The Future of News
The Editor-in-Chief of Bloomberg News, in his recent Convocation address to graduates of the Asian College of Journalism, Chennai, speaks, among other things, about harnessing Artificial Intelligence for journalism and countering fake news. This is a slightly abridged video and transcription of his talk.
I am choosing technology because, it strikes me, that is the thing which journalists are most worried about, with the possible exception of some governments at the moment. And it strikes me that when you talk about technology and journalism it pays to be humble. I am a good exponent of that. Back at The Economist - as it was mentioned, I was that magazine’s editor in 2006 - we published a cover which was written like a ransom demand and whose title was “Who killed the newspaper?” And it became a staple of media conferences around the world. Whenever people wanted to talk about the way the advertising had gone, from print newspapers to online, they would use that. So when, sadly, newspapers began to go bankrupt, I at least had the condolence compensation of feeling that I had been mildly prophetic about it.
Courtesy: The Economist
But was it actually fair? Just two weeks after I became editor of The Economist, a small company called Twitter was set up. It is now arguably the world’s biggest daily newspaper. It has 400 million people going to it. Or look around at the quality press nowadays. The New York Times just announced it has 3 million digital subscribers and is aiming for 10. Looking back, now, I would argue that the problem was not something to do with journalism, and it was not to do with the internet, it was just another technology that came along. It was the way, the suicidal way, that our industry responded to the internet and especially our embrace of it giving away our content away for free. Now that is being reversed, and you can see us and all our competitors, The Wall Street Journal, the Financial times, the Washington Post, are all expanding again because they are getting money from charging people.
So my point, to begin with, is that technology has always changed journalism, and not always for the worse. That seems relevant now that a new spectre is haunting our industry, and that is artificial intelligence. I think there are two basic fears. The first fear is that journalists are going to be replaced by robots. And the second is that the quality will go down as fake news proliferates and consumers gather in ghettoes of like-minded thought. Now I don't think that’s true. Yes, some jobs will be automated, but others, perhaps more, will be created, and journalism, I think, will get more interesting and better. As for fake news, yes, journalism will get more personalised, but not always in a bad way. And yes, we will still sometimes be hoodwinked by fraudsters and autocrats. But, to be honest, that has always happened. In general, I think AI will help deliver a world where we get more information and more truth.
But first, I think first we need to be fairly precise about what is actually happening in newsrooms around the world, which I think is a lot more than most people realise. And here I am going to make no apologies for using Bloomberg a bit. There are lots of news organisations, and I would try to pay tribute to them as I go through, who are doing some of the things that I talk about, but on the whole I think we tend to be doing slightly more than most, and that’s partly a function of size - we have 2700 journalists at Bloomberg, an array of platforms that go all the way from Tic Toc on Twitter to the terminal. Its partly also because we are a technology company. We have 5000 engineers working on the technology, perhaps 200 of them working on AI and news, and its partly, finally, because news is unbelievably valuable to us as somebody who feeds news towards the financial markets. If you want an example, it came - needless to say just before I arrived at the Bloomberg - when the Swiss central bank which used to have its currency, the Swiss franc, pegged to the euro, suddenly decided to drop that peg and set the Swiss franc free. What happened was that our reporter in the Zurich office of Bloomberg saw that and wrote a story immediately saying that this had happened. Sadly our opposite number, our main rival who shall remain nameless, happened to be in the loo at the same time. So, according to the legend, by the time other guy had returned from the loo around 1 trillion dollars had changed hands as the Swiss franc rose by 30 %. Now you can draw various lessons from that. One of those might be to never take a bathroom break, but the second one is that news is unusually valuable if you deliver it to the financial markets.
