"Robot journalism" is good for journalism in the same way that outsourcing is good for programmers and engineers: drudge code entry that needs to be done can be identified and contracted out at a much lower rate, therefore saving the in-house programmers to do innovative, business-expanding work.
However, that reality doesn't always follow, depending on the dynamics of the business and the forward-thinkingness of management. And this is in engineering/technical companies that are making viable amounts of money. Will the managers who run cratering news companies have the same foresight? Not going to hold my breath.
(note: I fully support the automation of writing/observation of digital feeds...I'm just skeptical that the journalism industry will apply it either efficiently or in a way that benefits their human workers)
>> "Journalism is becoming a more highly skilled job,” Doctor said. “Simply showing up, in the Woody Allen sense — being able to read a press release or interview a single person, and write up a story that is understandable in 750 words — that's not going to be enough."
Just wondering where these higher skilled journalists are going to learn their trade? Longer in j-school? So more debt?
The OA was I suspect given routine business reports to write as a cub reporter to get him used to deadlines, structures, systems, house style &c. Same with minor league sports reports in provincial newspapers in the UK.
The next attack on journalism might be great curation, not robots. Why ? because no single human has the capability to write the best story in any given case.Add to that the capabilities of unpaid bloggers, industry insiders and the like, you get a serious threat to journalism.
Alas ,there are no great content curation tools AFAIK.But when there will be...
Can you expand on this? What would great curation/curation tools look like? I feel like I agree with this insight, but don't know much about the field.
Giving you articles that are fun,strongly insightful and 100% right almost every time, on your exact interests is the goal of a great curation.Bonus points for short an to the point writeup when appropriate.
How do you do it ? not sure yet. The best that i can do today to do is constantly weeding out my blogs/twitter subscriptions. But i'm far from that goal.
> ... on your exact interests is the goal of a great curation
I am not so sure this is a good idea. It will trap people in bubbles, where they will find only an echo chamber for their existing opinions and never have them challenged. It will insulates them from serendipity and it will prevent them from exploring, acquiring new interests and perspectives and so on.
People keep worrying about filter bubbles, which is a rational but misguided fear. Having written some of these systems (Google News), it really does not happen the way people imagine it would.
At the end of the day, the factors that define a good curation framework do not actually create a filter bubble. People love serendipitously discovering new content, and the optimizing for the "best article" does not actually optimize for a singular viewpoint. As a result, any curation system that produces a filter bubble will not actually feel as good of a system to the end user, and will not get as much adoption.
Chris, in my view the problem with this reasoning is that you are assuming that people are good at noticing that they are inside a filter bubble.
I have a less optimistic perspective: I suspect people want to feel like they are being exposed to diverse information while not having their beliefs or preferences challenged. This makes sense, considering that our cognitive resources are limited and being exposed to information that contradicts our beliefs is a psychologically painful experience.
So I claim that this is a cultural problem that companies have no incentive to solve. I am not claiming that I have a solution, nor do I endorse regulation or (shudder) government intervention on cultural dynamics.
I suspect that we more and more belong to tribes that are divided across intelectual instead of geographical lines. Surely I have more in common with you than most of my neighbours (just going by the fact that we are both participating in the same niche forum). The problem is that nation states are still geographical and have to arrive at some democratic consensus to avoid tyranny. The avenues by which such consensus can be achieved are getting narrower. Maybe we will transcend nations, maybe we will devolve into tyranny. I am hoping for the former, but become worried when I see people believing that the problem doesn't exist.
I try do as I preach, so I am open to having my opinion challenged :)
Everyone's reality consists of unique connections between things. To someone "art" is more strongly connected to "abstract art" than "photorealistic art". For someone else "art" refers most strongly to "oil paintings". Our system tries to capture and understand what kind of unique connections an individual has. Random tries to understand what "art" means to you personally and what kind of related things might be interesting to you.
The app also allows you to connect things freely thus letting you express both your rational and irrational self. There're no universal categories or connections between different things - rather it's about an individual's own "ontology" that's created through usage. The "associative ontology" evolves continuously both through the actions of the individual and other people using the system.
Yep. I’ve heard complaints about “filter bubbles” before, but have always found these to be entirely theoretical — not a critique of an actual implemented system. In reality, as <chris_va> says, a “filter bubble” is different from a “viewpoint bubble.”
