monfrere

How is Bing a pile of garbage? It takes some effort to come up with a search for which it or ddg are markedly worse than google. FWIW I still use google.

That's how the BBC does their headlines. Anytime something is in quotes, prepend the headline with "Someone says:". For whatever reason they won't write headlines longer than 6 or 7 words.

I think it doesn't work well for questions with simple answers; it's meant for more open-ended questions. I got much better responses for some random questions I just thought up: "What will AI enable in the future?" "Why are so many depressed?" etc.

But spoken Chinese is quite different from written Chinese, which allows for more economy. Often a single character will be used in writing where a two-character word would be used in speech. And people's names usually can't be determined by the pronunciation; they are defined by the actual characters. If Chinese moved to a strictly phonetic writing system, a lot of culture would have to adapt: conventions around signage, poetry, proverbs, names (this is a big one!) of people and organizations, formal writing, wordplay, etc.

FWIW, I am American, and I would absolutely object if Google were building weapons for the Chinese military. I would stop using Microsoft products if it turned out they supplied weapons for the Russian takeover of Crimea. I would delete my Twitter account if they were found to be building special-purpose propaganda tools to aid Turkey's Erdogan. Etc.

Complying with laws in another country is one thing. Working with a military, which necessarily has implications beyond that country's borders, is another. And of course even here there are different degrees. You can build a general-purpose secure email client and sell it to a country's military, or you can design their bombs. Where the line is I'm not sure, but at some point your activity is inherently violent, inherently adversarial to some fraction of people in the world.

Your discussion of "legality" here is U.S.-centric. Google has offices in Germany, France, Poland, Russia, Turkey, U.A.E., India, and China, to name a few. How would you feel if Google worked on military technology for those countries? Would you point out that it was inevitable that their militaries would seek ways to use A.I.? Would you point out that the Turkish armed forces' activities are permitted under Turkish law? Would you lament the inability of Google employees to separate work from politics?

Google has users in almost all countries, and even our friends in other liberal democracies do not see the U.S. military the same way we do. Perhaps some Google users' family members have even been killed by the U.S. military. This presents a perfectly reasonable business reason (one that has nothing to do with "the personal, the moral, the legal, and the political") for Google to turn down AI drone contracts.

Not everyone thinks that the U.S. drone program has much to do with keeping free nations safe and prosperous.

How senior are you? I would first spend some time hunting for innovative teams within Google to transfer to. There are plenty of boring jobs at Google but also plenty of ambitious small teams working on startup-like projects. Even cute experimental features on big established products (search/gmail/maps/drive/etc.) often could plausibly be the basis for an entire startup, but there are also some pie-in-the-sky projects people are working on if you look around a bit.

I'm surprised Uber is funding general AI research given their reported financial situation.

> I have this theory that they've noticed that the platform is being used less and less, and is why we're getting so much ads mixed with our friends' posts. They're milking the cow before it dies.

Um, no. Facebook remains the most popular app in the United States with a comfortable lead. It's surpassed only by its own Whatsapp and Messenger globally. I think I read that in terms of time spent in an app, Facebook still beats Whatsapp, but I can't find the stat right now. Facebook just crossed 2 billion MAUs this year.

Its advertising revenue grows at close to 50% a year. Obviously part of that is showing more ads per user, but it's also growth in app usage. I'd say the cow keeps sprouting new udders and they are feverishly trying to keep up with attaching the pumps.

This should not be downvoted. One sees the source before the colon more often than after the colon, but it is common to see the source after the colon in some publications. I see several instances (e.g., "Corporate lobbying helped derail border tax: senior Republican") on the Reuters homepage right now.

I believe what she wrote, and I believe you do too, which is why we both keep quoting her. I disagree that either of my imagined scenarios are inconsistent with what she wrote. But thanks anyway for trying to answer my questions.

With complete sincerity and honesty, I think either sequence of events I described could match what Yeoh wrote.

If I said to someone, "You really gotta go, sorry", I might well describe that later to a friend using the words Yeoh used: "I told him he needs to leave."

If a guy took two steps right into my personal space as I was up against a sofa and a wall, I might well describe that as "he pushed himself onto me to the point where I was backed into a corner".

If a guy put his hand on my shoulder and leaned in for a kiss -- I could see that being what Yeoh described: he "made contact to kiss me". (Frankly, that's not the phrase I would use; it seems vague.)

I have experienced similarly persistent (or arguably even more persistent) unwanted sexual advances from other men, but I certainly didn't feel "assaulted".

As a socially clueless gay man, I was hoping someone could help elucidate the most problematic parts of McClure's behavior here. The two factors that stand out for me are

1) McClure controlled the funding of the agency Yeoh was heading, so he was in a position of power

2) McClure is a man and Yeoh is a woman, so McClure can be assumed to be more physically imposing

Would it be correct to say that these two factors are what make McClure's behavior inappropriate? In any other context, I would consider asking someone something a third time after two "no"s to be an example of annoying badgering, but not necessarily harassment.

Oh, right, subfield of the complex numbers. Now sigma is supposed to be an automorphism on that subfield. Which means sigma does take scalar values as arguments, contrary to what stablemap said...

OK, thanks. Your answer has allowed me to follow the paper a little further, though I think your answer may be inconsistent with the other two answers I got. Also I haven't been able to show myself that if the x_i are the roots of an irreducible polynomial, then Q(x_1,...,x_n) is symmetric with respect to x_1,...,x_n, in the sense claimed in the paper without proof.

> σ acts on rational functions like f and g

This is what I was missing. But the paper also says σ is an automorphism of Q(x_1, ... x_n). Which is weird, since I thought Q(x_1, ... x_n) was a subfield of the reals (not a field of rational functions). So I still don't get what's going on.