Now, I think there are four areas where this new internet technology or artificial intelligence is changing journalism. The first has to do with the automation of repeatable events. Here I am going to use the example of an earnings report. When I first arrived at Bloomberg four and a half years ago there was a team of young people, quite similar to the ones I met earlier today, a group of nimble-fingered, mainly youthful journalists, whose speciality was being able to write headlines the moment a company reported things. The moment a company like TATA or Apple or GM announced its results, this team would jump on it and immediately transmit things within a few seconds, again hoping to get the results out before our competitors. And then shortly afterwards, say 30 minutes, a beat reporter, the man or the woman who covered TATA would come in and write a secondary story summing up the whole picture, taking all the results and making it come together - TATA sales rose outside India, Jaguar Land Rover did well, costs were down... that type of story. It was basically an analysis of numbers with, typically, a quotation at the bottom from a stock market analyst. So that was how it was. Now, what happens is that the technology called Cyborg, perhaps an unfortunate name but all the same, scrapes the earnings from TATA’s website and it spits out headlines and then within minutes, certainly within a minute, usually within seconds, it can spit out a bullet-pointed story which contains most of the things that somebody used to write about after 30 minutes, minus of course the quotation from the analyst, and the computer keeps on looking for more. It no longer just looks for revenues and profits and what the earnings were last year. It begins to look for what we call the secondary data, the numbers of Land Rover Discoveries sold in China and those sort of things which move markets even more.
Roughly 30 %of all the things we do at Bloomberg includes some element of automation. But it is really worth stressing, really worth stressing, the use of that word ‘some’. Because we need journalists at the beginning to tell the computer where to look, you need to tell it what matters at TATA to set the templates in the right place so the computer begins to look for the right things. We also need humans to double check what comes out, both for accuracy and also for surprises. The shares of TATA might jump one way or the other because somebody senior resigned, and nothing to do with this particular piece of news. Rather than machines replacing humans, it tends to be humans and machines working together. And the main game so far has come in the breadth of the number of companies we can cover at the same time. What about that beat reporter, the man or woman who had to write up the story? Well, the interesting thing about that is, it ups the ante. You can no longer just list the numbers and pretend you've done a story. You now have to try and say what this means. You have to be able to say, look the reason why TATA’s sales went down was because something has wrong with its China strategy, or something has gone wrong with Brexit, or whatever you happen to choose on. You have to use your sources to say what exactly is happening inside the company and you have to also come up with some idea about what's going to happen next. You have to say what this means. And in financial journalism, I think this would change many things and it would change things in other parts of journalism.
In financial journalism, it tends to be a bit of an arms race. We are spending money on computers and so are others, but you can actually look, and you can see automation creeping into other parts of journalism.We had a piece the other day about a very small newspaper on the south coast of Britain which uses computers to pull information about weather, about traffic from the local government computers and automatically publish things that people get to pick up. I think the same is true if you look at the coverage of cricket nowadays. You see all the facts and statistics which come out: how close a cricket ball is to the wicket, which soccer player has run the furthest, and that by itself becomes part of the way that people begin to tell their stories.
The second area where AI is changing journalism is to do with the craft of journalism. This sounds rather small but I'm sure to the people who are graduating today it will save a lot of time, which those of us who a slightly older may have wasted. Now, at the moment, if you go and record an interview, you usually have to go back and spend the time for the tedious business of transcribing it. But, increasingly, with voice recognition technology you just put it straight through that and you'll get an immediate thing. Or in our business our reporters have to listen to briefing calls from CEOs. Now a computer can listen to several briefing calls and look out for particular words that can somehow trigger stories of someone or another. The biggest game changer perhaps in a market like this could be automatic translation. In many places, including India, people often report in one language but write in another. Now the machine can translate. In Turkey, for instance, our reporters break news in whatever language they want to, and then let the computer do most of the translation. Again, I say mostly because you need to keep humans in the process. But if you give a story first to a machine to translate, it can increasingly do so with ever greater accuracy. And I know that maybe people in this room who believe that the machines often get it wrong. They do. There's always a time when the first time you talk about the ties between China and India and the machine translates it as ties like this (indicating the neck tie he is wearing), but the second time it learns, and by the third time it's got better. And yes, some languages are difficult, including Hindi and Tamil, but what happens is, other languages are easier. Say, for instance, we can now do from Spanish to English by machine, it's harder to do English to Mandarin or to Korean but you can translate from English to Mandarin and then Mandarin to Korean. Everything in some ways is moving in that direction. To me, it is like a train coming down the tunnel towards us. These are all things that will help us push forward and we'll make the craft of journalism easier. It may even make a difference to help people hire people. You may end up now hiring the best local reporters in local language and then look for ways in which that can be translated.
