Matt retweeted
just saw a tiktok that said 'pov you just found a hidden cafe inside a bookshop in cambridge' and it's the Waterstones cafe inside Waterstones.
Matt retweeted
guy on the street hollered "you got some nice legs" at me, so i looked back to give him the evil eye. it turned out to be a man in a wheelchair with no legs, and he said, "you never appreciate them untill their gone." Confused, i just said "i hear that" and kept walking.
Matt retweeted
not even funny how some Americans talk with such authority and confidence on a sport they’re watching for the first time. it’s like me watching my first Rugby Union match and wanting piss-drinking to be banned
Matt retweeted
[average cbt enjoyer when i'm 2 hours late to meet them at the pub]: he probably has adhd. that he managed to overcome his executive function and time blindness enough to even show up at all is a sign he cares. should be grateful
[me]: i'm da king of da playing football manager
Matt retweeted
In two to three years time, when Starmer and his government are no doubt deeply unpopular, I hope we in the media will ask ourselves: "Why were we so supine during the long 2024 election; why didn't we hold Labour properly to account while we could, and ask more probing questions, and explore their records, rather than give them such an easy ride?".
Matt retweeted
Soccer I love when they hold up the sign and a brand new beautiful boy takes the place of a dirty sweaty ruined one
Matt retweeted
I ran a script to replace every instance of 'whale' in Moby Dick with the word 'lesbian', and I have to tell you, Captain Ahab does not come out of it well
Replying to @ZackPolanski
Replace "billionaire" with "Jewish" in all Polanski's polemic and propaganda, and his antisemitism becomes clear.
It's all just Zackshit.
Matt retweeted
You have secured funding from a Chinese state industrial bank to launch a new battery factory, but the catch is you may never leave the special economic zone where your plant is located. Which special economic zone do you choose?
Matt retweeted
Was not prepared for that follow up info or who it was the kid was printing out pictures of lmao
Matt retweeted
oomf urgently called to inform us that his friend ate shit biking on the west side highway bike path and a man helped him up + asked if he was ok and when he looked up it was none other than mayor zohran mamdani on his citibike. closest you can get irl to being saved by spiderman
Matt retweeted
This is just like being alive in the 1600s when they got good at making complicated clocks and deduced that every complicated thing in the universe probably functioned exactly like a clock
There's a quadrillion-dollar question at the heart of AI: Why are humans so much more sample efficient compared to LLM? There are three possible answers:
1. Architecture and hyperparameters (aka transformer vs whatever ‘algo’ cortical columns are implementing)
2. Learning rule (backprop vs whatever brain is doing)
3. Reward function
@AdamMarblestone believes the answer is the reward function.
ML likes to use pretty simple loss functions, like cross-entropy. These are easy to work with.
But they might be too simple for sample-efficient learning.
Adam thinks that, in humans, the large number of highly specialised cells in the ‘lizard brain’ might actually be encoding information for sophisticated loss functions, used for ‘training’ in the more sophisticated areas like the cortex and amygdala.
Like: the human genome is barely 3 gigabytes (compare that to the TBs of parameters that encode frontier LLM weights). So how can it include all the information necessary to build highly intelligent learners? Well, if the key to sample-efficient learning resides in the loss function, even very complicated loss functions can still be expressed in a couple hundred lines of Python code.

































