Amen to that! 20 years later this was my gateway drug into being addicted to computers and gave me my full stack understanding I still use every day at work. <3 Gentoo and the friendly geek at the coffee place I worked at when I was 14 who gave me my first hit and held my hand through what is effectively mental torture for most people
So a similar thing happened near some friends' in France. A military jet crashed into the forest near their house, but the air force couldn't figure out where it had gone. Eventually a farmer noticed that a new pond had appeared on his land. The jet made enough of a crater when crashing that drained the nearby swamp and created a new pond deep enough to conceal the full fuselage, thus completely hiding the airplane. Once the farmer alerted the air force, they were able to crane the remains out of the newly formed pond and recover the key parts of hardware onboard. Had the farmer not noticed the change in landscape they might have never found it. https://www.lefigaro.fr/flash-actu/2011/03/02/97001-20110302...
Ah the good old "it's not me it's the test" argument. These systems are not just next token predictors, they learn complex algorithms and can perform general computation, its just so happens that by asking them to next-token predict the internet they learn a bunch of smart ways to compress everything, potentially in a way similar to how we might use a general concept to avoid memorizing a lookup table. Please have a look at https://arxiv.org/pdf/2211.15661 and https://mobile.twitter.com/DimitrisPapail/status/16208344092.... We don't understand everything that's going on yet but it would be foolish to discount anything at this stage, or to state much of anything with any degree of confidence (and that stands for both sides of the opinion spectrum). Also these systems aren't exposed to the real world today, but this will be untrue very soon https://ai.googleblog.com/2023/03/palm-e-embodied-multimodal...
Now imagine the other side of this equation, where pandas seems too clunky, behold YOLOPandas https://pypi.org/project/yolopandas/ i.e. `df.llm.query("What item is the least expensive?")`
This is what pains me the most about the US. Out of sight and out of mind but these poorer parts of society represent xx millions of fellow humans. If you could have similar standard of living in Australia where the poorer parts of society perhaps have more accessible healthcare and better social safety nets, then why the hell can't we do that in the US!? (I know the answers to that question, but it just makes me sad and angry).
Source? My anecdotal experience is that in the US people without much income will simply avoid seeing a doctor until things get Serious because of cost, which makes the quality of care somewhat irrelevant (and my other anecdotal experience was that Uk hospital quality was quite good, and as you mentioned, free).
Honestly I think I may be even slightly more productive, but whatever, I make more than I can spend, I don't think any amount of money would make me work 5 days/week (I mean that quite literally, unless it was maybe for a couple months, and in the 1e6 ballpark, my time on earth is just too valuable to me).
Sounds like that was simply not actually a 4-day workweek company! My experience working 4-day workweeks (Thursday evening desktop goes off, corp phone gets left at desk) is brilliant, especially mixed with full remote, I would likely give up more than 20% of my salary to have this permanent 3-day weekend! (Shh, don't tell HR)
Also interested, I've always wondered what "the book" would be to learn to make a nice proper industrial control cabinet (PLCs, wiring, components etc.)
Indeed! Sorry, I was thinking more about the layers but of course JAX is way more than numpy on steroids. (Although it is also that: https://dionhaefner.github.io/2021/12/supercharged-high-reso...). JAX has a very nice vmap for easy parallelization on SIMD accelerators, and pmap even allows cross-device vectorization with a single line which is just beautiful !
Give it another try, I found the docs pretty good, you need to get your head around XLA tracing, and read "the sharp bits" section and you should be pretty set!
As a researcher in RL & ML in a big industry lab, I would say most of my colleagues are moving to JAX [https://github.com/google/jax], which this article kind of ignores. JAX is XLA-accelerated NumPy, it's cool beyond just machine learning, but only provides low-level linear algebra abstractions. However you can put something like Haiku [https://github.com/deepmind/dm-haiku] or Flax [https://github.com/google/flax] on top of it and get what the cool kids are using :)