Page 63 - The Economist

The Economist December 9th 2017 Business 63
2 own. But IBM’s weaknesses may not hold choose to use a service that is good ly that the incumbent tech groups will cap-
itback. Bossesofmostbusinessesfeel pres- enough, users are more likely to favour ture many of AI’s gains, given their wealth
sure to have an AI strategy, and they will self-driving cars with the best safety re- of data, computing power, smart algo-
pay handsomely to acquire one quickly. cord, meaningthatthe companiesthat best rithms and human talent, not to mention a
To date tech giants have mostly tried to employ AI to map out the physical world head start on investing. History points to
apply AI to reap profits from their existing and register the fewest crashes will enjoy the likelihood of concentration; both data-
operations. In the nextfewyearsthey hope outsize benefits. bases and personal computers ushered in
that AI will let them build new businesses. Each firm is approaching the problem ascendancies, if only for a while, of a tiny
One area of intense competition is virtual differently. Baidu, the Chinese giant, is try- group of tech firms (Oracle and IBM in da-
assistants. Smartphones know their users ing to create a self-driving-car operating tabases, Microsoft and Apple in personal
intimately, but AI-powered virtual assis- system, much like Google’sAndroid in mo- computers).
tants aim to take the relationship further, bile devices (although it is unclear how it By the metrics that count—talent, com-
whether through phones or smartspeak- plans to make money). Alphabet has its putingpowerand data—Google appears to
ers. Apple wasfirstto explore their promise own autonomous-car effort, as do Uber, be in the lead in AI. It can afford the clever-
when it bought Siri, a voice assistant, in Tesla, an electric carmaker, a herd of little- est people and has such a variety of pro-
2010. Since then Amazon, Google and Mi- known startups and, increasingly, estab- jects, from dronesto carsto smartsoftware,
crosoft have invested heavily: their assis- lished carmakers. (Apple is rumoured to that people interested in machine learning
tants’ speech recognition is better as a re- have scaled backits carambitions.) rarely leave. Other firms had to learn to
sult. Samsung, Facebook and Baidu are Self-drivingcarsare justone example of take AI seriously, but Google’s founders
also competingto offerthem. how technology firms’ AI strategies are were early devotees of machine learning
pushing beyond the virtual world of soft- and always saw it as a competitive edge.
One algorithm to rule them all ware into hardware. Many companies, in-
It is unclear whether standalone speakers cludingAlphabet, Apple and Microsoft, are AI’s spiritual home
will become a huge market, but it is certain also investing to build specialised, power- Some in the tech industry, such as Elon
that people will move beyond text to en- ful “AI chips” that can power their various Musk, the boss of Tesla and rocket firm
gage with the internet. “All these compa- activities. These will compete with those SpaceX, worry about Alphabet and other
nies understand that whoever owns that made by NVIDIA, a tech firm that has built firms monopolising AI talent and exper-
choke point for consumers will rule the an empire on powerful chips used in va- tise. He and a handful of other prominent
market,” says Pedro Domingos, author of rious AI realms, such as autonomous cars Silicon Valley bosses funded OpenAI, a
“The MasterAlgorithm”, a bookaboutAI. and virtual reality. not-for-profit research outfit focused on AI
Further into the future, augmented-re- It is unclear whether the likes of Alpha- with no corporate affiliation. MrMusk and
ality (AR) devices are another AI-infused bet and Apple will sell these chips to rival others are worried about what might hap-
opportunity. Mobile apps like Snap, a mes- firms or keep them for themselves. They pen when a firm finally cracks “general in-
saging app, and the game Pokémon Go are have an incentive to use their innovations telligence”, the abilityofa computerto per-
early examples of AR. But AR could more to improve their own services, rather than form any human task without being
radically transform people’s relationship renting or selling them to rivals—which explicitly programmed to do so. Such a vi-
with the internet, so that they consume could become a problem ifit means a very sion is probably decades away, but that
digital information notfrom a small screen fewfirmsdevelop a meaningful advantage does not stop Google from talking about it.
but via an ambient, ever-present experi- in brute computingpower. “We absolutely want to” crack general AI,
ence. AR devices will offer portable AI ca- That begs the broader question of says Jeff Dean, the boss of Google Brain. If
pabilities, such as simultaneous transla- whetherAI will furtherconcentrate power a firm were to manage this, it could change
tion and facial recognition. among today’s digital giants. It seems like- the competitive landscape entirely.
In the race for AR, big tech firms have In the meantime, much will depend on
not got much beyond the warm-up phase. whethertech firmsare open and collabora-
Google and Apple have launched AR soft- tive. In addition to publishing papers,
ware-development kits; they both want many companies today make their mach-
developers to build apps that use AR on ine-learning software libraries open
their platforms. There is also a rush to de- source, offering internal tools to rivals and
velop AR hardware. Google was early to independent developers. Google’s library,
launch a prototype for AR glasses, but they TensorFlow, is particularly popular. Face-
flopped. Microsoft has developed a head- bookhas open-sourced two ofits libraries,
set it calls HoloLens, but with a price ofbe- Caffe2 and Pytorch. Openness has strate-
tween $3,000-5,000, it is a niche product. gic advantages. As they are used, the librar-
Other firms, including Facebook and Ap- ies are debugged, and the firms behind
ple, are thought to be planning their own them get reputational benefits. “Beware of
offerings. Beingahead in AI could translate geeks bearing gifts,” quips Oren Etzioni of
into bigleads in these new fields. the Allen Institute for Artificial Intelli-
Nowhere is that truer than in the realm gence, anothernon-profit research group.
of autonomous vehicles. Tech firms are One guruofthe field worriesthat librar-
driving millions of miles to build up big, ies such as TensorFlow will bring in talent-
proprietary datasets, and are making use ed researchers but that their owners may
ofcomputervision to train theirsystems to start charging later on, or use them for pro-
recognise objects in the real world. The po- fit in other ways. Such caution may prove
tential spoilsare huge. Personal transporta- wise, but few think about the long term
tion is a vast market, worth around $10trn when a gold rush is underway. So it is now
globally, and whoever cracks self-driving in Silicon Valley. Most techies are too con-
carscan applytheirknowledge to other AI- sumed by the promise and potential pro-
based projects, such as drones and robots. fits ofAI to spend too much time worrying
Unlike search engines, where people may about the future. 7