Cyber journalist vs bio journalist: draw for now
In May 2015, Scott Horsley (Scott Horsley), a NPR correspondent in the White House and a former business journalist, defiantly challenged the Wordsmith writing algorithm of Automated Insights. “We wanted to know,” writes NPR, “how our best pen will show itself against the machine.” Since NPR is a radio, the bio journalist was trained enough to write quickly. Under the terms of the competition, both opponents had to wait for the publication of the financial report of the network of cafe Denny's and - go ahead. And Scott had an advantage - after all, he was Denny's regular customer. He even had a favorite waitress there, Genevieve, who knew his order by heart: a bacon and egg sandwich. Did not help. Although ... how to judge.
The robot managed in two minutes, Scott Horsley - in about seven. NPR publishes both notes and offers the reader a kind of Turing test - to determine which text is generated by the robot, and which by the person.
The robot wrote, of course, the left note. It obviously has a higher density of figures, and it is drier in style. Whereas Scotus, remembering whether the menu, or Genevieve, let the lyrics, a little optional for the financial report. For example, he inserted the phrase: "Sales growth shows that consumers are not averse to opening their wallets for pancakes, eggs and potato pancakes."
Formally, the dictionary of the robot is more, because it must include the entire vocabulary of the language (this is more than 1 million words for the English language). The dictionary of an English-speaking person can be up to 100 thousand words. But the robot must use the most frequent words, which makes its language "drier". In addition, the dictionary of this robot is also limited to financial specialization. He couldn’t even think that it was possible to use culinary or sports names (grand slam) in the financial report. Man, on the contrary, is not limited by frequency and is free to use arbitrarily rare and arty words, expanding the context and imagery. Moreover, a man-writer is a writer because he uses original images. Robot for the financial report is simply not necessary.
“But that could change,” this may change, writes NPR. If the owner decides to feed the Wordsmith arrays of more relaxed NPR texts and slightly tweak the algorithm, then the “Verbal smith” will quickly rebuild his vocabulary, more precisely, he will expand the boundaries and principles of its application. This is all customizable.
Who won the competition? The robot wrote faster, more efficient. Cattle Horsley, however trite it may be, is slower, but more humane. The audience for this post is the financiers. Whether the lyrical insert about wallets and pancakes is valuable for them? As long as readers are still humans, not robots, then, yes, probably, valuable.
In general, a draw. Although two minutes to seven ... For radio, for financial market news, this can be critical.
The Esenin competition of a foal with a locomotive was also conducted in an academic environment. In 2012, Christer Clerwall, a professor of media and communications at the University of Karlstad, Sweden, asked 46 students to read two sports reports — one written by a robot and one by a person. The human note was reduced to the size of a robot, but the robot note was slightly corrected by the editor: heading, lead, first paragraphs - as the editor usually does in the media. Students were instructed to evaluate the materials according to a number of criteria: objectivity, trust, accuracy, tediousness, interestingness, clarity, pleasure to read, usefulness, integrity, etc.
The results showed that one note won in some parameters, the other - in others. The human text received more points on the criteria of "well written", "pleasure to read." The robotic text, too predictably, received more points on the criteria of "objectivity", "clear description", "accuracy", etc. That is again a draw.
But the most important thing that revealed the study of the Swedish professor, - that the differences between the average text of a bio-journalist and the average text of a cyber-journalist are insignificant. This is a critical factor for assessing the future and already present robot-journalism. Cyber skeptics all the time say that a robot cannot write better than a human. But this is the wrong approach. “Or maybe the robot’s note doesn’t have to be better? What if it’s just enough good? " - Professor Klivalval shares his concerns with Wired.
Robot journalism, the third threat
The Internet has released private authorship. Millions of people themselves inform each other about everything in the world. What is the worst - for free, but with great desire. Yes, the Internet is full of rubbish, but we consume information carefully selected for our interests. Content on the web is not filtered before publication, but after - during distribution, thanks to a viral editor. As a result, the old media are deprived of a monopoly on the formation of the agenda. So business will not be limited to death of newspapers. The Internet threatens the old media not so much with the transition from paper to figure as with the involvement of the audience in the authorship.
