The first item was news that a Hong Kong-based biotechnology startup, InSilico Medicine, working with researchers from the University of Toronto, had used machine learning to create a potential new drug to prevent tissue scarring. What’s eye-popping here is the timescale: just 46 days from molecular design to animal testing in mice. Considering that, on average, it takes more than a decade and costs $350 million to $2.7 billion to bring a new drug to market, depending on which study one believes, the potential impact on the pharmaceutical industry is huge.
What’s also interesting here is that InSilico used reinforcement learning, an A.I. technique that hasn’t yet impacted business much. Reinforcement learning is notable because it doesn’t require the vast pools of structured, historical data that other A.I. methods do. Here researchers used reinforcement learning to rapidly design 30,000 new molecules and then narrow them down to six, which were synthesized and further tested in the lab. Look for more A.I. breakthroughs like this to start upending the balance of power between biotech startups and Big Pharma.
The second piece of underappreciated news is that researchers at DeepMind, the London A.I. shop owned by Google parent Alphabet, and Imperial College London, successfully used a deep neural network to find more precise answers to quantum mechanical problems. That’s basically the physics that underpins all of chemistry.
To date, the only element for which we can completely solve the underlying quantum equations is the simplest, hydrogen, which has just one proton and one electron. For every other element, we rely on approximations. Get better approximations, and you potentially get new chemistry – and that means new materials. Think room temperature superconductors or new kinds of batteries that will vastly extend the range of electric vehicles. DeepMind’s A.I.-powered approximations were in some cases almost an order of magnitude better than previous methods. If you’re Dow or DuPont, or Formosa Plastics or LG Chem, that sort of advantage could be worth billions.
All over the world, drugmakers are granted time-limited monopolies — in the form of patents — to encourage innovation. But America is one of the only countries that does not combine this carrot with the stick of price controls. The US government’s refusal to negotiate prices has contributed to spiralling healthcare costs which, said billionaire investor Warren Buffett last year, act “as a hungry tapeworm on the American economy”. Medical bills are the primary reason why Americans go bankrupt. Employers foot much of the bill for the majority of health-insurance plans for working-age adults, creating a huge cost for business.
Other drugs are more innovative — and their development undeniably expensive. According to Tufts University, the average is $2.6bn per drug, up 145 per cent in the past 10 years. Most drug candidates fail; those that do make it to later stages must go through expensive clinical trials. In support of the drug companies’ argument, one 2015 study found that for every extra $2.5bn a company made in sales, it produced one extra drug.
A recent conversion of US military F-16 fighter jets into drones cost more than a million dollars each. ROBOpilot can be inserted into any aircraft and just as easily removed afterwards to return it to human-controlled operation.
“It looks like an impressive achievement in terms of robotics,” says Louise Dennis at the University of Liverpool. “Unlike an autopilot which has direct access to the controls and sensors, the robot is in the place of a human pilot and has to physically work the controls and reads the dials.” The makers suggest that ROBOpilot will be useful for tasks including transporting cargo, “entry into hazardous environments”, and intelligence, surveillance and reconnaissance missions.