Can't comment on the model without knowing the details, but seems like you're on the right track with some exciting data.
Just be sure to be skeptical. Ya know, doctors notes on x-ray training data and all.
Yeah, looks a lot cleaner, no?
When / if you write it up, just be sure to discuss how you computed chance.
There is more than one way to do so.
Depends on the rubric
LLM hype aside, this is a fair point.
50% accuracy is chance for two classes, but say you have a multinomial classifier with 4 equal classes, chance can fall to 25%.
Never hurts to draw a thin plt.axhline across the plot to indicate where the chance level is to mitigate such misinterpretations.
I agree.
But to answer your question, persistently malleable memories may come at the cost of noise and intererence. In fact, there's pretty good evidence that that's what happens, no?
Maybe different animals, but perhaps philosophy asks questions we just don't have the tools to empirically study yet 😉
I see. Interesting take and thank you for the reference. I'll take a look.
And in your view, you don't see knowledge as a form of truth? (justified true belief, etc. Or if Truth is ultimately metaphysical, the only attainable form of truth we could hope for?)
I'd be really surprised if the answer doesn't involve the enterprise of science in some form or another... How could it not?
Good arguments, but then where does 'truth' derive from?
New paper! Out now in Neuropsychologia: “Sparsity and memory constraints interact with training sequence to bias learning of associative maps” with @dalezhou.bsky.social*, @kwcooper.bsky.social, @lovecrabmeat.bsky.social, and @aaronbornstein.bsky.social. authors.elsevier.com/a/1mipv6TBG9...
🧵:
@dieworkwear.bsky.social
You know what's urgently needed? Basic software. Reliable database systems, accessible interfaces, privacy oriented design. It's super mundane stuff like asking, "how can software make your life better", not telling people "this is how software is going to solve all your problems." Because it won't.
Congratulations!!!
Indeed, that has been their strategy for some time...
www.heritage.org/education/ev...
www.nytimes.com/2025/04/11/p...
christopherrufo.com/p/how-trump-...
and sadly, many many more...
Imagine if all Government data access was that simple...
"[...] This process can be described most appropriately as a memory reorganization process, as it seems likely that consolidation and forgetting of spatial memories never ends."
13. And anyone who has been part of this grand humanist tradition is someone who will not lie down and cower before these sad bullies.
I don't know what happens to science in the US over the next four years, but I have no doubt as to where the arc of history bends.
We aren't going anywhere.
fin
A photo of a globe logo, a circle, with an infinity sign in the middle. The text is ContinualAI UnConference October 19th 2023, Free for everyone
One week away from our celebration of Continual learning across brains and machines!
View our exciting line up of speakers and events here: unconf.continualai.org/program/
Oh, and did I mention it's FREE?
Please re-post to get the word out, and share widely with your networks!
💯 🔥
Howdy! I'm Keiland!
🧠🐀 🤖 🖥 ♾️ = 🧑🔬
I'm somewhere between a cognitive scientist and a neuroscientist studying learning and memory across humans, animals, and machines.
Ephys, engineering, and a lot of computation.
I also confounded the ContinualAI non-profit
#neuroscience #neuroskyence #AI
Exactly. We don't read with our eyes, but rather our memories. There is a far wider range of how media is interpreted, based on unique expirences, than folks give credit for.
A photo of the moon