Another threat to traditional journalism is corporate media and other content marketing. Corporations, too, got the opportunity to authorship. So, they need less traditional media as an intermediary. Corporations can now themselves.
And if among amateur authors, content “improves” through collaboration (viral editor), then corporations improve their media coverage because of competition for public attention. In the media arms race, they lure professionals from the media, apply innovations and, most importantly, move from direct advertising to socially significant topics. After all, brands need an audience: advertising only spends the audience, and the content is able to collect it. And although to the general public these processes are not very noticeable, but the media infrastructure of corporations, developing, causes traditional media no less damage than the blogosphere.
However, the blogosphere and corporate journalism are at least made up of people. On the unfortunate journalism comes the third threat, soulless, inhuman. If the blogosphere deprives the media of a subscription, corporations are deprived of advertising, then the algorithms threaten to take away the profession.
More recently, the topic of robot-journalism was strongly foreign. And in fact, those who are familiar with the life of the Russian editions, it is quite difficult to imagine the news algorithms in their everyday life. But at the end of October, Yandex announced that it was creating a news agency where robots would write news. Among the editors, exhausted by the struggle with the Internet and losses, this topic has caused a new bout of vague anxiety and, of course, rejection. “Well, okay,” say those who are a little familiar with the problem, “robots will someday write sports notes, financial analytics or weather reports. But they are incapable of anything more.”
Invalid rating. Robots are not something that "someday will" write about the weather, finances or sports - they are already doing it now, and in mind-blowing volumes. Accordingly, the question is whether they will be “capable of more.” Immediate answer: yes.
Earthquake note and tectonic shifts
This earthquake entered the history of journalism. On March 17, 2014 at 6.25 am, The Los Angeles Times journalist and programmer Ken Schwencke was woken by tremors. He ran to the computer, where in the publishing system he was already awaited by a note written by his Quakebot algorithm. Ken ran his eyes through the note and pressed the "Publish" button. So LAT became the first media to write about an earthquake - 3 minutes after the jolt. Robot journalist ahead of their bio-colleagues.
Here is a note on the website of the BAT, supplemented an hour after the second push. It is signed with the name Schwenke, but at the end it is stated: "This post was written by an author."
The Quakebot algorithm, written by Ken Schwenke, had already existed for two years. Quakebot is connected to an American geomonitoring tape. From it, he instantly takes the source data - the place, time, amplitude, compares these data with previous earthquakes in the area, thereby automatically determining the historical significance of the event. The data is placed in a suitable template, and the note is ready. The robot loads it into the publishing system and sends a notification to the editor. On Pulitzer, the note clearly does not pull, but the editor has a finished publication a few minutes after the event. It remains to add that earthquakes are included in the list of hot news in Los Angeles. The Los Angeles Times website has a whole section of Earthquakes, which is replenished with a robot-yunkor.
In the same Los Angeles Times, the robot also has another rubric - the criminal chronicle of the Homicide Report, and since 2007 (!) Of the year. As soon as the police coroner posts a report about someone's violent death to his base (it is obviously public), the robot takes all the data, puts it on the map, categorizes by race, gender, cause of death, police participation, etc. ., and publishes a note with a simple sample wording in the tape. Further, if the event deserves it, the journalist gets the information and appends a detailed criminal note. And if not, then no.
Media critics point out how much the robot - crime reporter has changed the newspaper style of coverage of murders. If earlier journalists wrote only about murders with resonant potential, the robot reports absolutely all murders. Accounting for all statistics, and not selected cases, allows you to see on the map the "density" of murders in different areas, both in general and by category (gender, race, etc.). This visualized statistics generates secondary content, missed by the “writer’s” journalist. Such a murder card gains additional value, for example, for the real estate market.
