The NY Times last week released results from its preliminary research into whether and how blockchain might be useful in combating the problem of misinformation and disinformation spreading on social media platforms. It also released a proof of concept of how a blockchain-based network for photojournalism could work, which it built in partnership with IBM Garage.

Called the News Provenance Project, the work is being conducted by the Times’ R&D Lab, with the help of IBM technologists and input from other news organizations. The goal, according to a post by the Times’ Sasha Koren, was to see if technology can be used to solve some of the problems technology has created, such as the production of ever-more convincing deep-fake videos and the easily manipulation of photographs to create false narratives.

The initial idea was that publishers could contribute to creating a healthier information ecosystem by surfacing information they already have about the work that they publish. Our main hypothesis was that adding context to news photos — in the form of metadata, with information that is often contained in a caption, but gets stripped out as a photo travels beyond a news outlet’s own sites and apps — might help people make better decisions about the credibility of the images they see on social platforms and elsewhere around the internet.

The project chose blockchain because its architecture enables the storage of metadata about an asset in (more or less) immutable form and the tracking of the asset’s provenance over time, including any changes made to it.

As a database, a blockchain can provide network members increased confidence in the reliability of its data, due to its enhanced security in the way it stores that data, which makes the records published to it more or less immutable. It also ensures transparency of all “transactions,” or updates to the records stored in the system, so that any major changes made to a record would be recorded and visible. In addition, it offers shared ownership of a database among a number of entities — in this case, publishers.

The Times deserves credit for investing the time, money and resources into addressing a systemic problem like the spread of fake news. But the results so far also illustrate the limits of blockchain’s capacity to solve systemic problems, at least within the media industry.

The Times’ researchers conducted in-depth interviews with 34 news consumers and daily social media users from a range of backgrounds, regions and preferences for left- and right-leaning news sources, to learn how they judge the credibility of news images.

They found that readers do apply critical thinking in interpreting images, although their approaches vary significantly.

Broadly, the subjects showed widely varying awareness of the contextual elements of an image on social media, such as its source, caption, date and other types of metadata. They also showed widely varying degrees of trust in news outlets. Using those data points, the Times sorted subjects into four broad categories:

Distrustful news skeptic (low trust, high awareness): Seeking to call out bias in mainstream media, a person in this category may use motivated reasoning to find any evidence to confirm their belief that the media is pushing a particular agenda.

Confident digital news subscriber (high trust, high awareness): A person in this category is digitally savvy and is comfortable distinguishing between true and false news when provided information from news outlets they trust.

Media-jaded localist (low trust, low awareness): This person may feel marginalized by mainstream media and uncritically accept hot takes from unofficial accounts as truths. They want news that feels local and authentic, but they don’t want to be misled by false information intended to deceive.

Late-adopter media traditionalist (high trust, low awareness): A person in this category may be more comfortable learning about news through older mediums such as television or newspapers, but less comfortable making sense of news online within the noise of social media.

While technology certainly influences how different people experience and react to news media content, the differences in their attitude and behavior are clearly not solely a function of technology. And no technology, by itself, is going to reconcile those differences.

The primary roles the News Provenance Project envisions for blockchain, moreover, such as surfacing more metadata and providing transparency into an image’s provenance, are likely to be relevant only to some categories of consumer.

The researchers also found that many of the signature concepts associated with blockchain — “immutability,” “encrypted database” — made little impression on subjects in the study.

A Heavy Lift

Another major challenge, which the Times’ researchers to their credit acknowledge, would be implementing a new blockchain-based infrastructure for photojournalism at a scale sufficient to make a meaningful difference.

To give a sense of what it would take to realize the full vision of surfacing provenance information on news photography, we’d need changes at every step of the process, from the time a photo is taken, to every instance of its publication and display. In brief, an ideal set of changes would be for:

• Camera makers to help photographers ensure time, date and location settings in cameras are exact.
• Every news publisher to modify their management processes for photo metadata so that they adhere to a common set of standards, such as those maintained by the International Photo Telecommunications Council (IPTC).
• All platforms such as Google, Facebook, Twitter and Apple, as well as chat apps like WhatsApp and Signal, to ensure the consistent display of this information.

Those issues will be familiar to anyone who has considered seriously the use of blockchain to address the challenges around data transparency and sharing in the music industry and other media sectors. The biggest challenges come down to questions of interests, incentives and inertia, which do not bend to technology no matter how innovative or powerful.

The ease with which misinformation and disinformation spreads and feeds back on itself today is in large part a function of the viral, peer-to-peer architecture of our new information ecosystem. It is a feature, not a bug.

That said, I hope the Times keeps plugging away with the News Provenance Project. The problem it aims to tackle, while larger than any one technology can solve, could not be more urgent. Even taking small bites out of a big problem can be worth the effort.

Total movie admissions in the U.S. fell 4.6% in 2019, to 1.24 billion, making last year the second worst through the turnstiles since 1995, according to the National Association of Theatre Owners.

The worst year was 2017, when only 1.23 billion tickets were sold.

The decline comes as theaters face growing competition from streaming services, not only for consumers’ time and entertainment dollars, but for their first-in-line position in the movie distribution system.

Two of 2019’s most high-profile releases, for instance, Netflix’s “The Irishman” and “Marriage Story,” which together helped the streaming service rack up an industry-leading 24 Oscar nominations, received very limited exclusive theatrical runs.

One reason for those limited runs, of course, was the reluctance by many theater owners themselves to book the films at all because they did not like Netflix’s proposed terms — a policy that increasingly looks self-defeating.

A recent survey by The Hollywood Reporter and Morning Consult found that a plurality of U.S. adults — 48% — prefer to watch new release movies via streaming service, compared to only 37% who would prefer to see them on the big screen.

