One thing that strikes me is that these auto-generated videos add no additional information to news reports; they merely present the content in a different way.
While that may be a way to broaden consumption of the content -- the auto-playing videos on Facebook feeds is a good example -- I don't think it does much to change readers'/viewers' perception of the quality of that content. If anything, it has the potential to cause them to perceive it as of lower quality than if they'd just read the article it was based on.
I see this as the next version of slideshows. News orgs used to like them because each new slide could be treated as a new page view. But readers found them to be, at best, of little additional value, and at worst, really frustrating to get through.
What news orgs really need is to invest resources to automate things that increase readers' perception of quality, and then figure out a way to monetize them, rather than looking at ways to monetize and then figuring out how to package content to fit the monetization strategy.
One initiative that can do that -- and this is where I think more resources should be invested -- is data science. Sifting through large public datasets and identifying trends, aberrations, etc. is something a computer can do as well or better than a human, and I would imagine that the fruits of those investments would be ripe for monetization.
Just one example: We have become conditioned to expect news content to be free, but we're more willing to pay for services. So imagine a news org that said, "Hey, we'll let you view all of our articles for free ... but then for a dollar a pop, you can sign up to receive an instant email notification when we publish a follow-up to a story you are interested in." I'd pay $1 to get notifications of follow-up stories to a particularly enthralling court case, or some local news of significant relevance to me. Best part is, news orgs already largely have the metadata that would make such a service work. They just need to invest in making computers do the work.
Can publishers handle advanced math? They are taking the same audience, same advertisers, and just dividing it up in to more pieces - and like you said, at a lower quality. I suspect the growth in online advertising has done a good job in masking the long term destruction being done to the publisher's revenue. Facebook has made it rain for a few years. That is coming to an end just as Google rolls its news traffic over to AMP. I don't want to own any publicly traded publishers right now.
Auto generated video is just a grab for using press release content for auto play video ads. Upon closer inspection, there is really no business left. It looks more like Demand Media's content farm now, to see who can provide the thinnest layer between traffic source & advertiser. The business I see is Google/FB providing the audience and Google/FB providing the advertisers. Making a higher quantity of videos no one wants to see is not going to recapture ad dollars from some kid playing a video game that an audience wants to see.
The only thing that amazes me is how many big publishers, including popular tech blogs, are able to run sites which barely function on a late model iPhone. I have no idea what metrics they are looking at. The past few months I've gone back to RSS feeds and bloggers so I can get readable content again.
An end to my worthless rant is publishers will probably have to deal with less glamour and focus on producing extremely high quality content that users will pay for. Most verticals have no market for paid content. Those probably have to turn in to donor based non-profits or business models shift to selling products.
Advertising has held fairly steady as a percentage of US GDP for ~100+ years (1-2%). However, GDP has grown so there is room for more players esp if the growth is in a tiny sector of advertising spending. Further, It's at the low end of it's normal 1-2% band we might see a bounce back if the economy starts taking off.
I'm curious, what does a former newspaper editor use, when he want low noise, high value content, for casual stuff(not necessarily extremely fresh), both in browsing and in search ?
My side-project is trying to achieve some semblance of this.
From your experience what makes you think people would be willing to pay? Wouldn't you need a massive user base consuming hard to create local content like what Patch was supposed to be? http://patch.com/
I would not to say that online video is surging, more like it is being forced. Nowadays, it is difficult to read an article on a news site without having some associated video being autoplayed (with advertisement of course). I want to read an article, not watch it.
On a side note, is there a way to disable these videos? I do not believe in blocking ads since a publisher should have a monetary gain for producing content, but I do NOT want video. Ever. Unless it is Youtube or Vimeo and the like of course.
Autplaying videos infuriate me. I've given up after trying to configure chrome to stop autoplaying videos. Now I just turn off javascript and only enable it for the sites I need javascript.
The most effective way I've found to disable autoplay is to simply block the video content networks which provide these. At my last update, it was a rather small set.
