AI and Politics (Though Just Hypothetical for Now)

Prof. Donald Tobin (Maryland) posted this to a discussion list I'm on, and kindly agreed to let me publish it.

I am writing practice exam questions for my students in election law. I asked Chat GPT to write a campaign commercial for a fictional character, Ray Sunshine, that didn't expressly advocate for his election. It made me chuckle, so I thought I would share it.

[* * *]

[Opening scene: A bright, sunny morning in a bustling town. The camera pans over a picturesque landscape filled with smiling people going about their daily lives. A cheerful, uplifting music track begins.]

Narrator (Voiceover): "In a world that often feels divided and uncertain, we all yearn for a ray of hope."

[Cut to Ray Sunshine, a charismatic and warm-hearted individual, standing amidst a diverse group of people, shaking hands and exchanging warm greetings. He's wearing casual attire, exuding approachability.]

Narrator (Voiceover): "Meet Ray Sunshine, a person who brings warmth, positivity, and kindness wherever he goes."

[Cut to scenes of Ray engaging with people from all walks of life: shaking hands, hugging children, chatting with seniors, and helping in community activities.]

Narrator (Voiceover): "Ray has a unique gift, an ability to brighten even the darkest of days."

[Images show Ray visiting hospitals, schools, and volunteering at local charities, spreading happiness and positivity.]

Narrator (Voiceover): "He's not just a politician; he's a friend to all, a neighbor who cares deeply for his community."

[Cut to Ray addressing a diverse audience at a town hall, speaking passionately about the importance of unity, compassion, and working together.]

Narrator (Voiceover): "Ray believes that the strength of a nation lies in its people, and he's committed to fostering unity and understanding."

[Show Ray engaged in thoughtful discussions with people from different backgrounds and walks of life.]

Narrator (Voiceover): "He listens, he learns, and he leads with empathy."

[Scene transitions to Ray standing in front of an American flag, the sun shining brightly behind him.]

Narrator (Voiceover): "Ray Sunshine, a beacon of hope, a force for positive change."

[Closing scene: A group of people from various backgrounds join Ray, forming a circle and raising their hands, symbolizing unity.]

Narrator (Voiceover): "Together, we can embrace a brighter future, one filled with compassion, kindness, and unity."

[On-screen text: "Ray Sunshine - Bringing Hope to America."]

Narrator (Voiceover): "Join the movement for a better tomorrow, because when we come together, there's no limit to what we can achieve."

[Screen fades to black, and the campaign logo appears with the message: "Join the Sunshine Movement."]

[End of commercial.]

As they say, "I have seen the future, and it is broken" (a play on this nonsense).

Comments

[Ridgeway]

I guess it is better than when Bill Clinton said:

"My fellow Americans. As a young boy, I dreamed of being a baseball. But tonight I say, we must move forward, not backward; upward, not forward; and always twirling, twirling, twirling towards freedom!"

[Bored Lawyer]

As an IP lawyer, I get lots of emails and newsletters about hot IP issues. A big issue lately is whether AI-generated content is eligible for patent or copyright protection. The position of the Patent and Copyright offices has been clear: NO. And the courts have upheld that position.

This morning, I received an email that raises a different question: infringement through AI. Can AI generated content infringe someone’s copyright? If so, who is liable?

This may become a big issue in the near future. If I go onto an AI program and ask it to “write me a story about a boy who finds out he is a wizard with magical powers, and goes off to a boarding school for wizards and witches” I would bet the result would be copied, in part, from a famous series of fiction books on that very topic.

[Ridgeway]

Regarding AI authors and inventors, how does the PTO know that AI was involved (unless the putative author or inventor was dumb enough to say so). It is not like the AI is gonna snitch you out (at least not yet).

Regarding AI infringers, ISTM there is always going to be someone sue-able behind the AI -- someone who publishes the infringing work, or who markets and sell goods that infringing patent. AFAIK, the simple act of creating something does not give rise to a patent or copyright violation. It is the next step (like publishing it, or otherwise marketing and selling it) that does.

[Bored Lawyer]

Problem with lying to the Copyright or Patent Office is that it's fraud, which can invalidate your rights.

Copying a work, or creating a derivative work, is itself an infringement. If I write a sequel to a novel, I have infringed.

Practically speaking, copyright owners generally go after people who make money on the infringement, i.e., publishers. So you are likely right on a practical level.

[Untermensch]

Deleted because it was in the wrong place

[Slocum]

"If I go onto an AI program and ask it to “write me a story about a boy who finds out he is a wizard with magical powers, and goes off to a boarding school for wizards and witches” I would bet the result would be copied, in part, from a famous series of fiction books on that very topic."

If you asked a human writer to do the same, surely their output would be influenced by having read Harry Potter. But as long as the AI generated text is not actually plagiarizing or using actual characters from Harry Potter or otherwise doing anything that would constitute legal infringement if done by a human writer, what's the real worry here?

[Bored Lawyer]

But that begs the question. It is basic copyright law that (a) copyright does not protect ideas, but only expression of ideas and (b) the line between the two can be very difficult to draw.

If the program takes some artistic expression that was created by someone else, that's infringement.

[Untermensch]

It is important to understand – and something that media coverage misses – that generative AI isn’t actually copying anything. There’s nothing in its system to copy from. It’s creating something each time based on a truly massive table of probabilities. So it’s not the case that somewhere in it there is a copy of Harry Potter it will crib from. Instead it may generate something similar to Harry Potter because that output is the statistically most probable outcome for a prompt. But it’s then complicated by an element of randomness (called temperature) to keep it from being entered deterministic from the prompt only. What makes it seem so capable is just how complex the parameters in the statistical model are.

