But all the tricks of developing a solid software stack to support the HW are already out no? The basic principles are there, to my understanding the main challenge is not doing this development in tandem with the HW people and the requirments to support older legacy device which makes it harder for example for Amd to compete. The only challenges,which intel are prepped to face is logistics and fabs. On a separate note, project like JAX are aiming to circumvent that abstraction layer cuda adds to nvidia, so having decent hardware competition is definitely an option. Just some time ago, vllm fully supported amd gpus! We need more competition.

It's definitely not gdpr compliant, because the identifier is still unique. Any data collected per session identifier for user interaction with the website are bound to user consent and not covered under gdpr legitimate interest.

I think captchas are just another lind of defense to make it harder for actors abusing the system. It's not a solution, just a little (getting outdated) fortification.

This was a fun read, very ignobel spirit!

Thanks!

It's one of those "to motivate the horse to run 1% faster, you add shit ton of weight on top of it" strategy.

Are we talking about NN Taleb? I am curious about the twitter persona.

Someone by the name of V. Minakhin. They have an irrational hatred of Bayesian statistics. He blocked me on twitter for pointing out his claim about significant companies do not use Bayesian methods is contradicted by the fact that I work for one of those companies and use Bayesian methods.

Netflix uses Bayesian methods all over the place. In a meeting presenting new methods, I called squinting at A/B test results and considering them in the context of prior knowledge "shoot-from-the-hip cowboy Bayes". This eventually lead to a Cowboy Bayes T-shirt, hat and all.

What's the selling point for it though? I don't get it?

Not to hijack's OP great work, but when you say 85% you mean true positives? How about the false positives?

My prompting is conservative to err on the side of playing an ad if there is a chance it might be part of the actual content, not really getting false positives at all yet. That being said while still in development I haven't reached the stage of running on a huge collection of podcasts to get more representative statistics.

I think the accuracy of my prompt/llm is also ~85%. I've got a collection 2500+ podcast episode transcripts (English language) with ads I'm going to try and analyze shortly to find out if I'm missing any ads, or tagging some falsely.

I remember learning about strides 10 years ago. Never gotten into gpu parallel programming, but that course informed a lot of my understanding about how GPUs worked.

I think there is a basis for such worry, "Why Nation Fails" highlight the aspects of corruption and non-independent/transparent judiciary system and absue of police force as a big reasons why countries fail. And if you want to compare western police force corruption to others, I invite you to check what happened in egypt 10 yo.