Big tech companies now trade at one of the smallest premiums in history.

So the “tech is over” crowd and the “end of the beginning” crowd are, perhaps, telling the same story: Tech stocks have fallen, because the media mountain has been scaled. Now the largest tech companies aren’t standing at a pinnacle; they’re plotting on a plateau. The challenge of owning the entire life cycle of our spending habits—that’s the real summit. And it’s just up ahead.

A rebours des grandes études, de l’université d’Oxford à celles de cabinets de conseil, comme l’Institut Roland Berger, qui quantifient les destructions d’emplois à venir, le chercheur déploie dans En attendant les robots une tout autre réalité : les algorithmes et les promesses de l’IA n’ont pas effacé la main de l’homme, et encore moins son doigt ! Derrière Uber, Facebook, Siri, derrière les milliards de recherches et requêtes des réseaux sociaux, des millions d’êtres humains à travers le monde créent, affinent, trient, corrigent. Et aident la machine à apprendre et à mieux fonctionner. Ces «millions de micro-tâcherons filtrent des vidéos, étiquettent des images, transcrivent des documents dont les machines ne sont pas capables de s’occuper», explique le chercheur. De «l’intelligence artificielle largement faite à la main», dit-il joliment. C’est ce qu’on appelle le «travail du clic». Notre imaginaire technologique est peuplé de blouses blanches et de types sympas en jean qui font tourner des start-up ? En fait, derrière chaque col blanc, œuvre une armée de cols bleus. Le grand bluff technologique !

In November, the Commerce Department released a list of technologies, including artificial intelligence, that are under consideration for new export rules because of their importance to national security.

Technology experts worry that blocking the export of A.I. to other countries, or tying it up in red tape, will help A.I. industries flourish in those nations — China, in particular — and compete with American companies.

“The number of cases where exports can be sufficiently controlled are very, very, very small, and the chance of making an error is quite large,” said Jack Clark, head of policy at OpenAI, an artificial intelligence lab in San Francisco. “If this goes wrong, it could do real damage to the A.I. community.”

In 2018 the blockchain/decentralization story fell apart. For example, a study of 43 use cases found a 0% success rate. theregister.co.uk/2018/11/30/blo…

Let's talk about some mistaken assumptions about decentralization that led to the blockchain hype, and what we can learn from them.

OP here. A small clarification: there are some legitimate criticisms of the Register piece that I cited in the first tweet, but I merely cited it as an example of why I think the hype is calming down. The arguments I make are independent of that piece; the limitations I point out have always been there, rather than something new that happened in 2018.
I also wanted to add a couple of points that I didn't get to in the Twitter thread.
Economics. Blockchain technologists seem to overestimate the extent to which new insights in economics are needed to understand cryptocurrencies and blockchains, as opposed to applying basic principles from economics and game theory. For example, a recent paper shows that thinking about miners and attackers in terms of stock and flow exposes important limitations of the security of Proof of Work. [1] I learnt of many other such examples at a recent conference on the economics of blockchains. [2] So I think a lot of the "cryptoeconomics" hype is misplaced.
Privacy. It's often taken for granted that decentralized architectures will improve privacy. This seems obvious given everything we've learnt about Facebook, but a better way to think about it is that decentralized systems exchange one set of privacy problems with another. I coauthored a paper a few years ago skeptical of the "decentralization ==> privacy" story in the context of social networks [3], but I think many of the arguments in that paper apply to blockchain/dApps that are being built today.
[1] http://faculty.chicagobooth.edu/eric.budish/research/Economic-Limits-Bitcoin-Blockchain.pdf
[2] https://bfi.uchicago.edu/events/cryptocurrencies-and-blockchains
[3] http://randomwalker.info/publications/critical-look-at-decentralization-v1.pdf

Yeah, deleting Fb is cool, but have you tried regulating it?

I admire the #DeleteFacebook impulse. But it takes so much coordination and scale to work, and most users are not American. And it also owns WhatsApp and Instagram.

It’s too much to take on as consumers.

This is why we have governments.

On immigration, the agency’s record is mixed. The USDS began working on the green card queue last year, noticing that the USCIS software classified routine issues such as a change of address just as seriously as, say, a green card holder being arrested. The team created a program to sift renewal applicants by risk category, feeding the change-of-address types into a faster lane. The digital service says the software has helped cut the overall USCIS backlog by about two-thirds.
Yet the reality on the ground hasn’t improved for everyone, says Denyse Sabagh, an immigration attorney at law firm Duane Morris LLP. “It seems like the agency is geared now to deny cases across the board,” she says, and it’s no clearer why her clients’ renewal requests are taking so long to process, with some basic cases that would once require a few days taking five months and complex ones taking years.

The only advantage Uber might have achieved is taking advantage of its drivers’ lack of financial acumen — that they don’t understand the full cost of using their cars and thus are giving Uber a bargain. There’s some evidence to support that notion. Ridester recently published the results of the first study to use actual Uber driver earnings, validated by screenshots. Using conservative estimates for vehicle costs, they found that that UberX drivers, which represent the bulk of its workforce, earn less than $10 an hour. They would do better at McDonald’s. But even this offset to the generally higher costs of fleet operation hasn’t had an meaningful impact on Uber’s economics.

But, you may argue, Uber has all that data about rides! Certainly that allows it to be more efficient than traditional cabs. Um, no. Local ride services have backhaul problems that no amount of cleverness can remedy, like taking customers to the airport and either waiting hours for a return fare or coming back empty, or daily urban commutes, where workers go overwhelmingly in one direction in the morning rush and the other way in the evening. Similarly, Uber’s surge pricing hasn’t led customers to change their habits and shift their trips to lower-cost times, which could have led to more efficient utilization. If Uber had any secret sauce, it would have already shown up in Uber revenues and average driver earnings. Nine years in, and there’s no evidence of that.

Typically, when workers speak out or organize, it’s because they want higher wages, better working conditions or stronger job security. Those aren’t your problems. You probably work manageable hours in tastefully decorated buildings with free food, ergonomic desk furniture and plentiful amenities. Your compensation is generous. And you’re in little danger of being fired or retaliated against, especially if you’re part of a large group.

The possibilities for you are so much greater. What if Facebook employees publicly took their executives to task for neglecting the real-world violence their products are causing in places like Myanmar and Sri Lanka? What if Google or Twitter employees threatened to walk unless their executives took major action against radical extremists and hate speech? What if Apple employees insisted that the company stop parking billions of dollars in offshore tax shelters, or Amazon engineers threatened to quit unless the company paid its warehouse workers higher wages?

Facebook employed a Republican opposition-research firm to discredit activist protesters, in part by linking them to the liberal financier George Soros. It also tapped its business relationships, persuading a Jewish civil rights group to cast some criticism of the company as anti-Semitic.

“So, there are online communities that exist under the umbrella of what we know as the Red Pill, which are men connected by common resentments against women, immigrants, people of colour,” she explains. “What I was surprised to find was the extent to which they are using ancient Greek and Roman figures and texts to prop up an ideal of white masculinity.”

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.