0.0.0.0 player.theplatform.com # Autoplay video
0.0.0.0 link.theplatform.com # Autoplay video
0.0.0.0 ci-2862d2c8d6-68f418d2.http.atlas.cdn.yimg.com # Autoplay video
0.0.0.0 big.assets.huffingtonpost.com # Autoplay video
0.0.0.0 ht1.cdn.turner.com # Autoplay video
0.0.0.0 ht2.cdn.turner.com # Autoplay video
0.0.0.0 ht3.cdn.turner.com # Autoplay video
0.0.0.0 ht4.cdn.turner.com # Autoplay video
0.0.0.0 ht5.cdn.turner.com # Autoplay video
0.0.0.0 ht6.cdn.turner.com # Autoplay video
0.0.0.0 ht7.cdn.turner.com # Autoplay video
0.0.0.0 ht8.cdn.turner.com # Autoplay video
0.0.0.0 ht9.cdn.turner.com # Autoplay video
Alternatively, uMatrix should allow you to globally blacklist, then optionally whitelist same.
That's been less effective lately, so I've also added an HTML5 autoplay blocker. It's fairly effective. Not sure how much it prevents from loading to save on data, but at least it prevents being halfway down the page reading and suddenly having noise come from the speakers.
Maybe turn on adblocker and send the news conglomerate a few cents in the mail? That's what I would do if I really felt the need to balance the scales with the websites I block. I'm not being sarcastic, that's really what I would do.
Just to remove all doubts--a large part of the reason the video is there and autoplayed has a lot to do with video CPMs compared to standard display CPMs.
Quakebot still has a byline and its own section on the LAT website [0], but it remains mostly unchanged since its first reveal (perhaps its implementation is more automated, but its public output is about the same). That Quakebot -- and all the other ways that public data can be efficiently harnessed as a service to readers and reporters alike -- isn't talked about in Tronc's buzzword-heavy press releases about machine learning and artificial intelligence is, IMHO, indicative of Tronc's questionable aims and viability.
Yes, "Tronc" was a huge step down for the once-famous Chicago Tribune. I wrote in to their "Colonel Tribune" columnist, and got back a rather sad reply.
The National Weather Service has an API endpoint [1] which returns weather forecasts in XML. Put in a latitude and longitude, and it calculates a forecast for that location. The data is returned in a structured format. This is the underlying data for most weather sites. One XML item is <wordedForecast>, which looks like this:
WEATHER FORECAST FOR 2 MILES SE SAN JOSE INTERNATIONAL AIRPORT
CA ON JULY 11TH AT 2:34 AM.
TODAY, JULY 11TH: SUNNY, WITH A HIGH NEAR 81. NORTHWEST WIND 6
TO 16 MPH, WITH GUSTS AS HIGH AS 21 MPH.
TONIGHT, JULY 11TH: MOSTLY CLEAR, WITH A LOW AROUND 55. NORTH
WIND 16 TO 21 MPH DECREASING TO 7 TO 12 MPH AFTER MIDNIGHT.
WINDS COULD GUST AS HIGH AS 26 MPH.
That's what you hear in broadcast radio weather broadcasts.
(Note that, if you have a NOAA-capable radio, you can also have this read to you in a the NWS's classic, distinctive robot voice. Brings back memories of waiting with my family for the tornado warnings to pass as we holed up in the basement).
> you can also have this read to you in a the NWS's classic, distinctive robot voice
That "robot voice" is relatively new. Not that long ago, it was a tape loop of a person reading the forecast, played over and over and updated every few hours. Definitely into the late 90s I think.
In my childhood home we had a radio with a "weather" button that, when pressed, would turn the radio on and tune it to one of the standard NOAA frequencies for a set amount of time (probably about a minute or so). It was astonishingly useful, on par with the Amazon Echo I have today and use for a similar purpose, but with probably six orders of magnitude fewer transistors involved. Simple, but effective.
In the early days of TV people often read newspapers into the camera. Early movies were often essentially vaudeville shows on the screen. Those things fill space for a minute then fade quickly.
I think one can say with confidence that nobody really wants to watch automatically generated title and clip art videos. Which means this trend will soon pass.
I saw Bloomberg doing this quite a while back so it is growing not shrinking. If it is auto-play, the content doesn't actually exist for the consumer, it is there to tell the advertiser they are buying an ad for content.
It is like a more advanced version of old school blackhat SEO scripts which would scrape content and then mash it up so it would appear original and unique to Google. In some cases these publishers may not even have real humans watching those ads.
Yup, you nailed it. It's robots making "content" for other robots to "watch" in order for yet more robots to auto-magically-bid on ads to show that second set of robots so the people who own the first set of robots can make a little bit of money.