An example: If our recommendation engine at http://recent.io/ concludes you’re interested in Barack Obama, it will (surprise!) recommend articles about Barack Obama. It doesn’t filter by political viewpoint. You’ll get articles from both lefty sources and right-leaning sources.
Also, of course, we offer “top news” if you prefer!
That's similar to what we're trying to do with http://grasswire.com, but not so much "articles" as "tweets/videos," and crowdsource fact-checking after the fact.
So, maybe not very much like it at all, but I agree with your original analysis that curation will be king. News/journalism feels very much like the Internet did pre-Google. There has to be some aspect of real-time curation/fact-checking built in IMO. But obviously I'm extremely biased.
> on your exact interests is the goal of a great curation.
You might be interested in signing up for the beta at http://recent.io (disclaimer: I quit my job at CBS/CNET earlier this year to found this company). We released an alpha version to early testers last month, with the beta to come soon.
We recommend articles based on what we learn your interests to be, and that selection of recommended articles is the best I’ve ever seen.
>Any differences from Zite (now owned by Flipboard)?
Yep, differences in design, approach, and implementation. It offers better recommendations and can do things Zite couldn't.
This is not to discount the groundbreaking work done by the Zite team at the time; the last CEO of Zite is a friend who managed the product well. It's always easier when you've been able to learn what other companies have done well and what they haven't. And we've been able to do things like using Google App Engine for the backend from the beginning, which wouldn't exactly have been possible when Zite was founded 9 years ago (and because of earlier GAE limitations and higher costs may not even have been practical even 2-3 years ago).
As a practical matter, Zite is about to be discontinued with some of its features, in theory, to be eventually incorporated into Flipboard. (The March 5 announcement of the acquisition said they'll be "shutting it down" as early as September 5: http://blog.zite.com/2014/03/05/zite-is-flipping-out/ ) So if you liked Zite, you might want to sign up at http://recent.io
On a totally different note this startup, Automated Insights, is based in Durham NC (where my startup Spreedly is also based). There are all the elements of a great startup scene; a clustering of talent, a decent incubator that does two sessions per year at around $150K per startup, cheap real estate, excellent food and affordable living. The greater area has 3 good schools. There are numerous other startups here but I can't help but give a plug for what's happening as I know many developers here wrestle with work/life balance. Right now it's very good in Durham. Biggest drawback is lack of local seed funding but things like AngelList are making that less of an issue than a few years ago.
Robot journalism isn't going to be great for journalists. Publications will quickly gravitate towards the content farm model where every article will be accompanied by 20 - 30 keyword permutations and no overhead for articles that don't pull social media traffic/shares/links. You can already see companies like Mashable pushing the limits of doing this by hand.
I wonder if/how Google will differentiate between automated news vs machine-generated spam?
It's easier than you might think. Machines are, if nothing else, good at identifying things other machines have created.
People also often forget that Google doesn't just have to look at the page content to analyze something. Reader behavior, for example, can tell you a lot.
Some people see tables of statistics and think, "AUGH! MATH!" Whereas the same information presented in natural language does not generate the same response.
I'm also assuming that the software can provide a little more context than just that sort of opening paragraph. If not, I agree with you that this isn't exactly earth-shattering software (then again, what is these days?).
Whatever a robot can do, they can do a million times an hour.
Next we'll have robot comments on robot blogs. We already have impersonal birthday wishes on fb -- where you can make an app that automatically wishes happy birthday. And the recipient can install an app to automatically thank everyone. It's as I was half joking in the past -- guys will outsource their robots to have sex with their wives' robots.
I wonder if you got a large dataset of these short human written articles, and trained a generative model to produce them character by character. This was done on wikipedia: https://www.cs.toronto.edu/~ilya/fourth.cgi
However, that reality doesn't always follow, depending on the dynamics of the business and the forward-thinkingness of management. And this is in engineering/technical companies that are making viable amounts of money. Will the managers who run cratering news companies have the same foresight? Not going to hold my breath.
(note: I fully support the automation of writing/observation of digital feeds...I'm just skeptical that the journalism industry will apply it either efficiently or in a way that benefits their human workers)