It remains to add that this robot covers with its criminal chronicle a territory with a population of 10 million people - about as many live in Sweden or Portugal. Of course, a bio-journalist simply could not make instant statistical comparisons on such a scale.
Analyzing these cases, media critics, trying to be radiant with optimism, note that the robot only helps the journalist - collects the invoice, gives the initial processing. "Freeing" thereby a journalist for creative research. It really is. A robot - a crime reporter hardly writes at all, Quakebot writes template notes. Good helpers.
More difficult to deal with financial and sports robot-journalism.
Here is a note written by the Wordsmith algorithm and published by the Associated Press (read optional, estimate the depth).
Alcoa swings to 2Q profit of $ 138M, revenue flat
Jul. 8, 2014 10:32 PM EDT
NEW YORK (AP) - Alcoa Inc. (AA) 2 weekly profit of $ 138 million, reversing a year ago, loss analysts' expectations.
It has been shown that it has been an engineered-products business for its aluminum-smelting segment.
It has been declining in the past few years. Two weeks ago, the British jet engine component maker Firth Rixson announced its $ 2.85 billion acquisition.
CEO Klaus Kleinfeld said the latest results show that the strategy is working. “We’re changing the portfolio,” he said on a conference call with analysts.
Alcoa has been closing down for business. However, you can climb up 2.4 percent from a year earlier.
The company left unchanged its forecast of a 7 percent increase in aluminum demand this year.
Alcoa said second-quarter net income was 12 cents per share. In the year-ago quarter it lost $ 119 million, or 11 cents per share.
The company earned 18 cents per share. The average estimate of analysts surveyed by Zacks Investment Research was 13 cents per share.
It was compared with $ 5.85 billion compared with $ 5.85 billion a year ago. That beat $ 5.63 billion, according to Zacks.
Before the report, Alcoa shares rose 11 cents to close at $ 14.85. They were up another 16 cents, up to $ 15.01, during after-hours trading.
While the Standard & Poor's 500 index has climbed 6.2 percent. The shares have climbed $ 7.04, or 90 percent, in the last 12 months.
The note was compiled in less than 1 second. The robot "ate" the invoice of the report, made the necessary comparisons on the market and in dynamics, and then generated a very detailed and fairly collapsible text.
Much more impressive is the fact that in 2014, the Verbal smith published 3,000 financial notes per quarter to the Associated Press. This is about 10 times more than what AP's own journalists wrote before. By the way, robots are already taking over a financial clearing in leading media. Another market leader in generative journalism (let's call it that), the company Narrative Science, supplies the services of its robot Quill to the portal Forbes.
In order to realize the gaping depth of this abyss, let’s look into its sports section. Here is a fragment of a report on a baseball game of a children's league, written by The Stats Monkey algorithm from the stable of the same Narrative Science.
Friona fell 10-8 on the Boys Ranch on Friday at Friona despite racking up seven hits and eight runs. It has been shown that it has been a day of the day for the Hunter Sundre, which went 2-2 against the Boys Ranch pitching. Sundre singed in the third inning and trip in the fourth inning ... Friona piled up the steals, swiping eight bags in all ...
The trick is that The Stats Monkey uses baseball slang. And that is not all. Stats of children's games are introduced by parents in a special application on the iPhone right during the game. And fans - relatives of small baseball players - get a detailed report on the match even before the rivals on the field have completed an exchange of handshakes. Of course, such reports, whatever their stylistic masterpiece, are more important for these fans than super bowl reports.
In 2011, the algorithm wrote 400 thousand reports for the children's league. In 2012 - 1.5 million. For reference, there were 35,000 journalists in the United States that year, who, of course, would not cover children's league matches for any price-gambling. Another aspect of robot journalism is that algorithms can provide a service to journalism where a human journalist does not come down because of the small scale.
