuff of science fiction. Intelligent machines that can harm humans out of ignorance, on the other hand, are a scientific reality today, as the four deaths attributed to self-driving cars demonstrate.

Making machines that are smart, yet safe, is the goal of leading edge roboticists like Josh Bongard. How do we build intelligent machines that are not only able to make decisions, Bongard asks, but also unfailingly make the right ones?  

To answer, Bongard relies on the VACC. He uses the facility’s vast computing power to stress test a growing assortment of virtual prototypes funders have asked him to investigate -- from self-driving vehicles to robots that fly by flapping their wings to sub-millimeter devices that deliver drugs to disease targets in the body. Using millions of simulations generated by the VACC, Bongard throws every imaginable adversity at the virtual machines, forcing weaker designs to fail. The VACC not only creates the challenging scenarios, it evolves the virtual prototypes so they’re progressively more likely to succeed. 

Without  the VACC, the work would be impossible. A typical project with 1.5 million simulations takes a week-and-a-half at the facility. A conventional computer would need 450 years.

What’s most exciting about the work for Bongard is the contribution he’s making to human welfare. In the case of the sub-millimeter drug delivery device, its ancestors will have experienced billions of simulations before human trials begin. “We want them to make the right decision 99.9999 percent of the time before a human ever swallows one.”

Josh Bongard,
Professor, Computer Science
http://www.cs.uvm.edu/~jbongard/
Tweets by ‎@DoctorJosh