Oddbean new post about | logout
 More reasons not to trust ChatGPT.. 🤦🏼 nostr:note1hlsm29lrea8hfqh38aucten3x04lujv4yejuqcl95tcg56p0tv8sxzfdkd 
 It’s actually 2.1 gazillion smh 
 ChatGPT got basic math wrong, and just apologized to me like no big deal?? Lyn Alden, my Bitcoiner crush, had to correct me…  WTF just happened. https://i.nostr.build/3IcwyXrLlijJD4Qo.jpg  
 Claude gets it right. Cause Claude is the goat 😂. 
 People overestimate chatgpt. It's a language model. It sometimes gives the impression to be smart, but it's just guessing words based on probabilities. And math is not a concept it's good at. At all. 
 But the word ‘million’… wow 
 Perplexity Pro uses Wolfram Alpha for math 
 LLMs are shockingly bad at math 
 It’s been doing great with all my daughter’s math, including long story problems.. Such BS. All the AI hype for this? 
 My experience is that they’re usually good with word problems or when there’s an explanation, but when you just ask it to solve a math problem they struggle. Had more luck with Claude than ChatGPT 
 It’s clearly going to take everyone’s job better start UBI now 
 😂 
 DYOR! 
 It’s notoriously (relatively) bad at math. Which from my perspective as a math teacher is really interesting. At one point, like a year ago, Stephen wolfram was talking about integrating wolfram alpha so it would recognize a math problem, send it to wolfram alpha, then return the result. In the screenshots I saw it even link to the wolfram alpha computations , effectively showing its work.