What our results show is that adults do not conform to what the robots are saying. But when we did the experiment with children, they did. It shows children can perhaps have more of an affinity with robots than adults, which does pose the question: what if robots were to suggest, for example, what products to buy or what to think?"

To make AI systems broadly useful for challenging real-world tasks, we need them to learn complex human goals and preferences. One approach to specifying complex goals asks humans to judge during training which agent behaviors are safe and useful, but this approach can fail if the task is too complicated for a human to directly judge. To help address this concern, we propose training agents via self play on a zero sum debate game. Given a question or proposed action, two agents take turns making short statements up to a limit, then a human judges which of the agents gave the most true, useful information. In an analogy to complexity theory, debate with optimal play can answer any question in PSPACE given polynomial time judges (direct judging answers only NP questions). In practice, whether debate works involves empirical questions about humans and the tasks we want AIs to perform, plus theoretical questions about the meaning of AI alignment. We report results on an initial MNIST experiment where agents compete to convince a sparse classifier, boosting the classifier’s accuracy from 59.4% to 88.9% given 6 pixels and from 48.2% to 85.2% given 4 pixels. Finally, we discuss theoretical and practical aspects of the debate model, focusing on potential weaknesses as the model scales up, and we propose future human and computer experiments to test these properties.

Thinking about artificial intelligence can help clarify what makes us human—for better and for worse.

For one thing, the company has, since 2015, proactively sought out candidates from outside traditional programmer pipelines like Stanford and MIT, recruiting through all-women’s coding camps like Hackbright, as well as programs that focus on training black and Latino programmers such as Code2040. Recruiters are trained to look at skills rather than a candidate’s university pedigree. In 2015, Slack worked with Textio, a company that analyzes job descriptions to ensure they appeal to the widest possible audience. (Slack’s job descriptions feature phrases like “care deeply” and “lasting relationships,” which statistically draw more applications from women. Microsoft’s, by contrast, feature words like “insatiably” and “competing.” Amazon’s keywords: “maniacal” and “wickedly.")

Over the last few years the European Union has been working on revising its rules on copyright. But the latest proposal from the head of the copyright committee would deny creators the right to refuse remuneration — the right to share a work without getting paid — which could undermine the use of CC licenses if approved.

I’ve long tracked Facebook’s serial admission to having SIGINT visibility that nearly rivals the NSA: knowing that Facebook had intelligence corroborating NSA’s judgment that GRU was behind the DNC hack was one reason I was ultimately convinced of the IC’s claims, in spite of initial questions.

Zuckerberg said that Facebook doesn’t allow fake profiles, contradicted by Sen. Chris Coons, D-Del., who literally found two fake profiles of himself the day of the hearing. Zuckerberg alleged that users prefer “relevant” ads — by which he means ads targeting users based on Facebook’s surveillance — impossibly speaking on behalf of hundreds of millions of people (and ignoring the millions of people who feel creeped out by ads that are too on the nose). He wrongly stated that Cambridge Analytica was banned from Facebook after the company became aware of the data harvesting; it turns out the firm popped up as an advertiser months later. He claimed users can port their data to other social networks, which is not really true since they wouldn’t be able to contact their friends, the whole point of interoperability. He was super-proud of stating that Facebook doesn’t “sell” its data to third parties — it allows third parties to use any slice of it for a price, making whether the data is “sold” a distinction without a difference.

He claimed that Palantir, the surveillance company chaired by Facebook board member Peter Thiel, didn’t have access to Facebook data; a former Palantir employee said the company did. He waved away suggestions that Facebook couldn’t manage the millions of examples of discriminatory advertising tactics and hate speech and deceptions happening on the site daily, saying that unspecified “AI tools” would fix things later. When asked questions to which he said he didn’t know the answer, Zuckerberg habitually said he would get senators the information later.

Under any normal circumstances, this would have been seen as a disaster for Zuckerberg and Facebook. But Facebook stock jumped nearly 5 percent on the day of the hearing, including while he was in the not-so-hot booster seat. And that’s because the Senate made it very clear that Facebook would be given every opportunity to continue governing itself, with barely any recognition that its power in the marketplace and business model of mass surveillance means that any serious fix changes the fundamental nature of what makes Facebook the profit-center it is now.

