A word about Claude

We should have started experimenting sooner with alternatives to ChatGPT. It’ll be a few episodes before we get to the Claude section of PastMaster Season 3, but I assure you it’s worth the wait.

On the face of it Claude, made by a San Francisco-based outfit called Anthropic, does the same thing as OpenAI’s GPT. It’s a large language model. But they do differ. GPT can do more, with features like image generation, while Claude apparently has a much larger context window, which means it can process more data at once. Some people say it has been programmed to be more conversational than GPT. I know that ‘personality’ is an illusion with AI like this, but based on our experience so far, GPT is like the detail obsessed big sister who organises the house party and will call your parents if things get out of hand, while Claude is the loose cannon who turns up drunk, tells an inappropriate dirty story and then downs a bottle of vodka and trashes the place before passing out in the corner.

Our GPT GameMasters have had the feel of a dotty aunt who’s trying best to entertain her nieces and nephews but sometimes gets in a muddle. Claude is a different beast – we’ve played two games so far with Claude at the helm and both times ‘he’ has put himself front and centre of the story, seemingly unable to resist the limelight, whether in a medieval occult dungeon or a feudal Japanese tavern. It’s the first time we’ve actually had a fight with our GameMaster.

I won’t give away any spoilers here, but seriously, keep an eye out for the Claude episodes when they drop this summer.

Have you had any similar experiences with Claude or any of the other large language models like Bard or GPT? What’s you take on it? Let us know - email [email protected] or message us on socials.