First it was the failure of robot vacuums. Then Tesla's autopilot slammed into a truck in broad daylight. Now boosters of AI have to deal with another fiasco. The Washington Post reports that Facebook's shiny new algorithm for selecting trending topics is severely broken:

As part of a larger audit of Facebook’s Trending topics, the Intersect logged every news story that trended across four accounts during the workdays from Aug. 31 to Sept. 22. During that time, we uncovered five trending stories that were indisputably fake and three that were profoundly inaccurate. On top of that, we found that news releases, blog posts from sites such as Medium and links to online stores such as iTunes regularly trended.

“I’m not at all surprised how many fake stories have trended,” one former member of the team that used to oversee Trending told the Post. “It was beyond predictable by anyone who spent time with the actual functionality of the product, not just the code.”

That's pretty embarrassing. However, I did a little back-of-the-envelope calculation, and at this rate I figure that Facebook's algorithm won't catch up with Donald Trump until 2038. Say what you will about Moore's law and neural nets and all that, but humans are just fundamentally better at bullshit and lies than computers.

I know this is pretty meaningless, but TPC has released its latest—and presumably final—analyses of the Trump and Clinton tax plans. There's nothing surprising in them: Trump's plan features gigantic tax cuts for the wealthy and Clinton's features big tax increases on the wealthy. For the record, here's how they pencil out:

TPC estimates that Clinton's plan would raise $1.4 trillion over ten years and reduce the federal deficit by $1.6 trillion. Trump's plan would cost $6.2 trillion and increase the deficit by $7.2 billion. Needless to say, that's just for the tax plans themselves. Clinton's overall budget proposal would be roughly revenue neutral once you account for her spending proposals. Likewise, Trump's plan might be slightly less of a deficit buster if he cuts spending somewhere—though he's ruled out most areas of the budget for spending cuts and hasn't identified any specific cuts in the few areas left.

Assuming that Trump has no sizeable budget cuts in mind, his plan would increase the national debt from about 90 percent of GDP to 115 percent of GDP by 2026. In other words, it's fiscally conservative, using the modern definition of the term.

How much does Donald Trump earn in a year? During Sunday's debate he said he made $694 million last year, but that's a lie. That figure represents revenue: that is, how much his various enterprises took in. You have to subtract expenses, interest on loans, and so forth to get to net profit, which is what gets passed through to Trump as personal income. Any child running a lemonade stand knows this. Selling $5 worth of lemonade doesn't mean you made any money if you also had $5 in expenses for sugar, lemons, cups, and payoffs to your little brother to go play somewhere else.

So then: how much did Trump's enterprises actually earn? Nobody knows, but MoJo's Russ Choma has uncovered a revealing nugget. Trump's records show that he's highly dependent on revenue from his golf courses: they pulled in 42 percent of his total revenue, or $296 million. But how much did they earn in profit? Nobody knows that either. But we do know how much two of his golf courses earned:

Trump's FEC financial form noted that his two Scottish golf courses earned him a combined $23 million in "golf related revenue" last year....But the public filings the courses submitted in the United Kingdom tell a much different story....When interest, depreciation, and currency exchange losses are factored in, Trump's Turnberry course lost over $2 million in 2015. And the corporate filings in the United Kingdom show that Trump's Aberdeen course lost about $1.6 million.

That means that Trump's reported income on the FEC financial disclosure forms regarding just these two projects is $26 million more than what they actually made. If these courses are representative of Trump's overall finances—$23 million in "golf related revenue" is really a $3 million loss—his declared $296 million in total "golf related revenue" may well be highly overstated.

"Highly overstated" indeed. In fairness, Trump's more established courses are likely to be more profitable than his newer Scottish ones, but even if his golf courses are well run, they'd be lucky to pull in net earnings of 10-15 percent. If Trump paid too much for them and has run them poorly—which seems likely—his golf earnings might be more like 3-4 percent. That's in the neighborhood of $10 million or so. Or maybe zero. Who knows?