With lawmakers and users on the warpath, the photo-sharing app could be Mark Zuckerberg’s way out of the latest data scandal.

SpaceX has a green light from the FCC to launch a network of thousands of satellites blanketing the globe with broadband. And you won’t have too long to wait — on a cosmic scale, anyway. Part of the agreement is that SpaceX launch half of its proposed satellites within six years.

The approval of SpaceX’s application was not seriously in doubt after last month’s memo from FCC Chairman Ajit Pai, who was excited at the prospect of the first U.S.-based company being authorized to launch a constellation like this.

Lawmakers and enforcement officials have been working to crack down on the site, the second largest classified ad service in the country after Craigslist that is used primarily to sell sex.

From the start, the FBI argued that Apple’s new security protections undermined public safety. It appears that the CEAU unit chief sought to use a “poster child” case to force Apple into building tools that would undo those protections. Yet, the premise of the going dark thesis is not nearly as strong as law enforcement would make out. Recent news shows that despite Apple's continuing efforts to secure phones, there are ways to open these locked devices. Forbes recently reported that the Israeli company Cellebrite claims to be able to open all models of iPhones, and that a new U.S. startup, GrayKey, makes similar claims. It seems that we're not going dark—or at least not nearly to the extent that FBI director Christopher Wray, Manhattan district attorney Cyrus Vance Jr., and Deputy Attorney General Rod Rosenstein have claimed.

UN investigators have said the use of Facebook played a "determining role" in stirring up hatred against Rohingya Muslims in Myanmar.

Facebook is using us. It is actively giving away our information. It is creating an echo chamber in the name of connection. It surfaces the divisive and destroys the real reason we began using social media in the first place – human connection.

It is a cancer.

I’ve begun the slow process of weaning myself off of the platform by methodically running a script that will delete my old content. And there’s a lot. There are likes and shares. There are long posts I wrote to impress my friends. There are thousands of WordPress notifications that tell the world what I’m doing. In fact, I would wager I use Facebook more to broadcast my ego than interact with real humans. And I suspect that most of us are in a similar situation.

Case in point: Breitbart News. The only reason Breitbart was able to be economically viable and politically influential is because of net neutrality: if their ability to contact their constituencies was contingent on approval from the establishment they decry, they'd have been strangled in their cradle.

This is true of many things I love and many things I hate: insurgencies rely on free, fair, open infrastructure and suffer under systems of control and incumbency. Boing Boing thrives because of the same free, fair and open network that let Breitbart arise.

What Breitbart doesn't realize is that the weapon they fashion for Trump and his FCC will remain locked and loaded for the administrations that come next: once you erode the principle of net neutrality, then a 2018 Democratic Congress or a 2020 Democratic President could simply turn Breitbart off.

Governments should “start with modest expectations” for manufacturing, says James Manyika of the McKinsey Global Institute, a think-tank. The policies that might help are mostly fairly obvious. Improve education to ensure that engineers and techies are in good supply. Provide more vocational training, along the lines that Germany uses to support its Mittelstand. And develop retraining programmes to refurbish the skills of current or former workers (see this week’s special report).

A measured pace is likely because what is technically possible is only one factor in determining how quickly new technology is adopted, according to a new study by the McKinsey Global Institute. Other crucial ingredients include economics, labor markets, regulations and social attitudes.

The report, which was released Thursday, breaks jobs down by work tasks — more than 2,000 activities across 800 occupations, from stock clerk to company boss. The institute, the research arm of the consulting firm McKinsey & Company, concludes that many tasks can be automated and that most jobs have activities ripe for automation. But the near-term impact, the report says, will be to transform work more than to eliminate jobs.

Automation is happening, and it will bring substantial benefits to businesses and economies worldwide, but it won’t arrive overnight. A new McKinsey Global Institute report finds realizing automation’s full potential requires people and technology to work hand in hand.