Sigh, I think I've forgotten more since school than I thought.

> Your sigma is a bijection on rational functions of a, b, not on the evaluations of those functions at particular values of a and b.

I see, this is the key point I was missing. But now I am confused as to whether Q(x_1, x_2) is supposed to be a subfield of the real numbers or a field of rational functions of x_1, x_2.

Can anyone help me understand what is happening at the bottom of page 23 (page 3 of the PDF)?

It says any permutation sigma of x_1, ..., x_n can be extended to a bijection of Q(x_1, ..., x_n) defined by

    sigma f(x_1, ..., x_n) = f(sigma x_1, ..., sigma x_n).

But I don't see how this definition can be consistent. For example, let

    f(a, b) = a - b
    g(a, b) = a/a + b/b = 2

    x_1 = 5
    x_2 = 3
    sigma x_1 = x_2
    sigma x_2 = x_1

Then according to the formula:

    sigma f(5, 3) = sigma (5 - 3) = sigma 2 = f(3, 5) = -2

But

    sigma g(5, 3) = sigma 2 = g(3, 5) = 2

Contradiction?

Troubling. Thanks for sharing. I wonder how the percent of innocents killed by drone strikes compares to the number for non-drone combat.

Is typing speed really the bottleneck when programming? For me, it's definitely either the speed of thought or the time it takes to look up names, interfaces, etc.

First, I would point out that bombings, drone strikes etc. have definitely been the target of criticism from sectors of the left. Although I would agree that the outrage has not reached this scale.

I can think of several reasons. I don't necessarily think they are all good reasons.

1) U.S. bombings are seen as a good-faith effort to stop terrorism, based on intel etc.

The current executive order seems more like a rash decision based on xenophobia and poorly implemented (in the early hours, border officials were giving confused and contradictory info on whether green card holders were affected, etc.)

2) The collateral damage from the bombings, added up over a decade, has probably been hundreds of innocent civilians' lives in some of these countries. This is upsetting but distant.

On the other hand, the collateral damage from this executive order is that 500,000 expats living in the US can't leave the country, or are stranded outside the country. For American "coastal elites", myself included, these are not remote people but our friends, coworkers, professors. Everyone in my social circle seem to have at least a second-degree connection to someone from Iran.

3) The current executive order threatens the standing of the communities "coastal elites" are part of: top American universities, Silicon Valley, the medical community, etc. These institutions are world-class due in no small part to how welcoming they are to foreign talent. That could change if the smartest around the world start heading to universities, hospitals, and companies in Canada or Germany instead...

I think you may misunderstand what the right to be forgotten usually refers to. It's not about a user's "data" such as emails or online comments; it's usually about news articles that mention a person.

Surely you would agree that no matter what French law is, there is no way a French judge should be able to force, say, an American company to not display links to American websites (which happen to mention a French person) for American users?

(disclosure: I work for Google, but not on anything related to RTBF, and my views are my own)

> The Times did the exact opposite and had a political campaign proofread stories before publication.

This is false, or misleading at best, if you're referring to the Mark Leibovich profile of Clinton. He sent an off-the-record transcript the campaign asking whether it could be put on the record, and the campaign only allowed him to publish part of it. Not ideal of course, but I'm not sure how you would avoid this, unless you think off-the-record interviews should be banned entirely.

I read the leaked emails referred to in these articles, and I don't see much that is troubling about NYT's behavior.

This is just how journalism works: the reporter has to request/negotiate access to the source. The PR folks will obviously try to spin everything positively to the reporter; that's their job. Afterward, the reporter writes up a story as objectively as possible. When some part of the story happens to be flattering, the PR people will obviously high-five themselves. When some part of the story is negative, the PR people will be disappointed and try to get the reporter to see their side.

See this article [1] from Mark Leibovich for an NYT reporter's account of one of the issues addressed in your links.

[1] http://www.nytimes.com/2016/10/14/magazine/anatomy-of-a-medi...

You might look into api.ai.

I took this class a couple of years ago. It's outdated and overrated imho, with nowhere enough mathematical rigor to be useful. For example, the discussion of support vector machines (and classification in general if I recall correctly) was limited to two dimensions so that you didn't need linear algebra. The class also spends a lot of time on problems like path finding that you should be able to solve with your standard CS algorithms toolkit or just "logic" rather than needing to reach for anything that deserves the name "artificial intelligence" (at least today). Prof. Winston furthermore spends way too much time on vague truisms that may sum up or organize what's in his brain but aren't helpful to students. ("What if the answer doesn't depend on the data at all? Then you've got the trying to build a cake without flour.")

I hate to dismiss something as ambitious as this course and just tell people to blindly follow trends, but my honest advice would be to just skim these notes if you're interested and go take a normal machine learning course instead.

Not a lawyer, but I always thought maybe someone could take your code, modify it to produce variations A, B, C, the three major use cases of your code, and slap patents on those modified versions of your code. So then anyone who wants to use your code with any of those three use cases in mind (i.e., most people who would want to use your code) would then likely be in trouble.

Or something like that.

A second issue would be one the author of the article seems worried about, which is that without a warranty-free clause people might sue you for bugs in the code. If you're distributing your code, with a line like "free for everybody to use," there may or may not be implicit quality guarantees, according to the article.

Calm down, it's a press release written by PR professionals, not the researchers. Others here have linked to the paper. But I would agree with you that code in academic research needs to be shared more. I haven't searched very hard, but if the code for "milk" is public, these researchers don't seem to have put it in an obvious place.

You should never have to use more than 9 moves for 3x3. Hitting the same square twice has no net effect, even if the two moves on the same square are separated by other moves.