Scary new Frieds
Chapter 1 - The chatbots obviously deserve #1 in our list. Grappling with AI chatbots is unquestionably leading into the fundamental potential of Artificial Intelligence. So far we have been shying away from probing too deep into the capabilities of LLM’s and focused on finding practical applications to help with our day to day work as architects and planners. Chatbots are giving probabilistic answers based on specific training data. In planning however we require specific and often precise metric results and not vague approximations.
Our Tests
▶️ Terms and Conditions
One of the first - and for us most conclusive - evaluation we have executed back in January 2024. In this first assessment we have established, that none of the services we reviewed would accept liability for the answers generated. In that sense we can assess that AI will struggle with taking over any engineering profession. In the end it is accountability that forms the basis of our work as architects and planners. And formulating answers that might be right but that need checking for context and consistency are generally of little use in our line of business.
▶️ Building Regulations
Beyond this very general approach at the beginning of 2024 we can report of an interesting conversation in July 2024. The question was how the usable area is to be calculated in the attic under the pitch of the roof. Struck had a quick answer, pointing out that it can only quote from the general building regulations and that at the given location other regulations might apply. Christina had searched the web the conventional way and noted across the studio that she had downloaded the relevant paragraph from the local regulations. Claude gave confidently a fully detailed report from the local regulation only to turn apologetic, when asked which paragraph this information was taken from. Only by cross-checking against the real source it was able to determine, that the quoted paragraph as well as the presented calculation method were a mashup of various sources. Concluding this accidental test: Perplexity was able to quote the correct paragraph as well as the correct method of calculation at the first attempt.