There's no question that Trump is a rich man. Even independent observers figure his net worth is around $3 billion. And with that kind of wealth, he almost certainly makes a lot of money even if he manages his businesses poorly. But $694 million? Not a chance. I'd wager that his income is in the range of $50-100 million.

While Donald Trump is doing his best to immolate the Republican Party, Hillary Clinton is...making proposals to assist people in need. Boring! This is, however, the kind of thing we typically expect from presidential candidates, and Clinton's new proposal is on a subject that's been close to her heart for her entire career: helping children.

In particular, she's proposing a two-part change in the child tax credit. First, instead of kicking in at $3,000, it would kick in at $0. This would help people in the deepest poverty. Second, for families with children four and under, it would max out at $2,000 instead of $1,000. The chart below shows what families would get per child compared to the current CTC (in blue):

The mainstream press pretty clearly couldn't care less about this, and I suppose that's hardly surprising given the Hindenburg-like dimensions of the meltdown of one of America's two major parties. Still, surely it deserves a little bit of attention?

UPDATE: My original chart was wrong. The current child tax credit starts at $3,000 and increases by 15 percent of income until it maxes out. The Clinton credit starts at $0 for children of all ages, but increases more steeply for children four and under. Apologies for the error.

A new Vox poll exposes the love-hate relationship that millennials have with their smartphones:

A slight majority of those under 45 say they agree with the statement “the ability to be constantly connected to the internet with a smartphone can make me feel stressed out.” In contrast, only a quarter of those over 65 agreed with the statement. Seventy-eight percent of people under 30 found the constant connectivity of their smartphones distracting.

I imagine this feeling of being stressed comes from a feeling that you have to stay absolutely current about everything. Every alert might be something important. Donald Trump just said something stupid! Val and Kim are moving in together! This is the cutest kitten ever! Even if it's just a quick OMG, you feel like you have to participate.

I sympathize. The difference is that as a blogger, my time scale for being current is measured in hours or so. There's a certain amount of stress in that, but at least it's limited to one thing (political news) and doesn't require me to literally respond within minutes. Social use of smartphones is different. A text requires immediate response, usually within seconds or minutes—which is sort of ironic since one of the benefits of texting is supposed to be that it's asynchronous. Technically it is, but in real life hardly anyone treats it that way.

But I imagine that things will all sort themselves out. Our phones will get ever smarter, and eventually our AI avatars will just respond for us. At some point, most communications will just be our smartphones chatting with each other while pretending to be people. That way, we stay in the loop, but we can catch up with things later—and hope that our smartphone didn't make any horrible social faux pas. Welcome to the future.

Ladies and gentlemen, Republicans are fleeing from Donald Trump en masse. The Republican Party is in turmoil. But where other people see chaos and doom, Donald Trump sees an opportunity:

Thus was born today's Twitter meme:

Newsweek reporter Kurt Eichenwald reports on the latest from Donald Trump:

At a rally in Wilkes Barre, Pennsylvania, Trump spoke while holding a document in his hand. He told the assembled crowd that it was an email from Blumenthal...."This just came out a little while ago," Trump said...."He's now admitting they could have done something about Benghazi," Trump said, dropping the document to the floor. "This just came out a little while ago"

Ah, Sidney Blumenthal, the unkillable Rasputin of the Clinton family. Conservatives sure do have a way of somehow putting him at the center of every scandal. This time, though, Blumenthal has an ironclad alibi: he never said this. He did email a Newsweek article to John Podesta—which was hacked and released by WikiLeaks a few days ago—but that's all. It was Kurt Eichenwald himself who said this stuff, not Blumenthal.