Digitizing the sport itself opens up new horizons for robot sports journalism. As in 2012, Steven Levy noted in the Wired article with examples of baseball reports that sports leagues weigh every inch of the field (and now the player) with cameras and tags. All the most unthinkable statistics collected by the computer: the speed of the ball, the range of the throw, the emphasis of the foot, the angle of inclination - all possible telemetry. A well-trained robot will easily notice that the pitcher's throws began to weaken slightly, or, say, he began to deviate to the left before the batter hit. Is this information important? It is important, but the human journalist will not notice it. In the old sports journalism it is simply not there - as there is in the old criminal journalism an interactive map with a density of murders.
In other words, the robot already many times exceeds the person in work with data, in speed, in coverage. But can he beat a man in style?
Two arguments about the unsuitable abilities of the robot
The main argument against the future dominance of robotic journalists is related to the inability of the machine to be creative. In fact, it breaks down into two theses: a robot cannot invent as a person, a robot cannot write as a person.
1) The robot can not invent
Yes. Indeed, serendipity - the ability of a person to make an accidental discovery "for nothing" (an apple fell on his head) looks like an occasional flash of a chain of excited neurons. A similar phenomenon is a eureka, a sudden insight, logically not deducible. Therefore, creative "tydydzh" will always remain the prerogative of man. Behind creativity, heuristics and other serendipity is a person's ability to be proactive, or will. The actions of the robot are fundamentally predetermined by the algorithm.
Therefore, skeptics will say, the robot will not be able, for example, to see a sensation in a series of similar events, as the human editor does. And even more so the robot will not be able to decide make a sensation like the people-editors do, snatching some event from the same type series.
But. But what if a robot can do something else that a human is not capable of? First of all, we are talking about cross-analysis of any size database and searching for correlations. For example, a cross-analysis of consumer and political preferences will suddenly reveal that the owners of the red Zhiguli regularly vote for Bush. A person would be knocked out right there, because a causal explanation must be found. And in the world of big data, causality that restrains our mind may be unnecessary. A robot has the opportunity to find fantastic, sometimes fantastically influential correlations for marketing, politics or journalism that permeate the world, but we don’t see them.
And what if the ability of the algorithm to detect correlations compensates or even replaces any serendipic abilities of a person, such as instinct, insight, or even a sense of humor? The texture obtained from Big Data may be no less interesting and irrational than creative human experiments.
2) The robot has no feeling (style)
Yes. Это верно, у робота нет целеполагания сделать красиво, а если бы и было, то что это - "красиво"?
Но. Градусником красоты для робота может стать человек. Если нельзя обсчитать красоту, то можно обсчитать реакцию людей на "красоту". Вообразим, что в память робота закачаны все тексты и все реакции людей на них: лайки, перепосты, комментарии, переходы. Роботы уже сегодня способны таким образом вычислять привлекательность заголовков, тем, ключевых слов и т.п. Редактор угадывает - робот знает. Теперь вообразим, что с помощью биометрии (eye tracking уже позволяет), робот будет анализировать реакцию на конкретные идиомы, эпитеты, синтаксические конструкции, визуальные образы. Подобные технологии все более доступны, вопрос лишь в объеме данных и скорости переработки.
Такая система автоматически начнет производить все более аттрактивные заголовки и тексты. Да, возникнут другие проблемы: 1) аттрактивность станет белым шумом, 2) погоня за реакцией читателя сделает генеративную журналистку дегенеративной. Но на переходном этапе важно другое: анализ человеческих реакций при достаточной массиве данных заменяет роботу отсутствующее у него чутье.
In other words, any argument about what a robot cannot do is possible to bring an equally convincing argument about what a person cannot. In this contest of abilities, there is also about a draw.
Competition at the very beginning, but already - a draw.
Robot Journalism Roadmap
Based on what we already know about robots, journalists, you can roughly imagine how they will take over the world. Well, or, for a start, a profession. Here are the key pieces of the future mechanism of robot-journalism.
1) Big data. The ability of the algorithm to work with big data, obviously, will only develop. The ability to detect correlations to some extent replaces human creativity with the robot. The best minds and the best investors are working on it. And not over the preservation of journalism.