Preventing movies from being more widely available on the big screen seems likely only to fuel consumers’ preference for in-home viewing.

It is a policy that also seems likely to grow increasingly untenable. As the major studios aggressively pursue their own direct-to-consumer initiatives, launching streaming services of their own and adopting increasingly proprietary distribution strategies for their content, pressure on the exclusive theatrical window will only increase.

If theater operators cannot find a way to accommodate themselves to the new strategic landscape, missing out on a few Oscar nominees will be the least of their problems.

Artificial intelligence has been a presence in Hollywood for many years, first as a subject of dystopian science fiction plots (“2001: A Space Odyssey,” “The Terminator”) and later as a CGI tool used in making many of the same type of films. Now, it’s moving behind the scenes entirely.

The Hollywood Reporter reported this week that Warner Bros. has signed a deal with L.A.-based Cinelytic to use its A.I.-powered project management system to provide predictive analytics on proposed film projects at the greenlight stage.

Cinelytic’s system can assess the value of star in any territory using data from multiple sources, and make projections of a movie’s likely earnings from theaters and ancillary distribution channels.

With the deal, Warner becomes the first major studio to sign up Cinelytic, following its earlier deals with Ingenious Media and Productivity Media.

Cinelytic is not the only developer shopping A.I.-powered predictive analytics to the studios, however. Belgium-based ScriptBook has trained an A.I. algorithm to be able to analyze a script alone –without stars, director or other creatives attached, — to generate box-office projections based on a range of story and character elements.

Although no studio has publicly signed on to use ScriptBook’s system, three of seven it has been presented to have expressed interest and are currently in discussions with the company, ScriptBook founder Nadira Azermai tells me.

Nor is Hollywood the only corner of the entertainment business where predictive analytics is finding traction. The music industry is also showing growing interest in the predictive powers of A.I.

Silicon Valley-based Chartmetric , for example, says its A.I.-powered A&R tool can shortlist which of the nearly 2 million artists it tracks will have a big career breakthrough within the next week. Pandora-owned Next Big Sound says it can predict which of nearly 1 million emerging artists it tracks will hit the Billboard 200 for the first time within the next year. Secret Chord Laboratories‘ dopr tool combines data from 33 years of Billboard charts with insights from neuroscience research to predict how a particular audience will respond to a song.

Big data

The use of analytics is not new, of course, particularly in Hollywood. The studios have long had green eye-shade types poring over past box-office, home video and ratings data to try to assess the likelihood of success for a new project. Even the determinedly non-empirical record industry will consult chart position and airplay to decide how much promotion to put behind an act.

The difference today is the immense amount of data, generated by streaming, that is available to analyze, and the computing power to sift through it all.

Neural networks and machine-learning algorithms make it possible for A.I. systems to analyze petabytes of data and discern patterns within it that elude humans.

Moreover, unlike humans, machine-learning algorithms get better at their job as they are fed more data to process. ScriptBook began training its algorithm with a library of about 30,000 produced scripts. But just as the natural language processing algorithms that power Siri and Alexa get better at recognizing human speech the more of it they hear, ScriptBook’s algorithm will be able to produce more refined and precise projections over time as it analyzes more scripts.

The limits of human processing historically left plenty of room for critical creative decisions still to be made by “ear,” or “gut” or “feel.” Data could inform but not supersede human intuition.

But with A.I.’s vastly superior processing power, and the mathematical improvement in its powers of discernment over time, that gap is starting to close, raising difficult questions about how and where to draw the line between man and machine.

A.I. developers, their eye on human purchasing managers, are careful not to put to fine a point on the question.

“Artificial intelligence sounds scary. But right now, an AI cannot make any creative decisions,” Cinelytic founder Tobias Queisser told The Reporter. “What it is good at is crunching numbers and breaking down huge data sets and showing patterns that would not be visible to humans. But for creative decision-making, you still need experience and gut instinct.”

Still, with potentially millions of dollars at stake, some mission creep is inevitable. ScriptBook, for instance, a tool it calls Deep Story, which Azermai describes as “the next generation writers room.”

By entering a few parameters (genre, character names, source material, etc.) Deep Story’s algorithm can generate an initial script, which human writers can then use as a starting point to produce a shooting script.

Similar technology is already in use in newsrooms, where A.I. systems sift through documents and real-time data feeds to suggest story ledes and angles to editors and reporters.

Even creative writers have begun experimenting with A.I.-powered story generation.

All of which is adding greater urgency to the questions being asked by the U.S. Patent & Trademark Office, the World Intellectual Property Organization, and other intellectual property agencies about how, whether and to whom copyright protection should be assigned or apportioned for works produced, initiated or assisted by algorithms.

I am currently preparing a larger report on the evolving role and implications of A.I. technology in the media and creative industries that will be published later this year by Penske Business Media’s Variety Intelligence Platform. Stay tuned.

The World Intellectual Property Organization this week issued a request for comments on whether copyright, patents or other intellectual property rights could or should be extended to works produced by artificial intelligence. The notice comes as part of a public consultation the United Nations agency launched back in September, and the comments will be used to refine its working draft (pdf) of the topics and questions to be addressed in the next, formal policy-development phase of the consultation beginning in May 2020.

The WIPO consultation parallels a similar process underway at the U.S. Patent & Trademark Office, which issued its own request for comments on the same topics in October. Other countries have also begun wrestling with questions of authorship and ownership in the emerging era of machine creativity.

The formal inquiries are at a very preliminary stage. Both WIPO and the USPTO acknowledge in their requests for comment that they are still trying to figure out what question they should even be asking and how they should be framed.

Read more and sign up for our weekly newsletter.