The
idea has been around for a bit. Jaron Lanier, the tech philosopher and
virtual-reality pioneer who now works for Microsoft Research, proposed
it in his 2013 book, “Who
Owns the Future?,” as a needed corrective to an online economy
mostly financed by advertisers’ covert manipulation of users’ consumer
choices.
It
is being picked up in “Radical Markets,” a book due out shortly from
Eric A. Posner of the University of Chicago Law School and E. Glen
Weyl, principal researcher at Microsoft. And it is playing into
European efforts to collect tax revenue from American internet giants.
In a
report obtained last month by
Politico, the European Commission proposes to impose a tax on
the revenue of digital companies based on their users’ location, on
the grounds that “a significant part of the value of a business is
created where the users are based and data is collected and
processed.”
Users’
data is a valuable commodity. Facebook offers advertisers precisely
targeted audiences based on user profiles. YouTube, too, uses users’
preferences to tailor its feed. Still, this pales in comparison with
how valuable data is about to become, as the footprint of artificial
intelligence extends across the economy.
Data
is the crucial ingredient of the A.I. revolution. Training systems to
perform even relatively straightforward tasks like voice translation,
voice transcription or image recognition requires vast amounts of data
— like tagged photos, to identify their content, or recordings with
transcriptions.
“Among
leading A.I. teams, many can likely replicate others’ software in, at
most, one to two years,” notes
the technologist Andrew Ng. “But it is exceedingly difficult to
get access to someone else’s data. Thus data, rather than software, is
the defensible barrier for many businesses.”
We
may think we get a fair deal, offering our data as the price of
sharing puppy pictures. By other metrics, we are being victimized: In
the largest technology companies, the share of income going to labor
is only about 5 to 15 percent, Mr. Posner and Mr. Weyl write. That’s
way below Walmart’s 80 percent. Consumer data amounts to work they get
free.
“If
these A.I.-driven companies represent the future of broader parts of
the economy,” they argue, “without something basic changing in their
business model, we may be headed for a world where labor’s share falls
dramatically from its current roughly 70 percent to something closer
to 20 to 30 percent.”
As
Mr. Lanier, Mr. Posner and Mr. Weyl point out, it is ironic that
humans are providing free data to train the artificial-intelligence
systems to replace workers across the economy. Commentators from both
left and right fret over how ordinary people will put food on the
table once robots take all the jobs. Perhaps a universal basic income,
funded by taxes, is the answer?
How
about paying people for the data they produced to train the robots? If
A.I. accounted for 10 percent of the economy and the big-data
companies paid two-thirds of their income for data — the same as
labor’s share of income across the economy — the share of income going
to “workers” would rise drastically. By Mr. Weyl and Mr. Posner’s
reckoning, the median household of four would gain $20,000 a year.
A
critical consideration is that if people were paid for their data, its
quality and value would increase. Facebook could directly ask users to
tag the puppy pictures to train the machines. It could ask translators
to upload their translations. Facebook and Google could demand quality
information if the value of the transaction were more transparent.
Unwilling to enter in a direct quid pro quo with their users, the data
titans must make do with whatever their users submit.
The
transition would not be painless. We would need to figure out systems
to put value on data. Your puppy pictures might turn out to be
worthless, but that college translation from Serbo-Croatian could be
valuable. Barred from free data, YouTube and Facebook might charge a
user fee for their service — like Netflix. Alternatively, they might
make their money from training A.I. systems and pay some royalty
stream to the many people whose data helped train them.
But
whatever the cost, the transformation seems worthwhile. Notably, it
could help resolve one of the most relevant questions coming into
focus in this new technological age: Who will control the data?
Today,
the dominant data harvesters in the business are Google and Facebook,
with Amazon, Apple and Microsoft some way behind. Their dominance
cannot really be challenged: Could you think of a rival search engine?
Could another social network replace the one all your friends are on?
This dominance might matter less if companies had to pay for their
users’ data.
Google
and Facebook and Amazon would not be able to extend the network
effects that cemented their place at the top of the technology
ecosystem to the world of A.I. Everybody wants to be on Facebook
because everybody’s friends are on Facebook. But this dominance could
be eroded if rivals made direct offers of money for data.
Companies
with different business models might join the fray. “This is an
opportunity for other companies to enter and say look, we will pay you
for this data,” Mr. Posner said. “All this is so new that ordinary
people haven’t figured out how manipulated they are by these
companies.”
The
big question, of course, is how we get there from here. My guess is
that it would be naïve to expect Google and Facebook to start paying
for user data of their own accord, even if that improved the quality
of the information. Could policymakers step in, somewhat the way the
European Commission did, demanding that technology companies compute
the value of consumer data?
In
any event, there is probably a better deal out there, in your future,
than giving Facebook free puppy pictures.