Math is off. Total number of people detained in 2023 was 270k out of 66million annual visitors. 0.4% chance. Chances of dying in a plane crash are roughly 1 in million (Google tells me so, not sure of the source). The ones I listed are the publicized ones but the story is the same.
Every MLE who didnt study Math really likes to downplay its importance. Yeah you dont need measure theoretic probability, but you need a grasp of Lin Alg to structure your computations better. Remember the normalization that we do in attention ? That has a math justification. So I guess yeah academics did have a role in building LLMs.
I mean computer scientists really do like to pretend like they invented the whole field. Whereas in reality the average OS, compilers, networks class has nothing to do with core ML. But of course are also important and these barbs dont get us anywhere.
I think at this point you should be more worried about being thrown in a detention centre. But the likelihood of being affected gun violence in be US especially if you stick to the right parts is low as long as you're there for a short while.
YCs success has nothing to do with just YC (not to take credit away) but rather from the fact that the US is a large unified well-regulated market. EU has sensible rules sure but the market isnt unified and the individual countries dont have big enough markets to sustain large numbers of software companies. YC itself is honestly quite open with their knowledge but its just that EU in itself is a pretty terrible place to run a business. Lots if regulation, some good ones.
Honestly most people want to be him. Everbody craves money and unchecked power. The problem is nobody understands that what he's doing is against their interests.
This is a little bit of a word soup. Its hard to see why the various redesigns were done without a set of requirements. I dont get why you'd trigger Airflow workflows for doing CDC. These things were designed for large scale batch jobs rather than doing CDC on a some Google sheets. Either way without scale numbers its hard to why PG was used or why the shift to BigQuery. Anyways the site uses Hugo, which actually sticks out for me.
Congratulations on the release. Although I hope the developers take the lesson that AI frameworks are unnecessary. You don't need frameworks to write HTTP calls. Just a good enough SDK would do.
Not the essay for me. I actually lost patience midway. The author lost me at "when we are young we dream of meeting our idols." I never gave a shit. I mean my idols were fictional so that obviously helped.
Its not fast without pre-segmentation as they do in WhisperX. It actually has terrible transcription speed. For speedup we have to use Ctranslate2 kernels. The decoding code is also a mess where its hard to plug your own custom language model. Not to mention streaming ASR requires even more tweaks. Whisper Small is very fast and quite inaccurate. If you deploy whisper on a GPU which costs around dollar per hr, you really to ensure that the cost savings are worth it.
Although all of this is from a production lens. For personal use, honestly nothing is as easy to use as Whisper (even works on a laptop).
Their interviews were one of the easiest things I ever did. Pay 150bucks for Leetcode, sort by company, memorize the solution and show up.
Even when solving problems in the real world, nobody sits in a dark room until moment of genius strikes. We always try some tactics others have tried and only then innovate.
I like t-shirt estimates but only when we all agree we known wtf we are talking about. I also really like hacks. They get the job done and not everything needs layers and layers of leaky abstraction. I also quite like easy things like Config fixes, esp if they get escalated. You can leverage these to show impact and get a good rapport going.
I hate meetings without agenda. Honestly send me a message and I'll make sure I reply. Meetings with written agenda are great.