This is strangely similar to generative adversarial training in neural nets. They are composed of a generator and a discriminator. The generator has the purpose to generate examples that fool the discriminator and vice-versa. They improve together, iteratively. It's a smart approach but has stability problems in training, because both the generator and the discriminator have to be balanced.
Maybe they could try to apply this kind of technique to improve their auto-generated articles to the point where they can't be distinguished from those created by humans, and at the same time, fool the FB bots which filter the feeds.
The future of our finance (as predicted by Charlie Stross): by robots, of robots, for robots.
If it reads like the "America InAction" definition of Corporate Robotocracy, feel good that it's least a more likely dystopian future than nuclear holocaust.
How about taking into account how consumption of this content has changed? This kind of stuff is great for mobile consumption, which, as far as I understand, is growing at a rapid pace.
> This kind of stuff is great for mobile consumption
To whom? Most people I know are more incensed at autoplaying video on their smartphones even more than on the desktop. Their audio is more likely to be turned up on their phone, and it's harder to turn it off, which is what most people immediately do.
Maybe future generations will develop some tolerance or actual interest for this kind of video-from-blog-post format [1], but I find it irksome. Speedup tricks aside, video is unskimmable, so I can't consume it at my own pace unlike I can with text.
But I understand, from a business perspective, the appeal of:
- preroll ads, which are (as of this moment) harder to block than other types of ads
- the promise of organic 'virality' brought upon by a visually stimulating format
- the desire to use automation to cut down on the human element of producing such converted content
This just absolutely stinks of "we want to create more content without having to pay for it, how can we convince people to look at ads when they view it". There's no attempt to do anything here but just hoover up dollars.
Oh, it's an automated video summary of otherwise existing content. Pardon me if I'm not really impressed. I was expecting something along the reach and challenge akin to 'automated' or 'machine generated' music, of which this might be similar but seems far, far less difficult. Cliff's Notes videos are useful I suppose.
Just today I noticed that autogenerated github activity visualizations are showing up as the first results when searching YouTube for various software projects. This is extremely annoying, and I hope YouTube bans or at least deranks them.
It seems like services like these could be more helpful with author involvement. The example DHL clip, for example, helps tell the story visually, and if the article's author were involved in assembling a video package (without needing to learn complicated video production software) he or she could write a complimentary script and have a great video summary right alongside the more in depth text article.
Is there a video production language/scripting tool for doing this sort of thing? Sort of like an ImageMagick for video. So you can put a certain graphic at a certain time, composite other source videos, do transitions, etc. All scripted in code, no hands on work required.
Just wanted to add that it's a super powerful tool with a ton of plugins developed by really knowledgeable people. The programming can have sections of parallel code and sections of serial code combined in cool ways, and you can pipe the output right to encoders like x264.
One thing that strikes me is that these auto-generated videos add no additional information to news reports; they merely present the content in a different way.
While that may be a way to broaden consumption of the content -- the auto-playing videos on Facebook feeds is a good example -- I don't think it does much to change readers'/viewers' perception of the quality of that content. If anything, it has the potential to cause them to perceive it as of lower quality than if they'd just read the article it was based on.
I see this as the next version of slideshows. News orgs used to like them because each new slide could be treated as a new page view. But readers found them to be, at best, of little additional value, and at worst, really frustrating to get through.
What news orgs really need is to invest resources to automate things that increase readers' perception of quality, and then figure out a way to monetize them, rather than looking at ways to monetize and then figuring out how to package content to fit the monetization strategy.
One initiative that can do that -- and this is where I think more resources should be invested -- is data science. Sifting through large public datasets and identifying trends, aberrations, etc. is something a computer can do as well or better than a human, and I would imagine that the fruits of those investments would be ripe for monetization.
Just one example: We have become conditioned to expect news content to be free, but we're more willing to pay for services. So imagine a news org that said, "Hey, we'll let you view all of our articles for free ... but then for a dollar a pop, you can sign up to receive an instant email notification when we publish a follow-up to a story you are interested in." I'd pay $1 to get notifications of follow-up stories to a particularly enthralling court case, or some local news of significant relevance to me. Best part is, news orgs already largely have the metadata that would make such a service work. They just need to invest in making computers do the work.