Simon @spoltier@qoto.org

The definition of madness is constantly misattributing quotes to Albert Einstein.

Extremely typical of @analog_ashley that she is capable of giving a talk in a bar that's both a full stand-up comedy routine that has people shouting with laughter yet also teaching how to engage with open datasets in neuroscience

Lord do I wish tech conferences had more talks like this

Costa Blanca was cool ✌️

Gotta come back to Costa Blanca another time to climb some of the cool mountains and big rocks they have 🥺 it’s like NM but with vineyards and proper trains, 1000/10

I posted about brain cancer and hope to find people to pick up the torch on my accessibility projects that I think could help make the web better. I have a major surgery May 29th and time is short. I will only become less able.

Summaries and links can be found here: https://github.com/alt-text-org/in-need-of-adoption

I'm happy to answer any questions, or just discuss things, but I do ask that kind words on cancer not be posted here, despite my appreciation for them.

Boosts and sharing elsewhere appreciated.

💜 Hannah

Thought about hypothesis testing as an approach to doing science. Not sure if new, would be interested if it's already been discussed. Basically, hypothesis testing is inefficient because you can only get 1 bit of information per experiment at most.

In practice, much less on average. If the hypothesis is not rejected you get close to 0 bits, and if it is rejected it's not even 1 bit because there's a chance the experiment is wrong.

One way to think about this is error signals. In machine learning we do much better if we can have a gradient than just a correct/false signal. How do you design science to maximise the information content of the error signal?

In modelling I think you can partly do that by conducting detailed parameters sweeps and model comparisons. More generally, I think you want to maximise the gain in "understanding" the model behaviour, in some sense.

This is very different to using a model to fit existing data (0 bits per study) or make a prediction (at most 1 bit per model+experiment). I think it might be more compatible with thinking of modelling as conceptual play.

I feel like both experimentalists and modellers do this when given the freedom to do so, but when they impose a particular philosophy of hypothesis testing on each other (grant and publication review), this gets lost.

Incidentally this is also exactly the problem with our traditional publication system that only gives you 1 bit of information about a paper (that it was accepted), rather than giving a richer, open system of peer feedback.

#science #philosophy

I think I've mentioned this before, but I will again because I keep seeing it. Why is "plan a vacation" on every list of cool things #AI could do?

Not even necessarily from people selling AI vacation planning. Just In general, when listing neat stuff you could do with AI agents it seems to come up.

Who are you people that take so many vacations that planning them needs automation??? Planning an itinerary is as close as I usually get to a travel vacation. Don't take it away from me 😂

New paper (in open access): "Philosophy of cognitive science in the age of deep learning" – in which I argue that although progress in DL is largely driven by engineering goals, it is far from irrelevant to (the philosophy of) cog sci, and vice versa.

https://wires.onlinelibrary.wiley.com/doi/10.1002/wcs.1684

Sending out immaculate datacenter vibes for your Thursday - with the correct aspect ratio 😉

I'm not sure how to turn these into coherent thoughts right now. I can't help but think about it in capitalist terms. And I know a lot of people hate that. We have an opportunity to create our own structures to do the kind of business that we find less objectionable. Things that we're happy to pay for. Exchanges that feel like value to all parties involved. But it really does require a change in our current expectations. And it's tough to find people who are ready to think differently.

this is not #AI

this is not a scene from a movie

this is simply a one in a billion video shot at the right time at the right angle by a teenager in #portugal a few days ago

fucking amazing! positively biblical

#space #meteor

https://www.news.com.au/technology/science/space/blue-meteor-falls-through-the-sky-over-spain-and-portugal/news-story/7bd39d794d6a5e79feade5723e4d4787

I paid 0.526USD/Hour on AWS and only get 2 FPS.

I rent a GPU on vast.ai for less than 0.5USD/Hour, and I got 12.5 FPS.

Today's lesson: Do not trust AWS.

I have resumed blogging. Here is the first for over a year.

https://paul.bone.id.au/blog/2024/02/13/poisoning-firefox-memory/

Hold me accountable to write more stuff.

I've spent the last several days trying to get a local mastodon build running on my laptop. I've gotten pretty far, but I have yet to see an actual UI in my browser.

So far, I've been trying out various attempts to run it in docker. That *should* be easier than trying to get the various components set up myself. But for some reason it's not. Getting mastodon configured properly is pretty complex. Getting everything right "out of the box" feels pretty error prone.

Did you know of any Granovetter diagram jokes? I need to find someone who knows one.

#CapabilityBasedSecurityHumor

Does anyone remember that mug that said "metadata kills" that was popular on here a while ago?

@Kensan @jnfrd @spoltier @kaoudis could possibly choose a language often used for medieval writing like brahmic scripts to be "write-only" lol

Surveillance Fédéral: the saga continues. Es ist viel sagend bei welchen Punkten die zuständigen Stellen sagen sie sind nicht zuständig und auf andere verweisen. Bravo @adfichter fürs dran bleiben. Diese Thematik verdient grössere Aufmerksamkeit!
https://chaos.social/@adfichter/112477506342997487

@jnfrd @spoltier @kaoudis @hipsterelectron What about Perl? It becomes unintelligible when you put enough of it in one place…