The third area where artificial intelligence is changing journalism is what I would call 'signalling'. Here the machine find something, it scours and it finds something which is interesting and it asks you, ‘Do you want to write a story about it?’. For instance, at Bloomberg, we have a unit called social velocity. It features a relatively small number of journalists and a relatively large number of computers, and what they do is, they scour social media all over the world looking for particular words, and I apologize for the grisly nature of this, but disaster, explosion, resignation, shooting... and when they find something, they then say to the journalist 'Has something happened? Then go and look at it'. Again, this is a work in progress but you can already see the results. We used it with the Colombo attacks recently, and it does give you a very clear advantage, but it does rely enormously on you knowing exactly who to trust, because you can spot that something is happening really quite quickly from all the tweets and all the social media, but to actually establish exactly what is happening, that is where you nearly always have to find a human in the end to verify it.
And also it's interesting that is much easier, we are much better, at verifying things in print than on video. To give you an example, a year ago, there was sadly another terrorist attack in New York on the subway. We were very quick, just purely by surveying social media things and police reports, to establish that this attack had happened. But it was very difficult to immediately to verify a picture, slightly grisly picture, that appeared on the web of somebody, allegedly the terrorist, lying in a pool of blood at the subway station. But that gets much more difficult. Because you know the people can fake pictures terribly easily, so you have to spend time trying to work out whether the subway floor looks like that, whether the number of pixels has been changed. But again it tends to be moving in the direction of the computer helping us. We don't always get it right, but it's getting much better. Another version of signalling has to do with data. The machine can send us a message, to our economics team, and say, 'Look, the GDP of Germany relative to that of Greece is now at its lowest level since 2009. Do you want to write a story about that?' Or, 'We can point out that the three directors of a company in Bangalore have all just sold shares in a company. Do you want to write a story about that?' Some of the football data you see on the TV is exactly the same. At the bottom of it, it's effectively saying to the person who's commentating, 'Look, Leicester has now won three games in a row, have scored more goals than everybody else - I support Leicester - and it immediately gives the commentator a chance to pick from it. It is the machine that analyzes how far a player runs. It is a machine that tells you how quick close the ball is to the wicket and then asks the commentator, in a fact, do you want to add anything. In all these cases, I would argue, the machine is acting much less like a journalist, and more like a commissioning editor. It is saying, how about doing a story on this. That also gives editors a way of holding journalists to account: there's a record showing that you were asked that question about a company share price, why didn't you write about it? What's your excuse when the takeover is announced?
The fourth area in which I think AI could affect what we do will be the issue of personalization. For me, I think this is the holy grail of modern journalism - the idea that you can find a newspaper that is particularly angled towards you. And it is incredibly difficult, and the main reason why is readers. Readers do not like handing over information. They'll fill out a couple of things saying that they are quite keen on cricket, and they happen to live in Bangalore, but that is about it.The other problem is a bit like the old joke about advertising.The joke about advertising was that somebody knew that half his advertising worked, but he just didn't know which half. The same, from an editor's point of view, is slightly true of journalism. We know that customers, clients, readers, like bits of our journalism and not others; we're not just quite sure what it is, and to make it even harder, the evidence is that people don't like journalism too much if it's solely targeted at them. I'll come back to this, the idea of ghettos, in a second, but it's true, I think, that in most of our lives we don't just want news that we thought was important, we also want to find out that bit at the bottom, that element of serendipity. There was a story this morning about the Russians having equipped a whale with what looks like spying technology and sent it into a harbour in Norway. There is no way that anybody in this room would have ticked a box saying Norway, whale and Russia. But if you read it, it was interesting, and that element of serendipity is something you need to maintain.
