Google should have the needed tech for good AI transcription, why the don't integrate them in their auto-captioning? and instead the offer those crappy auto subtitles
YT is using USM, which is supposed to be their SOTA ASR model. Gemini have much better linguistic knowledge, but it's likely prohibitively expensive to be used on all YT videos uploaded everyday. But this "correction" approach seems to be a nice cost-effective methodology to apply LLM indeed.
Are they crappy though? Most of the time it gets things right, even if they aren't as accurate as a human. And sure, they probably have better techniques for this, but are they cost-effective to run at YouTube-scale? I think their current solution is good enough for most purposes, even if it isn't perfect
I'm watching YouTube videos with subtitles for my wife, who doesn't speak English. For videos on basic topics where people speak clear, unaccented English, they work fine (i.e. you usually get what people are saying). If the topic is in any way unusual, the recording quality is poor, or people have accents, the results very quickly turn into a garbled mess that is incomprehensible at best, and misleading (i.e. the subtitles seem coherent, but are wrong) at worst.
I always wondered How YouTube distribute videos l, it's the smoothest video platform, even on when I had crappy internet it worked fine also not all.platform works well in south America.The closest to YouTube it's Netflix but lacks behinds a lot.
They have cache servers deployed into the ISP networks close to your home. The ISPs allow them to do this because they also benefit from it: Their bottleneck is the connection from their own network into the wider internet backbone, and having cache servers for the big CDNs takes a huge chunk of load away from that bottleneck. Here's documentation about a similar setup for Netflix: https://openconnect.netflix.com/
This is not a universal answer and I think deserves some correction.
1. Majority of ISPs do not host any cache for Google content
2. Credible ISPs do not have bottlenecks at the transit or peering level
3. Netflix makes use of much more local caching but their model works very differently to Youtubes
4. The concept of "internet backbone" does not really translate to reality. Peering is significantly more mesh-like than that, and transit more diverse.
I think that the main difference is that unix tools are intended to work which each other. Therefore needing less line of code. As opposed on other systems.
That was exactly part of the point I was trying to make with the article. The other, that I left unsaid but should probably add explicitly, is that Slack et al. run on a server with a kernel, but choose to ignore the access control capabilities of said kernel and instead chose to reimplement them. I think it is a shame, and more software should strive to be security agnostic instead of reimplementing access control for the umpteenth time.