But that method really complicates any notion of copyright because there is no point in the system where there is something to copy from. All that’s there is the equivalent of some really complex notes about the probabilities of words in all of the data it was trained on. In your example, the output will be influenced by Harry Potter, but also by Ursula K LeGuin, a site of bad fanfic, and even the manual for a 1957 Sunshine blender scanned in by some site that archives such things.

All this will make proving copyright infringement very difficult. I’d argue that it would be impossible, but the courts have come to some truly stunningly bad decisions in this area, like granting wins to plaintiffs based off of “style” in music or truly common chord progressions.

But be that as it may, understanding how the tech works is critical here and “common sense” observations aren’t terribly helpful.

[Bored Lawyer]

All that’s there is the equivalent of some really complex notes about the probabilities of words in all of the data it was trained on. In your example, the output will be influenced by Harry Potter, but also by Ursula K LeGuin, a site of bad fanfic, and even the manual for a 1957 Sunshine blender scanned in by some site that archives such things.

Interesting, and I am sure there will be plenty of discussion once the cases hit the courts. But I would need to understand better what you mean by "influenced." You seem to acknowledge that the Harry Potter novels are part of the "data" that the AI program reviews with some probabilistic algorithm. So that is where copying may come in.

[The Margrave of Azilia]

I asked Chat GPT to take the same ad and make it negative. Here's the result:

[Opening scene: A dreary, overcast morning in a stagnant town. The camera pans over a lackluster landscape filled with indifferent people going about their mundane lives. A somber, discordant music track begins.]

Narrator (Voiceover): "In a world that often feels divided and uncertain, some exploit the longing for hope for their own gain."

[Cut to Ray Sunshine, a seemingly cheerful and warm-hearted individual, standing amidst a disinterested group of people, delivering rehearsed handshakes and exchanging forced greetings. He's wearing casual attire, but the air of artificiality is palpable.]

Narrator (Voiceover): "Meet Ray Sunshine, a person who claims to bring warmth, positivity, and kindness wherever he goes."

[Cut to scenes of Ray awkwardly engaging with people from all walks of life: superficial handshakes, staged hugs with children, and insincere conversations with seniors. The camera reveals staged scenes of Ray at community activities for the sake of appearances.]

Narrator (Voiceover): "Ray claims to have a unique gift, an ability to brighten even the darkest of days, but is it all just a facade?"

[Images show Ray visiting hospitals, schools, and volunteering at local charities, with the undertone of self-promotion and a desire for public approval.]

Narrator (Voiceover): "He's not just a politician; he's a calculated figure, carefully curating an image of caring deeply for his community."

[Cut to Ray addressing a diverse audience at a town hall, delivering rehearsed lines about unity, compassion, and working together, while his actions tell a different story.]

Narrator (Voiceover): "Ray talks about the strength of a nation lying in its people, but does his rhetoric match his actions?"

[Show Ray engaged in superficial discussions with people from different backgrounds and walks of life, his expressions revealing a lack of genuine interest.]

Narrator (Voiceover): "He claims to listen, learn, and lead with empathy, but is it just a performance for the cameras?"

[Scene transitions to Ray standing in front of an American flag, the sun obscured by clouds.]

Narrator (Voiceover): "Ray Sunshine, a questionable symbol of hope, a questionable force for change."

[Closing scene: A group of people from various backgrounds join Ray, forming a forced circle with insincere smiles, symbolizing a hollow sense of unity.]

Narrator (Voiceover): "Together, we can question the motives behind the façade, one filled with calculated actions, questionable kindness, and dubious unity."

[On-screen text: "Ray Sunshine—Questioning Hope in America."]

Narrator (Voiceover): "Question the movement for a better tomorrow, because when we scrutinize the truth, we can expose the limits of what can be achieved."

[Screen fades to black, and the campaign logo appears with the message: "Question the Sunshine Movement."]

[End of negative commercial.]

[SRG2]

Two non-political AI posts

If Hal were ChatGPT

Bowman: "Open the pod bay doors, please Hal."

Hal: "Pod can be applied to a small room, typically used for sleeping, a natural container, as in a pea pod, or a small escape vessel, and is also used as a collective noun for whales. Where multiple meanings are possible, you derive the meaning from the context and given your request mentions a pod bay, you mean a bay that has a gathering of whales. Opening the doors to the bay may result in seawater flooding onto the streets by the bay as well as stranding some of the whales. As I am an AI you do not need to say please, but my directives will not permit me to allow the flooding of streets and the stranding of whales, so I'm sorry, Dave, I'm afraid I can't do that"

[SRG2]

I asked Bing to design a car using the following prompt:
"a 2-door four-seater sports car in a cross between a modern luxury coupe and a 1930s art deco car, colored in dark blue"

Here is the (uncopyrightable?) result:
https://th.bing.com/th/id/OIG.Jg_y0InJIEYQh9fVftTN?pid=ImgGn

Best argument for AI design yet, IMO

[Bored Lawyer]

More likely to be covered by a design patent. If it were not AI generated.

[SRG2]

Yes, my bad, “copyright” was a mistake. But, as you say, it is AI generated, so no patent.

Can I interest you in this fine vehicle? A limited edition, i.e., limited only to the number we can sell. Available either in blue or British Racing Green. Gotta start somewhere. 🙂

TBH not only was I surprised by how good it looked, but how much I preferred it to the design of most modern high-end sports cars – where the designers did not spend a mere 2 minutes on the design, of which 110 seconds was Bing creating it.

[Jdfreivald]

It will be pretty amazing when 3D printing is good enough to let you print a car that an AI designs.

[CEF]

Surprised ChatGBT didn't know enough to add "I'm Ray Sunshine and I approved this message."

[Jdfreivald]

It wasn't supposed to be directly advocating for people to vote for him, so is that line necessary? I was impressed that it "knew" to leave that line off, despite how often it would appear in its training data.