At Bloomberg we have a slight advantage in this, because when it comes to companies and things, people would tell us what they're interested in, because they want to get the data that they need. So, for instance, we now have fifty channels of particular forms of news aimed at particular things; and in the morning we produce something which I would argue is about as close as the first digital newspaper which is personalized, called Daybreak, and that has one section called ‘need to know’, which is the big headlines: Modi has done this, Gandhi has done that. And then it has an element of 'my news' which focuses on particular things that you're interested in, and then finally it has a 'nice to know' element, which is where you’d discover the stories about whales or Russians or princess Megan.
So those are the four basic ways in which I think artificial intelligence is changing Journalism. Again I'm not claiming that Bloomberg has a monopoly in any of them. But I do think we're probably as advanced as any people and I think these things, all these things, are likely to be a big part of your lives in journalism as it goes forward. Before I come back to deal with the issue of fake news, I've to just point out again that big issue: I do not so far see any evidence that these are computers that are taking away people's jobs, that it s still very very important that in all these things the human side of journalism still matters, and I think it will have to become even cleverer. What do I mean by that? One way of thinking about it is that the world is moving from 'what' to 'why' and 'what's next'. The 'what' of journalism is still incredibly valuable and by 'what' I mean what has just happened. If you can tell people that, as we seek to do at Bloomberg particularly, that is still incredibly valuable, but it is incredibly valuable for an ever shorter period of time.If we break the news that one company is due to take over another, it is immensely valuable for ten to fifteen seconds, because by then the share prices of the two companies would have moved - to use my story about the Swiss franc. Even fairly humdrum bits of news now travel all the way around the world, almost immediately.
So yes, we will continue to invest a lot of money into breaking things on the 'what' side. But we know that they will still be incredibly valuable for short-ish periods, and that's why the 'why' and 'what's next' parts of journalism I think matter enormously. The more robots help with the more humdrum ends of reporting, the more the value of the human end in terms of analysis and commentary. We think we're breaking more news than anybody else at the moment, but we are investing ever more in commentary; and one reason why is it is very difficult to replace really good commentators with machines. You cannot automate Martin Wolf for the F. T., you can’t automate The Economist or The Atlantic, and editors also matter,sww because the danger with machines is spam. They tend to hit people with endless amounts of things and the real cleverness often of being an editor is knowing what to leave out, what to filter. Virtually everywhere we look at, we see people who will pay a lot, or at least pay something to get news, but what they do not want is to be completely run over by the amounts of information they get. And that, I think, makes a difference not just in terms of editing, but also in terms of formats. We have to get smarter about those two. In the old days, when I began in journalism in 1987, a story always looked pretty much the same. Back then if you opened the New York Times, say, and I looked at a story on Indira Gandhi, it would be pretty much the same thing, a piece of news: she had just decided to do this or that. And then a degree of explanation about who she was for the people outside India, what the exact state of play in terms of different things was, and then at the end some version of a conclusion and typically it would go on about a quarter of a page of a piece. Now everything is multiplied. You have blogs, you have videos, you have graphics. But even in print, even in that basic story, that is being broken up. Nowadays the bit about Indira Gandhi and Congress would go into some kind of explainer, the reason being that people in India didn't need to have Indira Gandhi explained to them; people outside might need a bit more about the Congress party, but you can find that by clicking through. Well, you could follow an updated format where some bits are put into news, other bits are put into the explanation. The video I think is also splintering linear TV with a succession of programs which will give way, I think, to much more video on demand. In the same way as we go look at places like Netflix for films, we will increasingly look at it for news as well. And beyond all this, and this is an element of hope at least from my end, I think we are catering to readers or viewers who are normally more short of time than they are of money. And the reason is that today the sort of ideas and the information that journalism, a good journalist, can put across is still worth something. At The Economist, we would think it was the most expensive magazine in the world, but as we used to continually point out, it was only a cappuccino a week,and that is not a lot for the kind of information you have. Even at Bloomberg, on our consumer side, it is not much more than that. I think it matters, and you can see it in the way young people are picking up things like Spotify, Netflix that they are prepared to pay for information, for content, when they can, but they do not want to have their time wasted and so that's one reason why formats have to get tighter, and in general the trend is toward shorter, quicker bits of journalism. But not always. I would submit to you that if you want to understand the Indian election you could watch maybe twenty-five videos, you could what look at a variety of different data graphics, you could read lots of what we call blasts or quick things this way or that. But you might be much better sitting down to read one long piece in Business Week or in New Yorker; not necessarily because the writing would be superior, but because it would save you time. It would be more time- efficient for you to sit and read one piece than all those many shorter ones. So for all those reasons, I think that some strands of journalism are going to survive and the others would live very happily on top of what computers do.
