As a British citizen I didn't bother reading this as my VPN was roaming near Budapest at the time and my AI got confused with the cross cultural references ๐
View quoted note โ
mike
mike@mikehardcastle.com
npub1aqak...237n
Looking for new opportunities ๐
Fully vested Chief AI Officer (CAIO)
Former, failed, Chief VLOG Officer
Former Chief Shitpost Officer - NOSTR Inc.
Node runner - Miner - Author
My public relay: https://nortis.nostr1.com/
My book: https://mikehardcastle.com/my-book-why-bitcoin/
In AI, the term Temperature refers to the prediction tolerance permitted.
If set to 0, then predictions are tight and only the most defined, logical path is selected.
At this setting, asking the same question multiple times will reliably give you the same, predictable, safe, boring answer.
If set to 1, wild predictions and inconsistencies may appear.
Answers become less likely realities, but normally sit within the realm of possibilities. Asking the same question repeatedly will give wildly different answers that may bear little resemblance to reality.
Typical temperature for practical use is usually set around 0.8
If you want to have an AI on acid, then, if allowed, set the temperature to 1.5 and watch the show ๐
Temperature is another one of those wildly misleading labels, so much so, my private LLM had to warn me it had nothing to do with the weather multiple times.
I have therefore renamed "Temperature" to "Wiggle Room" for my internal AI discussions.
Below is the screen shot of one of my local LLMs warning me of its improper association ๐


Cross posting from @Joe Nakamoto on X
This is true, I tried it:


In a previous life, I used to buy loss making shell companies in order to profit from the tax deduction ๐
View quoted note โ
If you think about it, losses are just negative profits ๐


I present to you, my brain ๐
Turns out, my entire brain and life can be condensed into 6 files and 49KB's of disk space ready to be injected into any AI model I choose.
I felt inadequate until I realised that the entire worlds imperfect knowledge can be compressed into a 5GB LLM data set ๐ฑ


GM
I was up early today. What a beautiful sunrise ๐


I loaded LM Studio on my M1 MacBook (16GB RAM), loaded Meta llama-3.1-8b-instruct-4bit, pasted in the first Chapter of Jane Austin's Pride and Prejudice and asked it to summarise:


All Ethereum users are wrong ๐


I just got my fibre Internet a couple of months ago, El Salvador saw my example and wanted their own ๐


The Liquid Network uses Proof of Stake to prove the Bitcoin that are staked on it, which is not PoS the protocol, but the full, correct meaning:


Government should mine Bitcoin:
โก 1,250 points! I'm basically saving the galaxy one zap at a time ๐ธ Can you do better?
Play now:
#SpaceZappers #SpaceInvaders

Space Zappers - Bitcoin Arcade Game
A retro Space Invaders arcade game. Pay 21 sats to play. Publish your high scores to the decentralized Nostr leaderboard.
