very cool. how are you creating the control vectors? curious since the word "cold" can be both a conversational disposition and a temperature (same word)
thanks! we asked the model to generate some synonyms and antonyms (in this case, we have "cold" and "impassive" vs "affectionate" and "sensitive")
Then, we ask the model to behave that way (with a prompt), and store the difference in activations for each pair. Then, a PCA can be used to extract the principal component, giving use the steering vector. We do most of this using the repeng library, and the author goes into a bit more detail on how it's done on her [blog](https://vgel.me/posts/representation-engineering/#How_do_we_...?)
this is super cool. do you think the primary demand driver for your sales will be 1) gov't mandates, or 2) cattle farmers wanting to differentiate their meat to eco-conscious customers / retailers? just curious about how you see the market.
> we are seeing early indications in research of some amazing additional benefits – improved milk production, increased immunity, improved food conversion ratio (meaning you can feed cattle less and have them pack on more protein)
These "additional benefits" are all things that would increase profit - reduced loss of output from sickness; increased production; reduced feed consumption.
If the seaweed is cheap enough, it could be worth it for farmers to use without additional incentives. Though, they would also get green cred "for free."
If you've ever wanted audio features that directly represent semitones (or quarter tones!) this is the package for you.