But also, stop telling people how good it is up north, the southerners will ruin it
Posts by Arun Niranjan
How on earth do you make it up north the exact week I go on holiday?
If you can, go it to roker beach and fausto coffee shop, especially if you have the dog with you
I couldnβt wait to get back to my wild rogue rampage of hitching rides on bot-piloted transports and watching my serials.
Turns out Karl Friston was actually on social media the whole time π
If you're still dichotomising continuous variables, stop.
Dichotomising your exposure &/or outcome wastes power.
Dichotomising confounders introduces residual confounding.
If a variable has non-linear effects, use splines!
If you want 'clinically relevant' threshold effects, use marginalisation.
That's a decent expectation - that priors get sharper with more experience - but it's an assumption that's fairly easy to violate.
Also see any song by the killers or muse
The first one absolutely, the second one depends on your model of the world and the priors in it IMO
And a below inflation pay rise hurts like a kick up the posterior
The classic example I like to use to illustrate this is that I expect a pay rise every year, and if I don't get one I am disappointed.
My prior is on the trajectory, not the absolute value. And I don't like it when I get data that contradicts my prior π
Leave it long enough and you'll end up in a pickle
Previous active inference formulations focused on stationary distributions, but more recent path based formulations handle dynamic environments much better.
Whether they actually correspond to human behaviour in my opinion the jury is still out
Rob I did not expect active inference to come up in a uk politics thread! ππΎ
I would challenge that we are default biased away from change - we are default biased by our priors, which could be for a changing environment (but they don't have to be)
It would be nice if they interviewed 1400 people who don't have a house but would like one
Do you usually wake up around 7am? If so that's perfectly normal!
I highly recommend this book if you haven't read it already
www.waterstones.com/book/think-l...
I finally thought I understood the Monty Hall problem (kinda) and then I saw it in DAG form and now I'm pretty sure I neither understand the Monty Hall problem nor DAGs.
Please stop with the new year's resolution π
At some point I need to actually read the methods sections of your papers π
It makes sense if:
- you might want to swap one of those sub-functions out for another (e.g. a different algorithm implementation)
- it makes it easier to write tests (likely but not guaranteed)
But I think it really needs to be an extended period of time, one where it's long enough that there is actually a system in place to continue your work whilst you're gone, rather than you just be anxious that there will just be a larger pile waiting for you when you return
If you can, I highly recommend taking a sabbatical.
I scheduled a mini mid-life crisis to avoid having a larger unscheduled one π
Double-take detail in this interview with Edith Heard, head of the Crick
Her French husband decided against getting a spouse visa because of the UK's "intrusive" immigration rules
If that's happening with one of the UK's most senior scientists, how many others are being put off completely?
No I am currently in general habitat building mode now that we are in the post capitalism era
WhatsApp messages describing my rather capitalist approach to playing pokopia
So I have finished the main tasks in Pokopia, this chat sums up my approach, and yes we build the wall to set us free that's why we build the wall we build the wall to set us free
These KOs are what keep me on social media, thank you for sharing
I was literally just about to recommend just, glad you got there first!
It probably doesn't even need to be common advice, it could literally be on a single Reddit post that comes up in a tool search π€·πΎββοΈ
I think I can offer a partial explanation, as yer resident anti-AI understander.
Programming is wizardry: you go to school, you read the books, you learn the arcane languages.
Using AI is like being a warlock, making a deal with a demon to do the things you canβt do. And who knows the cost?
This is something I used to do - intellectually someone might know that self-harm is bad, yet they still feel an overwhelming urge to do it.
Swapping to physical discomfort was (for me) a way of tricking my own mind into accepting the urge was satisfied without doing permanent damage
Only one whisky away from starting your own podcast