WEBVTT - Fake News and Robots

0:00:03.800 --> 0:00:07.000
<v Speaker 1>Some people worry that artificial intelligence will become so smart

0:00:07.120 --> 0:00:10.959
<v Speaker 1>it will be dangerous, but turns out dumb artificial intelligence

0:00:10.960 --> 0:00:14.640
<v Speaker 1>poses a threat too. I'm Jonathan Strickland, and this is

0:00:14.680 --> 0:00:21.800
<v Speaker 1>tech Stuff Daily. On October, a series of odd stories

0:00:21.840 --> 0:00:25.560
<v Speaker 1>published on the dal Jones newswires. The stories were all

0:00:25.600 --> 0:00:29.680
<v Speaker 1>fake and clearly not intended for publication. The official word

0:00:29.720 --> 0:00:32.280
<v Speaker 1>from dal Jones was that a technical error was the

0:00:32.280 --> 0:00:36.440
<v Speaker 1>culprit and had erroneously published these fake stories. In other words,

0:00:36.680 --> 0:00:40.680
<v Speaker 1>this is a verified case of fake news. What exactly

0:00:40.720 --> 0:00:45.000
<v Speaker 1>was reported and what happened. The stories headline was Google

0:00:45.120 --> 0:00:49.560
<v Speaker 1>Apple joined to create tech giant. Such a headline would

0:00:49.600 --> 0:00:52.720
<v Speaker 1>be truly enormous news in the text sphere. The story

0:00:52.760 --> 0:00:55.320
<v Speaker 1>is a short one, so I'll share it in its entirety.

0:00:55.680 --> 0:00:58.520
<v Speaker 1>Here's how it read. In a surprise move to everyone

0:00:58.520 --> 0:01:01.520
<v Speaker 1>who is alive, Google said it's going to buy Apple

0:01:01.560 --> 0:01:05.160
<v Speaker 1>for nine billion dollars. Google chief executive Larry Page had

0:01:05.200 --> 0:01:09.520
<v Speaker 1>secret talks with the now deceased Steve Jobs to firm

0:01:09.640 --> 0:01:11.959
<v Speaker 1>up the deal. The deal was announced when Jobs as

0:01:12.000 --> 0:01:15.520
<v Speaker 1>will was read in Cupertino, California. The deal, which is

0:01:15.560 --> 0:01:19.160
<v Speaker 1>expected to close tomorrow, gives each Google shareholder nine shares

0:01:19.200 --> 0:01:23.720
<v Speaker 1>of Apple stock. Obviously, Google will move into Apple's fancy headquarters.

0:01:23.920 --> 0:01:28.640
<v Speaker 1>Google employees said, yeah, I bet you can tell through

0:01:28.720 --> 0:01:32.560
<v Speaker 1>context that this article was a joke. Why it was

0:01:32.600 --> 0:01:35.520
<v Speaker 1>stored away on the Dow Jones newswire at all as

0:01:35.520 --> 0:01:38.880
<v Speaker 1>a mystery, But due to whatever reason, the joke post

0:01:39.160 --> 0:01:42.319
<v Speaker 1>escaped into the real world and then it caused a

0:01:42.360 --> 0:01:46.560
<v Speaker 1>tiny amount of havoc for a very short while. Immediately

0:01:46.600 --> 0:01:49.680
<v Speaker 1>after the news published, Apple's stock price jumped up to

0:01:49.760 --> 0:01:53.080
<v Speaker 1>one fifty eight dollars. It took a few minutes before

0:01:53.120 --> 0:01:56.560
<v Speaker 1>the price popped back down to one hundred fifty six dollars.

0:01:56.600 --> 0:01:59.320
<v Speaker 1>According to the site nine to five mac, the culprit

0:01:59.360 --> 0:02:02.760
<v Speaker 1>for the jump asn't gullible human investors. Instead, it was

0:02:02.880 --> 0:02:08.320
<v Speaker 1>high speed trading computer algorithms. That's right. Algorithms, as in

0:02:08.440 --> 0:02:13.440
<v Speaker 1>sets of instructions are monitoring industry news and making corresponding

0:02:13.480 --> 0:02:16.320
<v Speaker 1>trades on the stock market. They do so its speeds

0:02:16.400 --> 0:02:19.000
<v Speaker 1>much greater than any human could hope to match. They

0:02:19.000 --> 0:02:21.600
<v Speaker 1>can also act in large volumes, causing the market to

0:02:21.680 --> 0:02:24.960
<v Speaker 1>surge or crash for a short time. In this case,

0:02:25.160 --> 0:02:28.880
<v Speaker 1>the algorithms responded to the news using natural language processing

0:02:28.919 --> 0:02:32.120
<v Speaker 1>to suss out the most important details that Google would

0:02:32.120 --> 0:02:36.080
<v Speaker 1>acquire Apple for nine billion dollars. The algorithms aren't smart

0:02:36.160 --> 0:02:39.680
<v Speaker 1>enough to detect when news isn't actually real. It acted

0:02:39.720 --> 0:02:42.400
<v Speaker 1>as if the news were legitimate, and the market responded

0:02:42.400 --> 0:02:46.800
<v Speaker 1>accordingly as these algorithms began initiating trades at a frenzied pace.