So how did Trump make this mistake? According to Eichenwald, the only news organization that reported this was Sputnik, a Russian controlled news agency:

This is not funny. It is terrifying. The Russians engage in a sloppy disinformation effort and, before the day is out, the Republican nominee for president is standing on a stage reciting the manufactured story as truth. How did this happen? Who in the Trump campaign was feeding him falsehoods straight from the Kremlin?

Who indeed?

News you can use from Aaron Carroll:

Perhaps no one in the United States has spent more time investigating the occurrence of bacteria on public surfaces than Charles Gerba.

According to Carroll, Gerba's research tells us that it's just fine to eat food that you've dropped on the floor. This sounds suspiciously like motivated reasoning to support the stereotypical male point of view, and I'm a little curious to learn what Mrs. Carroll thinks of this. I suppose we'll never know. In any case, the argument here is that your average floor is no more germy than any other surface in your house, and less so than many. Kitchen floors, for example, have about half the bacteria of kitchen counters.

That's all fair enough, but what about ordinary old dirt and dust? My kitchen counters have almost none of that. My kitchen floor has lots, thanks to the fact that I walk on it, the cats walk on it, the dust accumulates until I vacuum it, and so forth. It may be that dirt and dust aren't likely to make you sick, but it's still a little disgusting to have it all over your food. Or am I being a little too fastidious here?

Of course, it also depends on the food item. If a peanut M&M fell on the floor, I'd have no qualms about rubbing it clean with my shirt and then eating it. But a leftover piece of chicken? Probably not.

I wonder what Donald Trump would think of all this? He's a famous germaphobe, but he also apparently thinks that fast food is safer than other foods because it's highly processed and standardized. So what would he think about an M&M that fell on the floor?

UPDATE: Mrs. Carroll speaks!

A new paper from a trio of Fed researchers suggests that our recent sluggish growth is mostly a result of demographic changes and technological slowdown. The retired share of the population has increased, which means the working share of the population has decreased. Since workers are the ones who produce goods and services, it makes sense that GDP growth will slow down in an economy with fewer adults of working age. Ditto for an economy in which technological progress is slackening.

I've pointed out the same thing before in the case of Japan, and it makes sense. But how about in the US? The easiest way to see the rough shape of the river is to simply look at GDP per working-age adult. That eliminates most of the demographic issues. When you do this for the US, you get a trendline that still shows a decline in GDP growth: it's down by about one percentage point since 1978.

You can also look at total factor productivity, which gives us an idea of the effect of technological change that's independent of demographics. Over the past 60 years, it's been pretty flat.

Both of these are volatile series, so take them with a grain of salt. That said, productivity hasn't changed much, but GDP per working-age adult has steadily decreased anyway. This suggests that neither demographics nor technological progress really explains things. So what does?

NOTE: This bit of amateur economics was made possible by a grant from the Committee to Prevent Endless Blogging About Donald Trump. The author thanks them for their generosity.

This is only one poll, and the sample size is small. Still, it's the well-respected WSJ/NBC poll, and it suggests the possibility of unprecedented doom for the GOP in November:

In the new survey, Mrs. Clinton jumped to an 11-point lead over Mr. Trump among likely voters on a ballot including third-party candidates, up from 6 percentage points in September....The weekend survey found signs of women moving away from Mr. Trump. Mrs. Clinton’s advantage among women increased to 21 percentage points, from 12 points in the September Journal/NBC Survey. Mr. Trump retained a small, single-point advantage among men.

Eleven points! Among women, Clinton is now 21 points ahead, up nine points since the previous poll. This polling was done over the weekend, after the Pussygate tape was released but before Sunday's debate.

In other words, it might get even worse. In fact, since the rumor mill suggests that more videos of Trump are coming over the next few weeks, it probably will get worse. Trump seems to think that a press conference with Paula Jones will turn this around, but that's beyond crazy. Republicans are already jumping ship to save their own skins, and polls like today's will feed the panic. Soon Trump will have nothing left but the Old Confederacy—a fitting end for a racist, misogynistic, xenophobic creep.