WotC Hasbro CEO optimistic about AI in D&D and MTG’s future

[J.Quondam]

I'm with you on this, I'd much rather have a tool that spits out an answer rather than having to sort through a number of links. Chatgpt is great for this sort of thing.

I'd like both, personally. A summary is nice, but a list of links to the information it used is nice to have - even necessary - for deeper study and verification. Basically give me a bibliography when it generates a report.

 

[overgeeked]

I'm with you on this, I'd much rather have a tool that spits out an answer rather than having to sort through a number of links. Chatgpt is great for this sort of thing.

Too bad it's wrong so much. Getting the wrong answer quickly is worthless. Worse than worthless. It gives the recipient a false sense of the answer being correct despite it being false. Taking a few minutes to get the right answer is a good investment.

 

[Thomas Shey]

It also is good at hallucinating "answers" to things (as are the creators of many random webpages to be fair).

That's the biggest problem with it currently from my POV.

 

[cbwjm]

I'd like both, personally. A summary is nice, but a list of links to the information it used is nice to have - even necessary - for deeper study and verification. Basically give me a bibliography when it generates a report.

Wouldn't be surprised if it gets there at some point. I did ask where it got the information and it mentioned various books. Had to specify edition otherwise I think it just takes from the latest edition, or possibly it creates a mishmash of editions.

 

[Yaarel]

With regard to AI, the predictive models for what AI will do in the near future are grounded in hard science.

Pretty much all of it relates to Moores Law and its extrapolations. The Law itself is simple. In the context of electronic equipment, Moore in 1965 noticed that number of transistors on an electronic chip doubles about every two years. Moores Law continues to remain true today. Meanwhile, new technologies bypass the limitations of earlier technologies. This means that the amount of information computation is accelerating geometrically. Information processing is getting faster and faster and faster.

Here is a pretty good graph from Wikipedia, of Moores Law, last updated 2019. It shows the rate of acceleration. The microchips upto today 2024 remain on track at this same rate.

Visually this graph is linear, but look to the left at the numbers. More than merely geometric, the graph shows the information processing is ... accelerating ... at orders of magnitudes! Faster and faster and faster.

Correlating with this information processing, is every kind of human information. Every branch of science is accelerating at this same rate.

This acceleration of technology is relentless. Unstoppable.


Next, is a graph that depicts the implications of the acceleration of technology. Notice roughly 2025, a computer is about as smart as a human, in other words the Turing Test. Roughly 2050, a computer is more intelligent than every human on the planet put together.

AI is unstoppable.

We can ethically "steer" how the power of AI gets used.

However it is IMPOSSIBLE to slow AI down.

 

[Cadence]

With regard to AI, the predictive models for what AI will do in the near future are grounded in hard science.

Pretty much all of it relates to Moores Law and its extrapolations. The Law itself is simple. In the context of electronic equipment, Moore in 1965 noticed that number of transistors on an electronic chip doubles about every two years. Moores Law continues to remain true today. Meanwhile, new technologies bypass the limitations of earlier technologies. This means that the amount of information computation is accelerating geometrically. Information processing is getting faster and faster and faster.

Here is a pretty good graph from Wikipedia, of Moores Law, last updated 2019. It shows the rate of acceleration. The microchips upto today 2024 remain on track at this same rate.

Visually this graph is linear, but look to the left at the numbers. More than merely geometric, the graph shows the information processing is ... accelerating ... at orders of magnitudes! Faster and faster and faster.

Correlating with this information processing, is every kind of human information. Every branch of science is accelerating at this same rate.

This acceleration of technology is relentless. Unstoppable.


Next, is a graph that depicts the implications of the acceleration of technology. Notice roughly 2025, a computer is about as smart as a human, in other words the Turing Test. Roughly 2050, a computer is more intelligent than every human on the planet put together.

AI is unstoppable.

We can ethically "steer" how the power of AI gets used.

However it is IMPOSSIBLE to slow AI down.

The Soviets had a way in Frank Miller's The Dark Knight Returns iirc, at least for us - and for everyone if they had wanted to.

 

[cbwjm]

Too bad it's wrong so much. Getting the wrong answer quickly is worthless. Worse than worthless. It gives the recipient a false sense of the answer being correct despite it being false. Taking a few minutes to get the right answer is a good investment.

I dunno, I asked it questions I knew the answer to and it got them correct, granted these weren't obscure topics, the more obscure the NPC, for instance, the less likely youll get a correct answer. However, I think if WotC created a tool directly trained on all of their setting and adventure information, then those errors would be far less likely to happen and it would be an excellent tool for the table or DM prep, or even for players trying to adapt a character to one of the settings.

 

[Cadence]

In my news feed. Those companies making me feel good about their ethics again.

In Cringe Video, OpenAI CTO Says She Doesn’t Know Where Sora’s Training Data Came From

Wondering what data OpenAI used to train its buzzy new text-to-video AI? OpenAI CTO Mira Murati seems to be wondering, too.

futurism.com

 

[Cadence]

I dunno, I asked it questions I knew the answer to and it got them correct, granted these weren't obscure topics, the morw obscure the NPC, for instance, the less likely youll get a correct answer. However, I think if WotC created a tool directly trained on all of their setting and adventure information, then those errors would be far less likely to happen and it would be an excellent tool for the table or DM prep, or even for players trying to adapt a character to one of the settings.

LLMs don't know correct from incorrect, they know what should sound good, right?

On the other hand, making a tool that tried to parse your sentence and figure out which particular thing you wanted searched for and giving it to you is a different thing and would prevent the making stuff up. (And at worst I guess could be like a search engine that was bad with prompts). That doesn't seem much different than a search engine to me at first blush.

 

[overgeeked]

I dunno, I asked it questions I knew the answer to and it got them correct, granted these weren't obscure topics, the more obscure the NPC, for instance, the less likely youll get a correct answer. However, I think if WotC created a tool directly trained on all of their setting and adventure information, then those errors would be far less likely to happen and it would be an excellent tool for the table or DM prep, or even for players trying to adapt a character to one of the settings.

Here’s three non-obscure flubs I’ve seen. When asked about a famous painting by a famous artist, it gave me a whole paragraph about the subject’s face…despite there being no face in the painting. When asked about the career of a famous sports figure, it utterly flubbed the dates. When asked about the mechanics of just about any RPG except D&D, it gets them about half right and fills in D&D mechanics for the rest. These are all trivial and non-obscure. Things easily found on wikipedia, which these programs were likely trained on. I wouldn’t trust these programs with anything actually important. And they certainly aren’t a replacement for actually looking things up and reading about them.