0:02:47.400 --> 0:02:50.040
<v Speaker 1>This isn't the first time this has happened. On May SI,

0:02:51.040 --> 0:02:54.080
<v Speaker 1>the Crash of two, also known as the Flash Crash,

0:02:54.160 --> 0:02:58.000
<v Speaker 1>happened with a trillion dollar stock market crash. It lasted

0:02:58.000 --> 0:03:02.000
<v Speaker 1>for about half an hour. Frequency trading algorithms were partly

0:03:02.040 --> 0:03:05.080
<v Speaker 1>to blame, though the SEC determined that a stock trader

0:03:05.160 --> 0:03:08.880
<v Speaker 1>initiated the events that escalated in the crash. This one

0:03:08.960 --> 0:03:11.240
<v Speaker 1>was human directed, so we'll give it a pass as

0:03:11.280 --> 0:03:15.200
<v Speaker 1>far as blaming it on dumb artificial intelligence. In twelve,

0:03:15.400 --> 0:03:19.400
<v Speaker 1>a company called Night Capital launched its own trading algorithms

0:03:19.400 --> 0:03:23.080
<v Speaker 1>on the New York Stock Exchange. Unfortunately, someone must have

0:03:23.080 --> 0:03:26.360
<v Speaker 1>missed an important bug during the testing phase of the technology.

0:03:26.600 --> 0:03:29.680
<v Speaker 1>The algorithms began to buy stocks at high prices and

0:03:29.840 --> 0:03:33.519
<v Speaker 1>sell them at low ones, which, as I understand it,

0:03:33.600 --> 0:03:36.360
<v Speaker 1>is generally the opposite of what you actually want to do.

0:03:36.920 --> 0:03:39.800
<v Speaker 1>It took nearly forty five minutes to shut the system down,

0:03:39.880 --> 0:03:42.720
<v Speaker 1>but by that time the company had lost hundreds of

0:03:42.800 --> 0:03:47.200
<v Speaker 1>millions of dollars. Another incident in twelve raised concerns when

0:03:47.240 --> 0:03:50.680
<v Speaker 1>the share price for Pete's Coffee and Tea rose five

0:03:50.720 --> 0:03:54.480
<v Speaker 1>percent immediately after the market opened. That set off alarm

0:03:54.560 --> 0:03:59.360
<v Speaker 1>bells figuratively speaking, and NASDAC halted trading of the company stocks.

0:03:59.760 --> 0:04:03.400
<v Speaker 1>This suspected cause was a high frequency trading algorithm, though

0:04:03.400 --> 0:04:07.440
<v Speaker 1>it wasn't ever proven conclusively. The NASDAC exchange was forced

0:04:07.480 --> 0:04:09.600
<v Speaker 1>to cancel many of the trades that happened in those

0:04:09.640 --> 0:04:12.960
<v Speaker 1>opening moments. On the one hand, these algorithms can help

0:04:13.040 --> 0:04:17.240
<v Speaker 1>traders jump on opportunities with unprecedented speed, which can result

0:04:17.240 --> 0:04:20.480
<v Speaker 1>in large benefits for their clients. On the other hand,

0:04:20.560 --> 0:04:23.840
<v Speaker 1>computer programs can't always tell when things are going south

0:04:23.960 --> 0:04:27.640
<v Speaker 1>and can rapidly accelerate the problem to catastrophic levels, affecting

0:04:27.640 --> 0:04:30.799
<v Speaker 1>not just computer directed actions on the market, but human

0:04:30.839 --> 0:04:33.760
<v Speaker 1>ones as well. In some cases, it might at least

0:04:33.760 --> 0:04:37.120
<v Speaker 1>temporari lease in the company's valuation down the drain or

0:04:37.160 --> 0:04:41.000
<v Speaker 1>through the stratosphere. Typically, these problems when caught early can

0:04:41.000 --> 0:04:43.960
<v Speaker 1>be corrected with some fuss and bother, but without too

0:04:44.000 --> 0:04:48.360
<v Speaker 1>many other serious consequences. But sometimes, such as with Night Capital,

0:04:48.680 --> 0:04:51.599
<v Speaker 1>the problems can destroy a company. That's all for today.

0:04:51.800 --> 0:04:55.520
<v Speaker 1>To learn more about artificial intelligence, how high frequency trading works,

0:04:55.520 --> 0:04:58.840
<v Speaker 1>and the tech behind the financial sector, tune into Tech Stuff.

0:04:59.040 --> 0:05:01.280
<v Speaker 1>It's the twice week the series that takes a deep

0:05:01.320 --> 0:05:04.159
<v Speaker 1>dive on all things tech. That's all for today, and

0:05:04.200 --> 0:05:05.559
<v Speaker 1>I'll see you again soon.