Karpathy is biased when it comes to self driving. Example: You can't both have humans teleoperating like he claimed, AND have cones disabling Waymos. Waymo's mistakes such as one where they were going around a parking lot honking at each other tells you they don't have humans in the loop except in the most extreme cases. He's likely correct that it's not self driving 100% of the time, but what if it's 99.999% and in the 0.001% the humans have to tell the Waymo how to get out of a very tricky situation?
That's a strange position to take with such high certainty. Google has been talking about training on TPUs for a long time. Many ex and current employees have been on the record talking about how much nicer the Google internal training infra using TPUs is. GPU is an afterthought for Google's inference and non-existent in training.
Yep! A short summary of my n=1 findings (currently very speculative, but also sound I think, am meeting with my cardiologist next week to see what he thinks).
The type of ADHD I have seems to have an "autonomic nervous system impairment" component and a symptom profile overlapping with hyperadrenergic POTS.
1. I respond much better to Guanfacine ER (GFC) than stimulants alone (currently complementing with Vyvanse (LDX) 40mg, but I'd rate the Guanfacine as critical)
2. My blood pressure is very volatile, and GFC is supposed to have an impact but did not in my case, at least initially. I'd take GFC at bedtime and LDX in the morning, and on ChatGPT's suggestion, I asked my psych if I could take them together in the morning. Gamechanger for my blood pressure: the explanation seems to be that LDX makes my sympathetic nervous system extra simulated (on top of a poor baseline), and co-timed GFC balances it out.
3. I have poor cardiac endurance, and I find running nearly impossible. I'm a healthy young male who does weights and all. At ChatGPT's suggestion, I wore a Polar H10 and measured my resting heart rate while sitting, and then while standing still. I get a jump from 80bpm to 115bpm-ish, a strong indicator for something orthostatic.
I'm currently exploring rowing (with a concept2). I don't know why but it has a strong impact on my mental state that goes beyond general exercising: something about the rhythmic entrainment it produces, while being recumbent (good for POTS).
> Android is built from the ground up to provide surveillance data to Google-controlled adtech
I've always read this and it seems well accepted. But I'm curious what exactly does it mean? What's Android sending to Google? Surely it's not logging what I click on apps? It's not logging what I click on my browser since the websites themselves send this info for ad purposes. So what's Android doing that let's say my Linux laptop isn't?
Edit: Answering my own question. There is a cross-app unique identifier (ignoring any privacy sandbox stuff) so developers and ad networks can get a consistent id across apps.
Do you have any resources showing what this type of coding works in the real world? I keep running into this same types issue when I use languages without types specified everywhere, where the cognitive load gets too much for non-trivial projects. I would be pretty unhappy writing python without type hints and tools that check these for correctness built into my workflow for example.
This was from my experience that I thought may work for others too.
Types may help in some sense, but is not enough and sometimes even add extra load especially on heterogenous data structures.
Before using any tools beside basic IDE features, I think there are fundamental things you can do to reduce cognitive load, such as focusing more on code design, convention, structures, naming, testing, documentation, which can be applied whatever language you use.
I think people have had success with using PaliGemma for this. The computer use type use cases probably use fine tuned versions of LLMs for their use cases rather than the base ones.
I was wondering the same thing. I couldn't find any research on what increased baseline norepinephrine does to the pulses. I can see it being detrimental (pulses don't work as well) or neutral (brain adjusts to the new baseline).
Jeremy has a lot of cache with me with his amazing work in education and his excitement on this is a great sign in Mojo's favor. Solving the two language problem would be pretty amazing.
Assume you got source for a game written in a proprietary game engine. If you don't have access to the game engine itself, nor the API documentation, etc, how long will it take to get this game running in your manner of choosing?
The infrastructure in these companies is a huge amount of scaffolding that's non-trivial to replicate.