2) The corpus of all texts and analysis of accumulated auditory reactions. Monitoring the network or accumulating all journalistic texts in its own database, the algorithm can easily calculate which texts — with all their tactical and technical characteristics — collect the most likes, readings, reposts, comments.
Of course, there may remain some niches that are only possible for a human instinct. So they will be reserves of bio-columnists. However, on an industrial scale, these blind spots for the robot will not play a role, especially considering how much the total circulation of content will grow due to the robots.
3) Biometrics. Having gained access to unexpressed reactions, to body language, the robot will be able to take into account not only old reactions to old texts, but also instantaneous reactions to here-and-now perceived content. The man read the headline about Putin - the touchscreen senses how his hands sweat, the webcam captures how the pupils widen, the microphone hears the breathing more frequent. All clear. And so - on the arrays of thousands of readers. This household polygraph will replace the editor's flair.
Human substitution technologies will evolve with acceleration. And why should they stop there? Yes, and what is considered a level achieved? It is open up and forward road. Somewhere in front of it looms the specter of artificial intelligence.
Forecast for the world
Christian Hammond, co-founder and chief scientific officer of Narrative Science, is confident that by 2030 computers will be writing 90% of the news. Hammond also believes that in 2017 the computer will write material worthy of the Pulitzer Prize.
Since the exact dates are called, to call others is not interesting, it only predicts that quantitative and qualitative competition awaits us with cyber-colleagues. In quantitative competition, bio-journalists are losing right now. In quality, we will give way within 5-7 years.
It is curious that in the early stages of the transition of journalism from people to robots, native editors will become the killers of the profession. Editorial staff are forced to produce as much content as possible in order to increase traffic. The journalist has no time to deal with serious topics; he has to drive materials to the site: "motion for motion's sake" - movement for the sake of movement. This effect of journalism theorist Dean Starkman (Dean Starkman) called the hamsterization of journalism - from the hamster wheel, squirrel wheel. Hamsterization of journalism in the pursuit of traffic reduces the time the journalist works on the material for the amount of materials: “do more with less”.
One hundred thousand readers of one article is high-quality journalism. But is it necessary to fight for a hundred thousand readers for one article, if you can put a hundred thousand articles, each of which will bring ten readings? Who will the editor choose: a capricious journalist with a rising salary and three texts per week, or a trouble-free algorithm with a decreasing monthly fee and three texts per minute?
The Associated Press does not buy the services of Wordsmith because the blacksmith writes better than a man. And because he writes more and faster. So the dispute about the quality of the text is generally irrelevant. Robots will capture editorial due to their economic, not literary merits.
Journalist Robots in Russia
As for the forecast for Russia, you need to understand first where we are now. Botwork is actively developing in the political sphere and in commerce, but not in journalism. Known for the curious experiments Sports.ru - perhaps the most innovative edition in this field. For example, the robot keeps a sports chronicle there. Another robot writes short phrases and selects gifs for them, sometimes it turns out even funny.
Sports.ru made a header generator to help a sports blogger.
But the most interesting, in my opinion, Sports.ru experiment in this area is a robot that has learned to use popular comments from related topics. By reducing them and adding a couple of strong cheerful expressions, the robot quickly gained 300 "pluses" of popularity, and this is the first cut-off, which gives the portal's bloggers status for the "minus" of other participants. “I scored in two days, much faster than regular users,” Mikhail Kalashnikov, director of products for Sports.ru told me. “After that, we turned it off. We think what to do now with this knowledge.”
At first glance, the usual forum bot. But what is important is this: he used the analysis of human reactions (calculated the most popular formulations of readers) to produce his own content. It remains for him to learn from Wordsmith to generate texts - and this will be a full-fledged robot journalist, who knows better what the editor wants.
In the recently announced Yandex for Media project, the first in Russia robot journalist has already been presented, writing if not ready texts, then blanks summarizing information about traffic jams and weather. This service is reminiscent of the Los Angeles Times crime chronicle algorithm, which has been in operation since 2007. This is not yet a journalist, but rather an assistant journalist, who forms a tape of text notifications based on statistics.
