Hasbro CEO Chris Cocks Is Talking About AI in D&D Again

Chris Cocks, the CEO of Hasbro, is talking about the usage of AI in Dungeons & Dragons again. In a recent interview with Semafor, Cocks once again brought up potential usage of AI in D&D and other Hasbro brands. Cocks described himself as an "AI bull" and offered up a potential subscription service that uses AI to enrich D&D campaigns as a way to integrate AI. The full section of Semafor's interview is below:

Smartphone screens are not the toy industry’s only technology challenge. Cocks uses artificial intelligence tools to generate storylines, art, and voices for his D&D characters and hails AI as “a great leveler for user-generated content.”

Current AI platforms are failing to reward creators for their work, “but I think that’s solvable,” he says, describing himself as “an AI bull” who believes the technology will extend the reach of Hasbro’s brands. That could include subscription services letting other Dungeon Masters enrich their D&D campaigns, or offerings to let parents customize Peppa Pig animations. “It’s supercharging fandom,” he says, “and I think that’s just net good for the brand.”

The D&D design team and others involved with D&D at Wizards of the Coast have repeatedly stood by a statement posted back in 2023 that said that D&D was made by humans for humans. The full, official stance on AI in D&D by the D&D team can be found below.

For 50 years, D&D has been built on the innovation, ingenuity, and hard work of talented people who sculpt a beautiful, creative game. That isn't changing. Our internal guidelines remain the same with regards to artificial intelligence tools: We require artists, writers, and creatives contributing to the D&D TTRPG to refrain from using AI generative tools to create final D&D products. We work with some of the most talented artists and creatives in the world, and we believe those people are what makes D&D great.

 

[beaumontsebos]

For extremely specific tasks.

It can be used to "generate storylines, art, and voices" for an RPG. Note: I don't think AI produces writing and art at a higher quality than humans (yet). But it is cheaper, faster and more productive.

 

[Mistwell]

All the RPG books and novels for D&D have been available on the internet. Anything scanned got OCRed. The question of course is, was ChatGPT version x trained on that? But what about all the AI/LLM models? And even IF none were trained on that particular information, I expect that someone else will train an AI/LLM model on that before WotC does...

I disagree it's all OCRed. Having seen some scans of TSR stuff, most of it is pretty sketchy looking and not easily read by AI, particularly for training purposes not something they'd scan.

In addition, if your concern about AI is unethically taking authors properties without permission and using them to make money off them, then the ethical use would be what WOTC likely plans. Which is to take only materials they paid for and own under contract with the creators for this use. That has value.

 

[Whizbang Dustyboots]

It can be used to "generate storylines, art, and voices" for an RPG.

Eh, Cocks also claims he regularly plays D&D with 30 people. I would take anything he says with an enormous grain of salt.

Note: I don't think AI produces writing and art at a higher quality than humans (yet). But it is cheaper, faster and more productive.

I'm not sure I can agree that more writing of low quality is the same as productive.

I just had to go through four or five web pages to find out some information, because both Google's and DuckDuckGo's AI on their search pages was clearly hallucinating making stuff up.

Garbage writing that has to be immediately discarded (or that leads users badly astray) is not writing that's actually useful to anyone except for the people selling the AI subscriptions.

I do think voice actors and podcasters should be very nervous, though, as the technology there is already very good and very cheap.

 

[beaumontsebos]

Eh, Cocks also claims he regularly plays D&D with 30 people. I would take anything he says with an enormous grain of salt.

I'm not sure I can agree that more writing of low quality is the same as productive.

I just had to go through four or five web pages to find out some information, because both Google's and DuckDuckGo's AI on their search pages was clearly hallucinating making stuff up.

Garbage writing that has to be immediately discarded (or that leads users badly astray) is not writing that's actually useful to anyone except for the people selling the AI subscriptions.

I do think voice actors and podcasters should be very nervous, though, as the technology there is already very good and very cheap.

Maybe, for some reason I have yet to fathom, corporations in our uber-capitalist economy won't embrace a technology that is cheaper, faster and more productive than humans - even if the quality (right now) is comparatively low. I hope you're right. But judging how corporations have evolved, I doubt it.

