Playing God From 9 to 5
ResearchersattheBerkeley Robot
Learning Lab work to create machines
that can learn on their own and may
one day achieve human intelligence.to parking private vehicles. “AI is the new
buzzword,” says Jason Calacanis, an entre-
preneur in San Francisco. “You just use the
phrase ‘artifi cial intelligence’ in your busi-
ness plan and everyone pays attention. It’s
the fl avour of the month.”
That kind of scepticism is justifi ed. AI
can spot a cat in a photo and parse words
when you talk. But perception is not rea-
soning. Seeing is not thinking. And mas-
tering Go is not like liv ing in the real world.
Before AI can be considered intelligent,
much less dangerous, it must be taught to
reason. Or, at least, to have some common
sense. And researchers still have a long way
to go in achieving anything that resembles
human intelligence or consciousness.
“We went through one wall, we know
how to do vision now, and that works,”
says LeCun. “A nd the good news is we have
ideas about how to get to the next step,
which hopefully will work. But it’s like
we’re driving 50 mph on the highway in
the fog and there is a brick wall somewhere
that we’ve not seen. Right now we are just
happy driving until we run out of fuel.”mit physicist max tegmark, 48, has
a bowl haircut and a boyish eagerness that
make him seem younger than he is. In his
two-storey suburban house near Boston,
the living room is sparsely furnished, with
pictures of ducks and woodchucks on the
wall. As a physicist and cosmologist, Teg-ful ad platform? Better algorithms that can
predict ads you will click on. Even a 0.5
per cent diference in click-through rates
can mean enormous amounts of money to
a company with $50 billion in revenues.
Image recognition, which depends on ma-
chine learning, is one place where the com-
petition is now fierce between Apple, Mi-
crosoft, Google and cloud services like
Dropbox. Another battleground is perfect-
ing speech recognition. The company that
can figure that out first – making talking
to a machine as natural as talking to a per-
son–willhaveahugeadvantage. “Voice
interface is going to be as important and
transformative as touch,” says Baidu’s Ng.
Google and Apple are buying up AI start-
ups that are promising to of er smarter as-
sistants, and AI is crucial to the success of
self-driving cars, which will have a tremen-
dous impact on the auto industry and po-
tentially change the look and feel of cities,
oncewewillnolongerneedtodevotespaceed AI algorithms need to function. A new
kind of chip, called the graphics process-
ingunit–whichwasoriginallycreatedfor
video-game processing – has been partic-
ularly important for running neural net-
works that can have hundreds of millions
of connections between their nodes.
Thesecondbigchangeisthearrivalof
bigdata.Intelligenceinmachines,likein-
telligence in humans, must be taught. A
humanbrain,whichisgeneticallyprimed
to categorise things, still needs to see
real-lifeexamplesbeforeitcandistinguish
betweencatsanddogs.That’sevenmore
true for machine learning. DeepMind’s
breakthrough with Go and Atari games re-
quiredthecomputertoplaythousandsof
gamesbeforeitachievedexpertise.Partof
theAIbreakthroughliesintheavalanche
of data about our world, which provides the
schooling that AIs need. Massive databas-
es,terabytesofstorage,decadesofsearch
resultsandtheentiredigitaluniversebe-
cametheteachersnowmakingAIsmart.
In the past, the attempt to create a
thinkingmachinewaslargelyanexercise
carried out by philosophers and comput-
er scientists in academia. “What’s difer-
enttodayisthestufactuallyworks,”says
Facebook’s LeCun. “Facebook, IBM, Mi-
crosoft–everybodyisdeployingit.And
there’s money in it.” Today, whatever com-
pany has the best learning algorithms and
data wins. Why is Google such a success- Photograph bySpencer Lowell
87