So that brings me to the second part you might say, Okay that's fine I understand that we still need human journalists, and we'll get more news, but what about fake news? Won’t AI make it easier to fool us. I think there is a lot of truth in that. We all remember the cyber-utopianism of the 1990s where people dreamt about citizen journalism making us all purer. And that went the same way as all those dreams that it was possible to make news for free. Now we know the same technology that was used to democratize the world can also be used to suppress dissent. China is ahead of us, rather notably, in some versions of AI such as face recognition technology. And yes, it is possible, as we’ve seen, for bad people to use this technology to fool people, not least in journalism. If Robert Muller was correct about Russia, it was possible to help fix the last American election from a Moscow office building for a cost of a million dollars a month. And you can also herd readers and viewers into the same ghetto of opinion, so you only hear the sort of news you want to hear.
I think there are three basic types of fake news. The first is State-sponsored fake news. This is the world of Vladimir Putin. He used bots and other gizmos to spread untrue rumours or just deny the truth. This version of fake news has changed in quite an interesting way. In the old days, something would happen and then you would simply deny it: for instance, if the separatists in Ukraine had shot down an airline with what looks like a Russian missile, you would simply deny it once and move on. Nowadays the way the fake news is done is, you spread hundreds of different rumours in hundreds of different places, and you keep on as often as possible trying to not establish that there was one truth. That is the way the States on the whole do it. Next, I would argue, comes moronic news. This is where somebody comes up with some particularly weird or wonderful idea: that Pakistan has captured fifty Indian soldiers, or that Hillary Clinton is having an affair with Imran Khan, and they broadcast it. And what happens is sometimes the bad States pick it up and push it through. Finally, and most complicated, there is commentary that one sort of people thinks is fine and the other group thinks is fake news, and this I think is the most difficult area, and is one where I take a fairly fundamentalist attitude - that commentary is commentary and you should accept it even when you disagree with it. In America I often find myself arguing with people who talk a lot about the birth of Fox News as being something that was different and conservative. But for many people on the right in America, Fox news was giving them a form of commentary they hadn't had before, and so it does not surprise me they picked up on it. It's also worth noting, given the many people who point out or claim that the Fox has always been onTrump’s side, that to begin with at least, by far the toughest interviewing of Trump came from the conservative wing - you think of the Megan Kelly interview.
What about the charge of ghettoes? Here again, I think that that we happen to be overly simplistic. I know in India it's a difficult one. I haven't yet seen data on it. If you look at America, people tend to say, yes. there are two sorts of ghettoes: there's the ghetto on the right where people watch Fox, they listen to Rush Limbaugh, and they read the Wall Street Journal. On the left, you have a group of people who watch MSNBC, listen to NPR, and read The New York Times. And the evidence, academically, is, well, yes. There are a few people like that, but most people still want as part of their media diet something else, and what is particularly attractive to me is the idea that most of the conservatives I've met spend a large amount of their time reading liberal media such as the British Guardian in order to discover what the enemy does and, by the same token, people go the other way; except here it's a little bit more different, but in general, I think the idea of ghettoes is being exaggerated.
