Mind Readings: The WGA Strike and Artificial Intelligence

Mind Readings: The WGA Strike and Artificial Intelligence

Can’t see anything? Watch it on YouTube here.

Download the MP3 audio here.

Listen to the audio here:

Let’s talk about the very timely topic of the Writers Guild of America strike and one of their demands. In particular, this is about artificial intelligence. No surprise, I have no expertise in the rest of their proposal about how writers are paid, but I do have expertise in artificial intelligence. In their position document, the WGA said the following:

“In regards to artificial intelligence, they demand studios regulate the use of artificial intelligence on MBA-covered projects as a means of products that the union works on. AI cannot write or rewrite literary material, can’t be used as source material, and MBA-covered material can’t be used to train AI. Okay, let’s unpack this.

They’re saying that, from their perspective as writers, they don’t want machines writing or rewriting literary material. Writing, I can understand, rewriting machines are really good at that. The critical one is that their writing can’t be used as source material and MBA-covered material can’t be used to train any topic. That ship has sailed.

If you look at all the public, large language models like GPT-4, GPT-NeoX-20B, PaLM, LLaMa… all of them. There are tons of sites online that archive scripts and transcripts of TV shows and movies. There are subtitles available for every single major production going to open subtitles.org. Their writing is already in these models, decades of it in these models. So that ship has sailed.

The part that I think the WGA has underestimated, particularly with streaming companies, is that streaming companies are technology companies first, their entertainment company second, right? They are tech companies. Think about Apple at Apple TV plus, okay, think about Amazon and Amazon Prime. Think about Netflix. These are not legacy entertainment companies like MGM, Paramount, or CBS. These are tech companies that happen to have a foot in the wall entertainment.

They will, without question, use the WGA strike as a golden opportunity to replace writers as quickly as they can. And they have the tech and the know-how to do so. Yeah, legacy studios will struggle with getting the tech up and running. But Amazon, it would not surprise me if Amazon Studios already had custom large language models trained specifically on their massive content catalog ready to go. I mean, Amazon started selling their bedrock language model interface two weeks ago through their AWS service. It’s for others to be able to do so.

That I think this is probably not the position I would have gotten, right? Because think about will Hollywood adopt machines instead of humans for scriptwriting. Of course, they would. It’s an improvement of profitability. And let’s face it, Hollywood studios are not known for their super original content. A lot behind projects get greenlit for profitability reasons, not artistic ones. And if a production can cost $5 million, less just, you have machines doing 80% of the writing.

No studio exec in the right mind is gonna say no to that. And think about it. This is why we’ve got a never-ending supply of reboots and retreads and entertainment. This strike and the implications for AI is going to change the profitability models in entertainment. The companies can spin up the tech real fast; they’re going to shed headcount, and they’re going to move ahead of their peers, the legacy shops that can’t get up to speed with tech is going to fall behind the more tech-enabled companies. And when you think about copyright, yeah, the script, the writing itself might not be copyrightable if it’s purely generated by machine, but that script is not the final IP; the produced show is. So from that perspective, there’s not much of a copyrightation.

So what should the WGA consider as its negotiating position? For me – Now, again, I am not a writer, I do not work in entertainment other than the Save Warrior Nun campaign, and I do not have expertise in the entertainment industry. But I do know machines. So instead of that absolutist, no AI negotiating position, what they probably should have gone with, and they still can, is to make it a demand that WGA writers are required to oversee the use of AI in script production with a quota of one to one. One person for one instance of machinery for production is using a machine. With a technical person on a union production, a WGA writer is required to supervise its use and its output. And so it’s not a case where a tech division in a streaming company could just spin up a room full of GPT instances and scrape a whole bunch of scripts. Now, you still need to do a decent amount of prompt engineering for that and custom model training. But it would be trivial for the WGA to say, “Yeah, our writers are going to be attached to the hip to your tech folks, our writers are going to be in the system itself, looking at what people are putting in for prompts and what comes out.”

