flawn

* open-weights

(sorry for nitpicking, i think it's important to emphasize - only OLMo has been the most most prominent fully open source release)

Also tokens/sec are not only dependent on parameter size, so that is kind of a brittle theory.

Then all this 3x amount would have had no ROI, China would have outdid the US comparatively.

I mean, I validate them by opening my window and seeing what the floor is. For me around 430ish usually - inline with what's in the atmosphere globally.

Additionally, I'd imagine quantization to have more side-effects than just slightly lower performance (on whatever task). You are basically removing information, and that information could be by chance what the model needs to fulfill it exactly the way you'd want to do - although it's still fully capable. I am not sure if this is really different from "lower performance" but open to hear your opinions.

Dumb question here - how can I manage effectively edges across the whole world without the huge maintenance overhead? Which tools would be recommended for that? I e.g. have a VPS at Hetzner with Coolify but users from the US have high latency. I wouldn't know how to not use CloudFlare?

Or for 100$, get a 5x increase on all failure points - maximum vibes, maximum excitement.

This does not play a role - even if you lock your bootloader Play Integrity Checks still fails, and that means you can't use certain apps, MDM and overall restricts your usage. Thank Google for that.

Social Media is actually one problem, and that's not just cellphones. I don't disagree with the premise that this could be all a smokescreen by the UK spying on citizens but Social Media is a huge issue. If you are keen, read up on the intersection of Epistemology, Sociology and Social Media research.

I wonder, how well does Yak-Shaving work for you with AI, how does it look for you specifically and how do you make sure it's not undermining the friction for learning things properly? I want to try some more yak-shaving soon too.

They are part of the latest YC Batch. //s

Not sure if there is any in-depth analysis across more countries about this pattern, but also in this votum, densely populated urban areas had a clear stance against this votum, whereas the majority of votes have come from rural areas. So I am not as certain as you are whether the trend extrapolates as outlined - I rather imagine that the polarization will increase even further across these two segments, with lived and experienced realities drifting apart.

What is the evolutionary advantage of this? I mean, if the host dies subsequently that's pretty bad for both parties, or?

That's wild misinformation. There was an outdated military database at play, and not just Claude. It doesn't exclude AI interference of course but your statement is just not correct.

Let's abstract this further: It's about the user's existing power and intentions, meaning if I am already in a position of power, AI will multiply it to levels way beyond a peasant could. Power dynamics just get exacerbated.

Then check V-Dem, you might argue they're flawed as well but then I'd suggest you to provide counterexamples for why the US should be considered a functioning democracy, and is not on the way to a fully authoritarian state.

The CCP knows, whatever the heck this technology will bring with itself, the current power dynamic inside of the country is on their side, and AI will solidify it.

I hypothesize that, rather than slowly having it disperse in society and allow people to harness it in ways they don't want, they might as well accelerate everything until AI becomes the totalitarian swiss knife - which they can make use of in the best way of course.

Let's see what will happen.

Owe a lot to them.

If you have kiddos in Germany: jugendhackt.org

https://github.com/tombonez/noTunes

this is the OG app.

Good old times, same for me. Just ended up not bothering and deactivating any additional services - digital detox time ＼ ( ° o ° ) ／

The fact that you hear this often (also not only from the victims of this) should signal that, even without quotes, there is likely some truth to it.

If you check blog.archive.*, a russian landing page with a warning that the page has been blocked by government authorities comes up. Maybe related?

If you sell a blackbox that you constructed yourself, then you are also liable for anything that happens.

If you sell a blackbox from a third-party (e.g. from China), you are liable for somebody else's decisions that you cannot scrutinize.

So, that's kind of the argumentation that underlies sovereignty and why Chinese Models are not being used in critical infrastructure.

The issue is that you wouldn't be able to even transparently get to any evidence, as these models are blackboxes.

They might start scheming behind employees backs as soon as they realize they are being used in critical infrastructure of adversaries. And nobody would know until it's too late.

That would mean regulation of social media companies seems appropriate.

That would mean regulation of social media companies seems appropriate.

That would mean regulation of social media companies seems appropriate.

Corruption is like cancer.

And as if there isn't a middle-way here to have short-term contractors _not_ from large consultancies.

Or rethinking approaches, and doing such work via OSS, and paying maintainers to keep code up to date, which France has been doing iirc.

Latest macOS does have an integrated local LLM shipped on device! Not SOTA but something.

I would assume, because the title and parts of the book suggest reducing human connections to utility and personal gain.

I am biased towards 67