Ugglan @ugglan@pinkroom.biz

Kind of having my mind blown by embeddings based semantic search.

I have an archive of The Economist's The Intelligence podcast transcribed using Whisper.cpp. Indexed all March'23 episodes using @simon 's #llm (https://simonwillison.net/2023/Sep/4/llm-embeddings/)

Searched the index for "Tesla" - and the first hit is the 20th Mar episode that doesn't mention the word "Tesla" even once but is all about mobility that mentions cars and electric cars several times.

https://stitcher2.acast.com/livestitches/f904bd5e459f111404303194ccc6a784.mp3?aid=641836ce3883b200100b559a&chid=d556eb54-6160-4c85-95f4-47d9f5216c49&ci=sEcwO_y58RmiNfiWNfu5yDZ4-rUWHN5nzM-06dzbbXMOkuhnH_nzfg%3D%3D&pf=rss&sv=sphinx%401.178.0&uid=66d039322df320fb2be0692ddcc5c994&Expires=1694285457956&Key-Pair-Id=K38CTQXUSD0VVB&Signature=C4~l5fGNMSFI56rSniqEIdcv1E56EmB19JcCnmwsXikcD9XKVvLXg2OXah0WpW1HwM0qCWt2Rlb3o0eHttl6Qufi3csuI3dTKAdXWm1qOfXuskJyLCUf2PdS-SqHbSEjPEHIoKQe9P2omRV6cxJq6IX-9hArLfyMjvoiZQXs7JREKGzmedsWkHJc6M8iVhhIhGvrwQxjd3SjqxiViZCHBPZiGpWQeVWSfurKeWoy~os8-ji6~5ekFX3WYuj9NB0uAfSVQYu3mcgywncPW93XSHdUK6rpEQ7pQVNcHqYFE37R7ykGoyL4z38dG0tL8uF97ZM2B10xYFtoWPa1o6IrLQ__ (~11:30 mark depending on ads)

You know what I want AI and robots to do? Sort trash so we can just dump everything in one bin and not be confused and paralyzed every time we throw something way.

Is there a simple video or document that could explain to an idiot how this service works? #toot #mastodon #confused