 

[Clint_L]

So, this is the kind of thing I use generative AI for. I'll then go in a tweak/alter/replace these as needed, but as a starting point, this saves me a ton of work:

Can you create a random encounter generator, with brief descriptions of 20 encounters, suitable for a party of 4th level adventurers (D&D 5e), set at sea on the Lucidian Ocean of Matt Mercer's Exandria

ChatGPT said:Here's a random encounter generator for a 4th-level party sailing the Lucidian Ocean in Exandria. Each encounter is designed to be quick and dynamic, with some offering combat, intrigue, or environmental challenges.​

---

Lucidian Ocean Random Encounter Table (Roll 1d20)​

d20	Encounter Name	Description
1	Sahuagin Raid	A band of sahuagin emerges from the depths, attempting to board and seize the ship. They bear a bloodstained banner with an unknown crest.
2	Phantom Schooner	A spectral ship appears in the mist, crewed by ghostly sailors. They seem lost, seeking guidance—or revenge.
3	The Stormcaller’s Wrath	A violent storm descends suddenly. A ghostly voice within the winds calls out to the party by name.
4	Shipwrecked Survivor	A desperate figure clings to driftwood—an exiled Clovis Concord smuggler who knows too much.
5	Kraken’s Offspring	A juvenile kraken briefly rises, testing the ship's defenses before vanishing into the abyss.
6	Pirate Pursuit	A ship flying the black flag of the Revelry appears on the horizon, closing in fast. Their demands? Surrender or die.
7	The Sea Speaks	The water churns unnaturally, whispering secrets of the leviathans below—or perhaps, something else listens back.
8	Merrow Ambush	A pod of merrow lays in wait beneath the waves, ready to strike with harpoon traps and magic.
9	Sunken Vault	A galleon wreck juts from a reef. A sealed treasure chest glows faintly beneath the waves, guarded by something unseen.
10	Siren’s Song	Beautiful voices echo across the water. A trio of sirens in disguise offer companionship, but their true hunger soon emerges.
11	Giant Sea Turtle	A massive sea turtle, with an overgrown coral shell, offers safe passage—but only in exchange for a rare offering.
12	Elemental Surge	A rift to the Plane of Water opens briefly, releasing an angry water elemental seeking vengeance against all landwalkers.
13	Fog of the Forsaken	A thick magical fog surrounds the ship, causing visions of lost loved ones to appear before the crew.
14	The Silver Serpent	A legendary sea serpent passes beneath the ship, watching, waiting. It does not attack—yet.
15	Shark Frenzy	A feeding frenzy of bloodthirsty sharks circles the ship. Something—or someone—lurks among them.
16	Marid’s Favor	A regal marid (water genie) appears, offering a cryptic prophecy in exchange for a gift of great personal value.
17	Lost Cargo	A floating crate drifts by, marked with a Concord sigil. Inside: contraband, or a trapped mimicwaiting for prey?
18	Whale Watchers	A pod of playful whales follows the ship, but something massive looms behind them in the deep.
19	Abyssal Rift	A sudden whirlpool pulls the ship toward an opening in the ocean floor. Are those demonic eyes staring up?
20	Messenger of Uk’otoa	A strange, golden-eyed revelry cultist is found floating on wreckage. He whispers one word: “Uk’otoa…” before vanishing in a blink.

 

[Cergorach]

I disagree it's all OCRed. Having seen some scans of TSR stuff, most of it is pretty sketchy looking and not easily read by AI, particularly for training purposes not something they'd scan.

A LOT of the TSR stuff was done and OCRed 20-30 years ago by a company working for WotC (Jim Butler ran that, might even have been Bastion Press), what was missing has been done since. We had the Dragon Archive. Probably all the novels in epub format. The big issues were the 3e/4e books, but those have already been done and available via DTRPG.

And if you think that's difficult to convert to workable text, look at olmOCR.

Even back 20-30 years ago, the prices for those official D&D pdfs were very low, it started at $2.99 per product (no matter what size). I got all of them, made an index via Adobe Acrobat Pro and that was easily searchable. All the 2e books, all the Dragon magazines, Forgotten Realms Atlas, easily searchable and would link you to the right pdf and page for the sources.

 

[GnomeWorks]

So, this is the kind of thing I use generative AI for.

This sort of thing isn't the problem.

The problem is that the techbros have everyone convinced that these Chinese rooms can bootstrap themselves into producing outputs that they just ... can't.

I've literally heard people say "oh you just have to engineer your prompts better."

I saw a diagram a guy drew up of connecting an LLM to a Q-learning algorithm, and when I asked him why he would do that, his explanation was such that it was clear that he did not understand what either model could do, or why connecting them in the way he wanted wouldn't actually accomplish anything, and -- in fact -- completely lost the point of using Q-learning to begin with.

We've got people out in the wild who have no expertise in this field who think we can work literal magic, and then you've got people with no experience in it taking six-week bootcamps and thinking they're experts and know exactly what's going on to the point of questioning someone with advanced degrees in this crap. It's not just dangerous to people trying to do crazy things with LLMs: once it's obvious the techbros are absolute-- well, I can't say that string of words here. But once it's obvious to everyone that they've been selling snake oil, there's a decent chance it's going to completely screw up real, legitimate AI R&D.

