The new machine-learning techniques transforming how digital systems operate don’t work like old software programs. Once upon a time, programmers wrote code, and the code worked — or it didn’t, and the programmers debugged. But how do you debug a neural network? In Technology Review, Will Knight explores “the dark secret at the heart of AI”: We can’t really know why AIs do what they do. They are not programmed, but trained.
When a machine-learning program provides an answer, a decision, or a choice that its human operators believe is wrong, they can tell it so, and it will incorporate that data into its next choices. But most AIs can’t turn around and tell us how they reached a particular outcome. This “explainability” problem poses practical, legal, and moral questions we’re only beginning to scope out.
Let’s circle back one more time to Treasury Secretary Steve Mnuchin’s extraordinary comment that artificial intelligence is “not even on my radar screen” and won’t affect the job market for 50 to 100 years. What was he smoking? And please keep it away from us, okay?
The double whammy of AI and robotics, what observers are calling “the fourth industrial revolution,” is certainly on the rest of the world’s radar, and already having an impact on transportation, manufacturing, retail, medicine, education — everything. We can’t know how this wave of change will play out; scenarios range from utopia to doomsday, and we’re already beginning to live them.
Artificial intelligence experts gathered in Asilomar, California in January to ponder the ethical and philosophical challenges their field presents. The choice of location echoed a historic 1975 conference at Asilomar that grappled with the future of genetics and biotech in cosmic terms. While the recent event was private, Cade Metz in Wired reports that the nightmare scenarios it raised were less of the science-fiction-apocalypse variety and more centered on mundane economic concerns.
The headline: AI-driven automation will “eliminate far more jobs far more quickly than…expected.” In other words, the long-term threat to the middle class comes much less from globalization than from technology, and the Trump-era focus on closing the borders is a waste of effort.
Globalization and outsourcing have already reshaped the American workforce. Next up: work-on-demand and automation. That’s the future-of-labor picture according to venture capitalist Fred Wilson (AVC), who moderated a discussion on the topic at our just-concluded NewCo Shift Forum.
What we’re seeing, when you put these trends together, is the unbundling of employment. “Job” has always been a name for the package our economy has used to tie together a bunch of different things: work that an organization needed to get done; payment for that work to individuals with the skills to do it; benefits that provided long-term security to the worker; and laws that defined the obligations on both sides of the “job” relationship. All of that is now in flux.
It makes a great sound-bite: For every new rule federal regulators under President Trump want to add to the books, they’re going to have to scrap two existing regulations. That’s the gist of an executive order Trump signed Monday — “a big one,” as the president called it — in a move to fulfill a campaign promise to reduce the regulatory burden on American business (The Atlantic).
Sounds fine. But wait a minute: Um, how do you actually count a rule? Is there some standard unit of regulation? Most regulations are complicated and include a lot of different standards, guidelines, and conditions; that complexity is why businesses complain so much about them. Does a particular environmental rule — about, say, coal power-plant emissions — count as one single rule or many? And rules vary immensely in the burden they place on businesses; if you institute one humongous new rule and scrap two trivial ones, what kind of win is that?
Twitter was a potent weapon in Donald Trump’s campaign arsenal. But the power of social media constantly mutates, and right now it’s working against the Trump team’s efforts to control the flow of information from the ranks of the federal civil service out to the public.
Most incoming administrations go through rough patches as their political appointees take the reins of career-run agencies. But this transition is the first one to feature rogue Twitter accounts pushing back against the dictates of new bosses — or at least, as in the case of the National Parks social media managers who decided to Tweet-blast straightforward climate facts, do what they view as their public-information jobs. The Ringerhas a good summary of January 2017’s “digital insurrection” (the Associated Press’s apt phrase).
Voice computing has been revving at the starting gate for some time. It’s finally having its breakout moment with the market success of Amazon’s Echo device and Alexa voice assistant. The emergence of voice has implications well beyond the opportunity to bark out Amazon orders and song requests to a compliant digital helper (The Economist): “Computers without screens and keyboards have the potential to be more useful, powerful and ubiquitous than people can imagine today.” At its best, it feels like casting spells.
But the magic of voice commands comes with a cost. Echo is a corporate owned microphone listening in on your living room. The whole system works better the more fully you let it in on all the details of your life. How prepared are you for that?
