First up (and in the third time this has happened so far in the last couple of months and the second time it's happened for this particular topic), I've been vindicated about an opinion I expressed in a past Filmism.net Dispatch after the fact.

Watch this video where Robert Rodriguez interviews James Cameron about Avatar: The Way of Water, from the about the 29 minute mark.

Cameron talks about how when we see a movie in a theatre we make a 'contract with yourself to give it it's time to work on you', and how we give it our full attention in a different way from the way you do when we watch something at home.

I'm positive Cameron read this recent Filmism.net Dispatch and he was just agreeing with everything I said. I similarly talked about how you adopt a moviemaking 'stance' that's different depending on how and where you watch the movie, but how I advocate using the theatre-going stance even if you're at a home (ie by giving it your undivided attention).

As he says, and in what might be the most beautiful phrase ever spoken to describe the act of seeing a movie in a cinema; 'you've got to bear witness'.

But to lighter matters. I was very interested to read this post on the Netflix technology blog recently, about how the company makes decisions about what to produce or buy with the help of AI.

I thought it was interesting because it brings an old chestnut bubbling to the surface about whether movies are pieces of artistic artefacts or content commodities, and how the way we see them affects the way we treat (and make) them.

As you'd expect, it talks about how greenlight executives (or whatever Netflix's equivalent job title) are deciding what movies and TV to make based partly on algorithms.

To many, it will make sense. If you're one of Netflix's 195m subscribers around the world it means that what you watch, what you've watched in the past, the exact point you stop watching something you lose interest in, even how long you hover over a specific title on your device is all carefully collected and stored because it's the exact stuff big data was made for.

There's too much information there (and more all the time) for a human to mine, catalogue and understand, and somewhere in all that is the perfect and unadulterated answer to the question 'what kind of movie or show will audiences like?' Using technology to delve into it is a way of getting as close to that answer as possible.

It's also very much Netflix's modus operandi. If you read No Rules Rules, the 2020 book CEO Reed Hastings co-authored to lift the lid on how Netflix works as a company, you'll know he's a guy who trusts systems and data, not gut instinct or intuition.

Among its most revealing gems was the way feedback both up and down the chain of command is so formalised. With every comment systematised, spreadsheeted and logged as actionable and intended to be revisited to check progress and improvements, the company was trying to make an algorithm even out of relations between colleagues.

But to a lot of people for whom the art in the entertainment business is important, the idea of letting computers tell us what to create based on impenetrable metrics about what's worked before is the very model of technology gone mad and overstepping its remit.

Surely we make (and consume) art because it means something to the artist and hopefully the audience, and if it doesn't connect with people there are a million reasons why, few of them measurable or even visible in such a complicated supply chain of production, marketing, exhibition/broadcast, etc.

Now, if you're tempted to dismiss is as the tech industry taking over Hollywood and making every decision a data-driven SWOT analysis, we should remember that the movie industry's always measured and acted upon what's successful.

Why else have we always been so obsessed with box office, and why else was legendary critic Gene Siskel complaining about there being too many sequels and remakes as long ago as 1976?

Is it too uncomfortable a leap for us when computers do what panicked or elated studio executives used to do on Monday morning after poring over the weekend's figures (ie decide what to make next)?

As I alluded to before, it's all mixed up in the longtime unease we've had with the way we use the word 'content' nowadays. It was popularised in the early web era to refer to any text, imagery or video that appeared on the early HTML pages of the Web 1.0 (and to differentiate material created by the website owner from banners, pop-ups and other advertising).

Now it's interchangeable with the word 'art' in the way it refers to the stories produced for screens (or paper, canvas and beyond), and a lot of people still don't like it.

Now, to be fair to Netflix, the blog post tries to assure the reader right out of the gate that they still make art the old fashioned way. Content, marketing and production executives make the key decisions, and the second paragraph leads with the assertion that commissioning a title is a creative decision, and that the AI engineers merely support them in that effort.

But it has to be said that the company takes such technological approaches to hit prediction to extremes. One such passage; 'we learn a model on a large set of historical titles, leveraging information such as title metadata (e.g., genre, runtime, series or film) as well as tags or text summaries curated by domain experts describing thematic/plot elements'.

In other words (unless my knowledge of technology and systems terms is way off), a human being extracts as much information as possible about a movie's intangible elements like theme, aesthetic, design and various genre- or style-specific milieu, modularises them and affixes them with endless tags.

You can then imagine a computer spitting out a 'yes' or 'no' (one day maybe even spitting out an intact outline or screenplay) based on the synthesis of how successful those elements have been based on how many subscribers interacted with them in other titles before.

I have to admit the entire blog post is a bit above my expertise overall. It's about probabilistic statistics and not movies, after all.

But in talking about 'similarity maps' that relate to each other through a spatial metric called 'Euclidean distance' and results called 'learned embeddings', you can't help coming away with the feeling that writers and producers are entering Netflix's offices for pitch meetings and being met with a 60s-style mainframe with magnetic tape reels sitting in a huge leather chair and chomping on a cigar.

It might be the perfect economic expression of what data can do. It might be another step towards a 1984-level technocratic nightmare where we're all cogs in an inhuman machine, all sparks of individuality, expression and humanity extinguished by the impenetrable decision making powers of distant electronic overlords.

It's probably somewhere in between, but what do you think?

Now on screens, it seems movie production as a whole still hasn't cranked back up to pre-COVID levels because there just hasn't been much worth seeing on movie screens lately.

Like the rest of the industrialised world I duly lined up for Avatar: The Way of Water. After another viewing of the original and a few trailers my expectations were lower than ever, so I wasn't disappointed to find it only mildly diverting.

In fact I've been feeling out of sorts a lot lately because of how several very critically successful films have failed to grip me like they have the rest of the world, among them Everything Everywhere All At Once and The Worst Person in the World.

One that's been very worth watching is Mystify: Michael Hutchence, a documentary about the late singer's life, art and struggles that eventually and tragically claimed him.

And if you're a fan of the Godzilla myth and enjoy films that do a good job of using a fantastical or genre element to comment on society, I also recommend 2016's Shin Godzilla.