And this is going to do two things. One, it keeps the WGA writers meaningfully employed. And second, it will show both the studios and the writers what the strengths and limitations of these large language models are because they do have strengths like rewriting things; they’re really good at that. Writing new stuff? Not as good as that. They can’t really generate truly original new ideas. But they absolutely could take, you know, a script from an old Knight Rider episode and transform it into a Pokémon episode. These machines are extremely good at rewriting.

Prompt engineering, which is the discipline of writing that programming language, plain English code that goes into these machines to get them to do what we want them to do, is something that requires expertise. There is no one better qualified conceptually to be a prompt engineer than a skilled writer. So the WGA’s position should be that they are going to be doing the prompt engineering as well as supervising. I think that would give the studios the ability to use the technology to reduce the time to output and speed up production without eliminating WGA writer positions. And WGA writers will be able to supervise and put a leash on AI without outright saying, “Nope, it’s not allowed here,” because the reality is, it’s already in the technical studios, and those who use it are just gonna march right ahead. And they will use the strike as an excuse to say, “Well, I don’t have human writers, so we’ll just use the machine writers,” and that’s not a position that’s going to help consumers.

If we’re going to get retreads of retreads, it’s not going to help writers. And in the long term, it might not help the studios either because they will be more profitable with it, but the content won’t be as good. So, again, if you work for the WGA and you are a union member, please feel free to send this along to the leadership. If you work in entertainment and you’ve spotted some obvious flaws in my argument, please let me know. Leave a comment or whatever. But that’s where things stand right now, and I think there’s going to be a sea change accelerated by the strike. So hopefully folks can adopt a more nuanced position on AI sooner rather than later.

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.

Christopher Penn 0:00

In today’s episode, let’s talk about the very timely topic of the Writers Guild of America strike and one of their demands.

In particular, this is about artificial intelligence, no surprise, I have no expertise in the rest of their proposal about how writers are paid.

But I do have expertise in artificial intelligence.

In their position document, the WGA said the following.

Christopher Penn 0:23

In regards to artificial intelligence, they demand studios regulate the use of artificial intelligence on MBA covered project as a means products that the union works on.

AI cannot write or rewrite literary material can’t be used as source material and MBA covered material can’t be used to train AI.

Okay, let’s unpack this.

Christopher Penn 0:41

They’re saying that, from their perspective, as writers, they don’t want machines writing or rewriting literary material writing, I can understand rewriting machines are really good at that.

Christopher Penn 0:52

The, the critical one is the their writing can’t be used to source material and MBA covered material can’t be used to train any topic.

That ship has sailed.

Christopher Penn 1:01

If you look at all the public, large language models like GPT-4 GPT, Neo x 20, be long on all of them.

There are tons of sites online that archive scripts and transcripts of TV shows and movies, there are subtitles available for every single major production going to open subtitles.org.

Christopher Penn 1:25

Their writing is already in these models, decades of it in these models.

So that ship has sailed

Christopher Penn 1:32

the part that I think the WGA has underestimated, particularly with streaming companies, is that streaming companies are technology companies first, their entertainment company second, right? They are tech companies think about Apple at Apple TV plus, okay, think about Amazon and Amazon Prime.

Think about Netflix.

These are not legacy entertainment companies is not MGM, or paramount, or CBS.

These are tech companies that happen to have a foot in the wall entertainment.

Christopher Penn 2:06

They will, without question, use the WGA strike as a golden opportunity to replace writers as quickly as they can.

And they have the tech and the know how to do so.

Yeah, Legacy studios will struggle with getting the tech up and running.

But Amazon, Amazon, it would not surprise me if Amazon Studios already had custom large land of models trained specifically on their massive content catalog ready to go.

I mean, Amazon started selling their bedrock language model interface two weeks ago, through their AWS service.

It’s for others to be able to do so

Christopher Penn 2:46

that I think this is probably not not the position I would have gotten, right.

Because think about will Hollywood adopt machines instead of humans for script writing.

Of course they would.

It’s an improvement of profitability.

And let’s face it, Hollywood studios are not known for their super original content a lot behind projects get greenlit for profitability reasons, not artistic ones.