Vsevolod Pula mentions an algorithm that compiles stock quotes reports on finmarket.ru, some other algorithms-assistants. Surely in Russia there are robots writing texts for other tasks, say, Ashmanov. But apparently, there is still no journalistic text, even financial or sporting, written by a robot in Russia. Otherwise he would have thundered. If so, then the backlog of our market from the West can be estimated at about 5-7 years.
Forecast for Russia
The threat of robots for Russian journalism must be considered in the context of other threats. The Russian media have more than enough experienced all the misfortunes that the blogosphere with a viral editor causes journalism. She takes away from the media exclusive, audience, market.
Nevertheless, the development of corporate media is becoming the main threat to journalism in Russia. The problem is that normally corporate media must compete - this will improve their social sound. But our main corporation is the state. His corporate media competition does not imply. Here is his media presence and drives journalism out of the market. Deprived of market sources of media are turning into a broadcast channel for this corporate media. At best, they creatively rethink given topics. This is not exactly journalism, but rather the laundering of reality on a particularly large scale.
So there is good news: the algorithms will not have time to destroy Russian journalism. She did not get them.
Wordsmith or Quill are unlikely to learn Russian. And for the cultivation of a Russian-speaking robot of such capacity, first of all, a strong venture capital market is needed. In general, the algorithms will not take away the profession from Russian journalists.
However, to stay away from technological innovations will not succeed. News of the invasion of robots one way or another will alarm journalists. For editors, the following strategy will be advantageous: to be among the first at the beginning of robotization and among the latter at the end.
Now the editorial use of a text-generating algorithm can be an interesting PR action - for the audience, for the investor. But when and if the algorithms already saturate the market, then the human voice that sounds amongst their metal gnashing will certainly get a special appeal to the public.
In this sense, oddly enough, the consumer value of human journalism in its final stages will be ... an editorial mistake. Until the robots understand this and learn to simulate the same mistake.
In 2014, Wordsmith, one of the two most powerful news algorithms, wrote and published ONE BILLION Notes in the media. This is more than what is written by human journalists. Part of this billion went to a physical increase in media content. Well, the other part has already gone to replace, that is, ousting people from the profession.
The market will demand more, more. And nothing will prevent the robot from writing as much as it wants, except for the limited ability of people to read it all. But this restriction will be lifted when robots will be reading too.
- I do not believe in journalistic robots, but I know that robots can do a part of journalistic work. The founder of the startup Narrative Sciencе Stuart Frenkel told Slon.ru about how computers can replace media workers. Peter Birger. Slon.ru, 26.05.10.
- Journalists who never sleep. Robo-hunter, October 1, 2014.
- Journalists are no longer needed: which robots work for us. Oleg Akbarov. Look At Me, July 2, 2014.
- In Case You Wondered, a Real Human Wrote This Column. By Steve Lohr. The New York Times, Sept. 10, 2011.
- More News Is Being Written By Robots Than You Think. By Jason Dorrier. SingularityHub, Mar 25, 2014.
- Can an Algorithm Write a Better News Story Than a Human Reporter? By Steven Levy. Wired 04.24.12.
- Robots have mastered news writing. Goodbye journalism. By Liat Clark. Wired, March 06, 2014.
- Robots Are Invading the Journalists. By Kevin Roose. New York Magazine, Jul 14, 2014.
- Automation, Algorithms, And Android News Readers: How To Robots Are Changing The Face Of Journalism. By Edward Sinnott. Junkee.com, 7/8/2015.
- Can Robots Do Public Interest Journalism? By Nicholas Diakopoulos. European journalism observatory, February 11, 2015.
- Enter the Robot Journalist. Users' perceptions of automated content. By Christer Clerwall. Journalism Practice, Volume 8, Issue 5, 2014.
- Should media outlets tell readers? By Matt Sutton. J-source, May 12, 2014.
- 5 white-collar jobs robots already have taken. By Erik Sherman. Fortune, February 25, 2015.