Yes, but this updates weights and you have to have the reward model / loss function.
06.03.2026 20:19
👍 2
🔁 0
💬 1
📌 0
Yes, but this updates weights and you have to have the reward model / loss function.
Richard catching strays!
Yes, and the users of the tool are likely updating their prompt (to modify the tool’s behavior) as they review the results and iterate on their search. (My guess is that the model artifact itself is static and the prompt is what has changed, to the extent this matters.)
Certainly possible in principle. I'm not aware of any deployed system like that. You'd have to define a loss that you're using to backpropagate to update those weights.
bsky.app/profile/buil...
dj don t da queenz boi:
A driver backing up a vehicle struck a person riding a scooter in San Francisco’s North Beach neighborhood Thursday.
How fast was this driver backing up that THIS is the state of the car? (source: www.reddit.com/r/sanfrancis... )
DJT to the stove:
I don’t currently got that dawg in me.
Point of clarification, AI tools are trained, but when you’re deploying/using them, they’re no longer learning.
In Woke 2, this is the energy we’re gonna bring.
bsky.app/profile/buil...
If your Mark be Biblical, you're invincible
If your Mark be Wayne, he's fuckin on ur plane
"🔜"
The Ceaușescus are being recognized more and more.
- Tronald Rumsfeld
If your Ryan be Grim, the wits may be dim.
Magnitude 4.3 Earthquake Strikes Gerāsh, Iran - USGS
This would be pretty unnerving if you were in the area.
The American people yearn for Donald Trump’s “forever” war.
That's probably what, a $24,000 burst from the C-RAM?
It's an F-15 so I assume there are just 12 left.
Evil, done in my name, funded by my taxes.
INDOPACOM in disarray.
Seconding this question.
Yeah if their purpose were to make the government more efficient or effective, which it was not, that would have been nice.
Guy on the right is like a skinny Windom Earle
