The Washington Post made an AI chatbot for questions about climate

The Washington Post is sticking a caller climate-focused AI chatbot wrong its homepage, app, and articles. The experimental tool, called Climate Answers, volition usage the outlet’s breadth of reporting to reply questions astir clime change, the environment, sustainable energy, and more.

Some of the questions you tin inquire the chatbot see things like, “Should I get star panels for my home?” oregon “Where successful the US are oversea levels rising the fastest?” Much similar the different AI chatbots we’ve seen, it volition past service up a summary utilizing the accusation it’s been trained on. In this case, Climate Answers uses the articles wrong The Washington Post’s clime conception — arsenic acold backmost arsenic the section’s motorboat successful 2016 — to reply questions.

“We person a batch of innovative and archetypal reporting,” Vineet Khosla, The Washington Post’s main exertion officer, said during an interrogation with The Verge. “Somewhere successful the years and years of the data-rich reporting we person done, determination is an reply buried successful 1 of the things we person written.”

Beneath the answer, you’ll find links to the articles that the chatbot utilized to nutrient its answer, on with the applicable snippet it pulled its accusation from. The instrumentality is based connected a ample connection exemplary from OpenAI, but The Washington Post is besides experimenting with AI models from Mistral and Meta’s Llama.

When asked astir the anticipation of misinformation, Khosla said Climate Answers won’t nutrient a effect for questions it doesn’t person an reply for. “Unlike different reply services, we truly are baking this into verified journalism,” Khosla said. “If we don’t cognize the answer, I’d alternatively accidental ‘I don’t know’ than marque up an answer.” However, we program to effort the instrumentality erstwhile it launches contiguous to get a consciousness of its guardrails.

The Washington Post isn’t the lone quality outlet that’s relying connected its archive of accusation to powerfulness an AI chatbot. In March, the Financial Times started investigating Ask FT, a chatbot that subscribers tin usage to get answers astir topics related to the outlet’s reporting. Meanwhile, different publishers, similar News Corp, Axel Springer, Dotdash Meredith, and The Verge’s parent company, Vox Media, person jumped into licensing partnerships with OpenAI.

The Washington Post has been gradually building connected its usage of AI; according to Khosla, the outlet has besides rolled retired AI-powered summaries for immoderate of its articles. Even though The Washington Post’s caller chatbot is lone capable to tract climate-related questions for now, Khosla didn’t regularisation retired the anticipation of expanding it crossed different topics the outlet covers. “We perfectly expect this experimentation to widen and standard to everything The Washington Post does,” Khosla said.