So I think these are all ways in which bad news machines will come up with ways to manipulate good machines, and they will be there, and we have to live with them. But I think there are several points in our favour as to why we will win in the end. The first is that the victory so far is that it has made us much more vigilant. You only have to look at the way anything to do with American elections is now policed and questioned. And don't forget of all the different ways in which AI will help journalists to check things. You can scour the Panama papers in seconds. You can check whether the picture on the right really is Donald Trump or a picture of a town bombing in Pakistan and it will help us, on the whole, disseminate news. The line on the Washington Post that ‘Democracy Dies in Darkness’ I think is true, but AI helps us shine a light by getting more news to more people. It is harder in general, not always in particular, but it is harder in general for bad people to hide.
And in the end, to go back to what I said at the beginning, fake news did not begin with the internet. Humans have always found ways to manipulate humans; the Trojan horse was arguably the original piece of fake news; or you could say it was the apple in the garden of Eden. If you go back to ancient Rome, Octavian poisoned the minds of the people of Rome about Antony by publishing a fake will that showed that that he was leaving on his money to Egypt. Shakespeare made a very convenient villain out of Richard the Third. That's a nice way of looking back at the way. There was really no golden age where all journalism was fine and unrestricted. The very first newspaper in India, James Hickey’s Bengal Gazette - I was reading about it this morning - was set up to expose the horrors and the evils of the East India Company and he inevitably ended up in prison rather quickly. It wasn't as if then there weren't people trying to push back.
Mahatma Gandhi had a very nice line about English newspapers. He said, to the English voter their newspaper is their Bible. Then he said, they take their cue from newspapers that are always dishonest. The point is that many times newspapers have been used in good and bad ways. I think if you look at India at the moment, yes, there are some areas where I see fake news being spread by new technology. I can see it with WhatsApp, with the way the people try to persuade people that one group or the other is deeply wrong. But I don't think you need to go all the way to the internet to find examples of things that are wrong. If you want one, just open the newspaper, any newspaper in India, and look for a map. You are surrounded in this country by countries that are less democratic than you are: China, Pakistan, Russia... all these ones. I've dealt with all of them. They have many faults. But the one thing they have never insisted is that we draw a border where it doesn't physically exist. And that is something India sadly does. It is fine always to show the lines where people claim particular areas that are part of it, but it is also a journalist’s job, I would submit, to say where the border, for better or worse, currently exists, and that is something that I hope the people in this room will help change in Indian journalism. I would also submit that there was nothing particularly digital about the ouster of my friend and colleague Bobby Gosh- who now works for us - from the Hindustan Times. I think, again, that is something that goes back to an older thing that I don't think has anything to do with technology or newness. Nor has it particularly to do with one party exerting control over the media; you can see that on the other side with what happens in Calcutta. In short the same things are being used against journalists everywhere. It has not changed, and it will not change, and the job of many in this room will be to fight that and to push back.
Journalism has always been interrupted by technology. In 1840, the Times of London started using a steam-powered press that effectively raised the amount of papers you could print tenfold. Cheap newspapers, many of them scandalous, sprouted everywhere. The biggest was Benjamin Day’s New York Sun which sold for a penny. It specialized in extravagant fake news. It reported that half-bat-half-humans were living on the moon. When people questioned Day, he said, prove me wrong. But gradually the tide turned, for two or three reasons. One was the number of people, like hopefully the people in this room who wanted to be good journalists to change it. The second was advertises who thought they no longer wanted to have that soap and their various other products advertised alongside the pictures of the half-bat-half-human, and the last was readers, who said I will pay for good news.
India I think needs good journalists. It needs the facts... and it needs investigative journalism, regardless of what the people in power think. So go out there, I think, and explain the truth or facts. Follow the line of James Hickey, no matter how difficult it was. Be brutally honest with people, not least with yourselves when it comes to your experiments with truth. I'll end with a quotation from Gandhi himself. He said, “In the very first month of Indian Opinion, I realized that the sole aim of journalism should be service. The press is a great power, but just as an unchained torrent of water submerges whole countrysides and devastates crops, even so an uneducated pen serves but to destroy.”