7734128

Sure, gyros alone are not doing that, no.

They are correctly using position. The dead reckoning a phone does over its accelerometers absolutely yields a position. It would of course drift quickly, but it would know its position relative to where it was was a while ago.

It's basically what VR headsets and controllers do, sans their periodic error correction from cameras or lighthouses.

Is this garbage actually using a broken Google Drive link as their source?

Could we please ban this tabloid, and the user who posted this?

That's not a "not good guys" event. That's a "lifetime in prison for all people involved" event.

Because the RLVR probably uses formalization to judge rollouts.

At the very least diplomatic condemnation and a South African level sanction.

On the day that it happened both Egypt and the US confirmed that Egypt had informed Isreal in advance.

https://www.bbc.com/news/world-middle-east-67082047

This article is from a few days after that, but I remember that there were confirmations already during the day.

I don't think appealing to how good a country is at breeding is a valid counter to its genocidal tendencies. That's probably the strangest argument I've ever seen.

You don't need a counter example to a counter example.

It's not that I think the opposite of the original statement is true, I'm suggesting that's it's just all noise as there are both successes and failures at both extremes.

Bell Labs would be a good counter example. All the resources in the world and no demands.

However, it's Ubisoft and therefore tainted.

Anyone can write a blog and they can write anything they want there. There's no more value in those links than in a Reddit comment.

No, they do not.

There are many people who deserve to feel ashamed for what they do for a living.

A lot of people should be shunned for what they do to others, regardless of whether it's for making money or for other purposes.

It's pretty much the biggest complaint of Claude compared to its competitors, so they really should adress it .

It's not a regret. Android would never have been popular in the first place if it had not been open source. If phone makers had been able to anticipate how much the demons at Google would be able to lock down the Android then they would never had used the OS back in ~2009.

90% of screen door effect disappeared when we switched from pen-tile OLED to ordinary RGB matrix.

The remaining 10% still remained with the HP Reverb and the Pico 4 which I've used quite a bit, so it's not a solved issue but it's a world of difference from the old Vive or Quest 1 if that's your reference frame.

("pancake lenses" is a similar upgrade over the old fresnel lenses)

I'm sorry but this is garbage. It's just a cobbled together mess using LangGraph and other overly complex libraries.

You can't teach fundamentals by just calling libraries.

If they were not using actual Norton trademarks I would have thought this was a parody.

Why would anyone use Norton? It's just malware.

They "burn" a lot when they do benchmarks, while these runs can become valid roll outs for training. Perhaps less efficient than other data creation, but hardly burned in the same way.

I'm not sure how viable it still is. Perhaps it's still possible, and perhas that's exacly what they did in wtich case my objection falls, but I don't know.

They would know the specific thoughts released by the "stolen thought" paper, which became part of the public internet on August 10.

Unfortunately those are the only thought examples you can use to perform this experiment, as no other are availible.

But as the model should have seen those specific examples, it's not a good signal that Qwen was exfiltrating thinking traces.

This is probably the 20th time the german wiki is posted here.

It's not a new incident, nor a new discovery of an old incident.

Sure, but TFA is trying to use Qwen's reaction to the thoughts as proof that they did indeed extract thoughts to train on.

My point is that any model trained after August 10 will know of those specific thoughts.

The problem with this is obviously that the only GPT 5.5 thoughts that we have access to are from stolen thought.

Qwen 3.8 0902 was trained after the release of the paper on August 10, so it should have seen those specific thoughts.

How the hell could that cookie consent form ever be legal? Fuck them.

The correct response should be to actively disregard all IP owned by corporations participating in that behavior.

It seems that all competitors are rushing to release before Astra, which would suggest that they think it's going to be major.

Practically free for "contributors" at 0.2 usd/mtok. That's going to be hard to say no to for hobbyists.

They're absolutely right – and this is is why it's a load bearing observation that cuts to the heart of the issue.

There have been many blogs like this over the last years.

Yes, embeddings are computationally heavy, but they are not at all complicated and they provide a lot of benefit.

90% of "document" based RAG projects should view semantic search with embeddings as their primary method.

It's very powerful and so easy to implement that you could try it out and discover whether performance would be an issue rather than trying to anticipate it.