AI chatbot errors intentional, claims developer
- eppursimuove
- 10 minutes ago
- 2 min read
The errors sometimes made by AI chatbots are a deliberate feature, according to a spokesman from Uberdork, which produces the Gerard app.
”Sure, these things have access to the whole internet, along with tools to help them distinguish trustworthy sites from Fox News. In theory, they could be 100% reliable.
“But we also want them to be relatable, like having a chat with a friend. And we all have that friend who may know a fair amount about some things, but doesn’t admit it when he comes to the limit of what he knows - maybe doesn’t even realise he has - but just keeps barrelling on, making stuff up to look clever.
”That’s why we called ours Gerard, and I guess why Anthropic chose the name Claude. They both suggest someone a bit full of himself who might bullshit just to impress you, so you should double check everything they say.”
Far from being a problem, the spokesman continued, Uberdork’s market research shows that, in today’s fragmented media landscape, many people find chatbots too accurate.
“People are used to getting news that panders to their prejudices, and don’t like a chatbot that sets them straight. That’s why we’re going to produce one for MAGA types called Billybob, which will give them nothing but Trump’s view of the world.
“The difficulty will be keeping it up to date, so you’re always getting the latest version of what the Iran war was intended to achieve, and how we can tell the US has won it. Again.”
Why not join our WhatsApp channel, NewsBiscuit Daily Posts? You can access the latest output from NewsBiscuit and share posts easily with friends, family and colleagues using WhatsApp. Get some biscuits in your pocket (sorry about the crumbs, though) by clicking here./