If you’ve been reading all year about “machine learning” but still feel you only have a fuzzy grasp of what it is, relax. Having a fuzzy grasp of what something is turns out to be exactly what machine learning is all about. You can glean that and much more from Gideon Lewis-Kraus’s epic artificial-intelligence narrative in The New York Times Magazine, whichtells how Google sharpened its translation skills by plugging in a new machine learning engine.
As he traces the trail of Google’s researchers and engineers, Lewis-Kraus also provides a beautiful summary of how machine learning evolved from a renegade strain of artificial-intelligence theory into a practical tool for delivering useful services — and a powerful harbinger of socioeconomic disruption.
Bring on the cyber. If you’re old enough, you remember the brief moment two decades ago when referring to the online digital world as “cyberspace” actually seemed ahead-of-the-curve. That ended fast, but somehow the prefix “cyber-” found a survival niche in the world of foreign-policy wonks and security pundits. It came roaring back to life in last night’s presidential debate, tumbling out of Donald Trump’s mouth in a muddled monologue that left jaws agape and younger viewers, particularly, in giggles (The Verge). Trump railed against ISIS recruiting and sang the praises of his 10-year-old son’s computing prowess, and by the end of the segment, #TheCyber had become a meme. Once the laughing subsided, we could all glumly realize that neither of these aging candidates has a visceral understanding of the digital world that shapes so much of our experience today. Trump has a thing for Twitter, and Clinton may have had her issues with email. But for both of them, “the cyber” seems a forbidding alien landscape — while for a growing proportion of the electorate, it is simply the ground on which we must build our work and our lives. (Props, though, to Clinton for bringing in a crew of digital natives to craft her technology policy.)
Palantir charged with bias against Asians. Palantir Technologies, the Palo Alto-based security and analytics firm that’s high on anyone’s list of “cyber” companies, is being sued by the U.S. Department of Labor for discriminating against Asian job applicants (Reuters). Federal agencies like the CIA, the FBI, and the Pentagon are also among Palantir’s biggest customers. The suit comes as Silicon Valley faces growing criticism for its failure to diversify its work force. Palantir co-founder Peter Thiel has made headlines recently for funding the lawsuit that brought down the feisty Gawker media business, and also for his support of Republican candidate Donald Trump (Thiel spoke at the Republican convention). Will Palantir claim it’s being persecuted because of Thiel’s controversial profile? Will the lawsuit provide the media with new insight into the inner workings of the secretive company? Is tech industry discrimination against Asians the next big diversity story? There’s a lot to play out here.
What’s at stake in the automation-prediction game. “Intelligent agents” is what we called them, once upon a time, these programs that would perform tasks for us without our telling them. And that sounded great! An agent works for you, right? But now those software bots have joined up with hardware robots, drones, and autonomous vehicles. And if a new report from the Forrester research outfit is right, they might be putting us out of work (The Guardian). Forrester says that, “by 2021, a disruptive tidal wave will begin”: Robots will already have eliminated 6 percent of U.S. jobs — starting with customer service reps and taxi and truck drivers — and more will quickly follow. If true, it’s a grim prognosis: We don’t have nearly enough public resources in place to retrain that volume of unemployed workers or to give them a safety net. But wait! Here’s another report, this one from Goldman Sachs (Bloomberg), that says not to worry: We’ve survived similar industrial transitions before, and they will open all kinds of new opportunities for those who know where to look. We just need the government and other public institutions to manage the transition in a smart way. That’s when you might take a look at our deadlocked Congress and trivialized presidential election — and wince.
Big Ag keeps getting bigger. Bayer’s $56 billion buyout of Monsanto, if approved by regulators, will further consolidate the hold of a handful of agro-chemical industrial giants on the world’s food supply, writes Brad Plumer in Vox. Monsanto holds a big slice of the global seed market, but its trademark Roundup-Ready GMO crops are already losing ground, as bugs gain resistance to the pesticide. The BigCos engaged in this frantic mating dance — not just Bayer/Monsanto but Dow/Dupont, Syngenta/ChemChina and others — are all betting that the bigger they are, the better they’ll be able to influence governments and regulators in their favor. They might be right. But global consolidation only deepens the industry’s commitment to the path of mammoth monoculture, even if that hasn’t worked so well for Monsanto’s pesticide-resistant crops. If we hope for sustainable diversity in our food supply, we may need a little more of it in the companies that support our farmers, too.
