Applying data science to predict unrest. AI that can
anticipate the next variant of COVID-19’s structure. Reducing carbon emissions
from planes using algorithms. That’s a few of the headlines in AI this week,
which ran the gamut from the dour (how AI might prevent the next attack on the
U.S. Capitol) to the uplifting (making air travel greener). It’s caveated
optimism, but nonetheless a breath of fresh air in a community that’s becoming
increasingly cynical about the technology’s potential to do good.
Wired first reported that a researcher at the University of North
Carolina ran simulations using AI systems, including Alphabet-owned DeepMind’s
AlphaFold and the University of Washington’s RoseTTAFold, to predict the
protein structure of the Omicron variant of COVID-19. Ford managed to predict
one structure that was “pretty much right” – an impressive feat, given that he
arrived at his conclusions before scientists were able to map Omicron’s
structure properly.
AI promises to expedite certain processes in drug discovery and virology, for example identifying
compounds to treat conditions for which medications remain elusive. But
as Sriram Subramaniam, a professor at the University of British Columbia who
studied Omicron samples, told The Register, having access to a real sample still
beats algorithmic models. AI still can’t predict things like the strength of
new virus variants’ binding to host cells, for instance, or the infectiousness
of those variants.
Predicting social unrest
Could AI perhaps predict events like the January 6 attack on
the U.S. Capitol? A piece in The Washington Post this week investigates
the premise. While the consensus is mixed, some researchers believe that
algorithms can serve as early indicators of violence in regions ahead of major
political conflicts.
Unrest prediction, also known as conflict prediction, is a
burgeoning field in academia and industry. It and its practitioners, such as
the University of Central Florida’s CoupCast, aim to design systems that
consider variables (e.g., the role of a leader encouraging a mob, long-term
democratic history) to determine whether, for example, electoral violence might
occur.
Those who are bullish about the technology say that it’s
already revealed surprising insights, like the fact that social media conflict
is an unreliable indicator of real-world unrest. But others caution that it’s
little better than chance in terms of accuracy – and that it could be used to
justify crackdowns on peaceful protests.
“Actors react,” Roudabeh Kishi, director of innovation at the
nonprofit Armed Conflict Location & Event Data Project, a group engaged in
conflict prediction research, told The Post. “If people are shifting their tactics,
a model trained on historical data will miss it.”
Reducing jet emissions
The global aviation industry produces around 2% of all
human-generated carbon dioxide emissions. If they were a country, all the
airlines in the industry – some of which run thousands of nearly-empty flights to keep valuable
airport slots – would rank among the top ten in the world.
Like other greenhouse gases, carbon dioxide drives climate
change, leading to extreme weather, larger wildfires, disease from smog and air
pollution, food supply disruptions, and other effects. In an effort to combat
this, some airlines, including Air France, Norwegian, Malaysia Airlines,
Cebu Pacific, Go Air, and Atlas Air, are turning to algorithms trained on data
from billions of flights to identify emissions-reducing opportunities.
Openairlines’ SkyBreathe – the system recently adopted by Air France – can
reportedly reduce total fuel consumption by up to 5%.
Other startups, like Flyways, are creating AI-powered
platforms that attempt to optimize aircraft routing, giving suggestions on how
and where to fly planes. During a six-month pilot program at Alaska Airlines,
Flyways claims to have shaved off five minutes from flights and saved
480-thousand gallons of jet fuel on average.
Some critics argue that airlines aren’t going far enough;
they call for a phase-out of short-haul flights in Europe, among other
footprint-reducing measures. But considering the long road ahead to meaningfully
cutting the world’s carbon output, every bit helps.
“If you went a teeny bit slower, you were on time, you had a
gate, and because you went a teeny bit slower the airplane actually burned less
fuel, that might be a win/win combination for both the guest and the operation
and sustainability impact,” Diana Birkett Rakow, senior VP of sustainability at
Alaska Airlines, told ABC News.
.gif)
Tidak ada komentar:
Posting Komentar