 

[Cergorach]

I use r1 671b for writing read aloud texts from simple room descriptions (without asking for DM notes!). Example:

Certainly! Here's a read-aloud text for the room at the bottom of the well, blending atmosphere, hidden clues, and sensory details:

Read-Aloud Text:
"The dim light from above barely pierces the gloom of this square, dusty chamber. A thin layer of grit crunches underfoot—sand spilled from some long-forgotten hourglass or distant desert. The walls are cluttered with dented, rust-pitted shields, hung like trophies of battles no one remembers. Scrawled across them in jagged charcoal and knife-carved grooves are names, obscenities, and crude symbols: a grinning skull, a broken crown, and the words ‘Turn Back’ repeated like a mantra. To the south, a jagged tunnel slopes downward before twisting west into darkness. The air feels unnaturally still… save for a faint, cold draft whispering from the northern wall, where the mortar between stones looks just a little too clean."

Key Details Embedded:

1. Secret Door Hint: The "cold draft" and "clean mortar" subtly hint at the one-way secret door to the north for perceptive players.
2. Graffiti Clues: The symbols and phrases ("Turn Back," the skull, and crown) could tie to faction lore, a past adventuring party, or foreshadow dangers ahead.
3. Sensory Immersion: Gritty sand, rusted metal, and the oppressive stillness contrast with the chaotic energy of the tavern above.

DM Notes for Flexibility:

• If a player investigates the shields, you might reveal one bears a fresh scratch or a hidden compartment (if you want to add a loot hook).
• The "cold draft" could carry distant sounds (e.g., echoing voices, dripping water) to hint at what lies beyond the secret door.

Need adjustments or additional layers? Let me know!

I then run the read aloud text through ElevenLabs (Callum v1) to generate a sound file that reads this aloud for the players.
Next step would be to run the same read aloud text through the LLM again to generate a prompt for generating an image with Midjourney in a certain art style.

That's without the thinking process. The Thinking process is absolute gold for new DMs! Ideas, thought processes exposed on how to do things. I know how to do it, I just don't want to do it for all the rooms of Undermountain! And I don't mean the 5e adventure, I mean the big two boxed set version (with add ons)...
========================

I had some chats with our previous DM about my character via What's app about six months ago. It was in Dutch and English, lots of little snippets. Our new DM asked for our characters backgrounds, so I needed to rewrite what I had. I had planned to do that manually, but that fell through. I just asked the LLM to translate everything to English and make it a coherent whole. I asked it once to do it again and got better result, first one wasn't that bad, but asking again was free and I could always choose the first result if worse. It had some new twists and turns, sure it wasn't 100% what I had in mind, but that's the point. It doesn't have to be perfect, it needs to be good enough. That's imho how DMing works.

These are NOT magical instruments, they are tools, all tools (and humans) have limitations. Are these writing literature? No! But in many a case they do write better then many of the official D&D books. I'm a creative person, I can write these things myself, but we're all limited by time, and I feel I can spend my time better on ideas, details, encounters, or maps. And if I don't like a specific part, I can always rewrite it or regenerate it. The end result is that I can do WAY more with these tools then without them.

The quality of the promt and the priming of the LLM is imho also very important. It can make a night or day difference in results. And if you have the access, some of the underlying settings can also be very important. I see it as the art of googling, some people have it, some don't. Often don't knowing how to even formulate the question... It won't suddenly do all your house work, but depending on what you ask it, it can work very well. And not all LLMs do all the things equally well. I've been feeding the same series of assignments into all kinds of LLMs, noting the results for each and evaluating each LLM. For how I use it for D&D, DeepSeek r1 671b is the 'bomb'. If I were looking at coding, I would evaluating the different models again.

And regarding solo adventures (like those chose your own adventure novels) with LLM, I found this: AI Dungeon very impressive! Not really my thing, but fun to play around with a bit.

 

[Jensaarai]

There's just something kinda 'bleh' about using AI to come up with cool ideas and worldbuilding. There isn't a sense of personality or interesting perspective on the creator's thoughts, or how things can percolate in your own mind in fun/weird directions.

 

[Whizbang Dustyboots]

But once it's obvious to everyone that they've been selling snake oil, there's a decent chance it's going to completely screw up real, legitimate AI R&D.

Not to mention these folks will be trying to apply AI to serious, real-world problems that it simply is unable to handle, especially after they get rid of the data sets that AI would use to get an accurate picture of what's going on. I think we're going to see some serious disasters before everyone changes course.