And if a production can cost $5 million, less just you have machines doing 80% of the writing.

Christopher Penn 3:13

No studio exec in the right mind is gonna say no to that.

And think about it.

This is why we’ve got a never ending supply of reboots, and retreads and entertainment.

This strike and the implications for AI is going to change the profitability models in entertainment.

The companies can spin up the tech real fast, they’re going to shed headcount and they’re going to move ahead of their peers, the legacy shops that can’t get up to speed with tech is going to they’re going to fall behind to the more tech enabled companies.

And when you think about with copyright, yeah, the script the writing itself might not be copyrightable if it’s purely generated by machine but that script is not the final IP at the produced show is so from that perspective, there’s there’s really not much of a copyright ation.

So what should the WGA consider as it’s a negotiating position for me now, again, I am not a writer, I do not work in entertainment other than the Save Warrior Nun campaign, right.

And I do not have expertise in the entertainment industry.

But I do with machines.

So instead of that absolutist, no AI negotiating position, what they probably should have gone with and they still can that could change the position.

So if you’re with the WGA group, please consider taking this input and feeding it back to the folks who are making decisions

Christopher Penn 4:32

make it the demand instead that WGA writers are required to oversee the use of AI and script production with a quota of one to one.

One person for one instance of machinery for production is using

Christopher Penn 4:45

a machine with a technical person on a union production a Wi Fi router is required to supervise its its use and its output.

And so it’s not a case where a tech division

Christopher Penn 5:00

Have a streaming company could just spin up a roomful of GPT instances and scrap a whole bunch of scripts now, you still need to do a decent amount of prompt engineering for that, and custom model training.

But

Christopher Penn 5:12

it would be trivial for the WGA.

To say, Yeah, our writers are going to be attached to the hip to your your tech folks, our writers are going to be in the system itself, looking at what people are putting in for prompts and what comes out.

And this is going to do two things.

One, it keeps the WGA writers meaningfully employed.

And second, it will show both the studios and the writers what the strengths and limitations of these large language models are, because they do have strengths like rewriting things, they’re really good at that, writing that new stuff, not as good as that, right? They can’t really generate truly original new ideas.

But they absolutely could take, you know, a script from an old night writer episode and transform it into a Pokeyman episode.

Maybe these machines are extremely good at rewriting,

Christopher Penn 6:08

filing

Christopher Penn 6:10

front engineer, which is the discipline of writing that fleeting language, plain English code that goes into these machines to get them to do what we want them to do.

Christopher Penn 6:21

There is no one better qualified conceptually, conceptually, to be a prompt engineer than a skilled writer.

So the WGA is position should be, we’re going to be doing the prompt engineering, as well as supervise, I think that would give the studios the ability to use the technology to reduce the time to output right to speed up production without eliminating WGA writer positions.

And WGA writers will be able to supervise and put a leash on AI without outright saying Nope, it’s not allowed here, because the reality is, it’s already in the technical studios, and those who use they’re just gonna march right ahead.

And they will use the strike as an excuse to say, well, I don’t have human rights, so we’ll just use the machine lens, and that’s not a position that’s going to help consumers.

If we’re gonna get retreads of retreads, it’s not going to help writers.

And in the long term, it might not help the studio’s either because they will be more profitable with it, but the content won’t be as good.

So, again, he worked for the WGA.

If you are a union member, please feel free to send this along to the leadership.

If you work in entertainment, and you’ve had spot some obvious flaws in my argument, please let me know, leave in the comments or whatever.

But that’s, that’s where things stand right now.

And

Christopher Penn 7:43

I think there’s going to be a sea change accelerated by the strike, so hopefully folks can adopt a more nuanced position on AI sooner rather than later.

Thanks for watching.

Talk to you soon.

Christopher Penn 7:57

If you’d like this video, go ahead and hit that subscribe button.


You might also enjoy:


Want to read more like this from Christopher Penn? Get updates here:

subscribe to my newsletter here


AI for Marketers Book
Take my Generative AI for Marketers course!

Analytics for Marketers Discussion Group
Join my Analytics for Marketers Slack Group!