Slightly different, but Charlie Brooker (of Black Mirror fame) had a series called Screenwipe, which lampooned the tropes of television in the mid 2000s. It was pretty funny, and quite dark in places.
That's interesting - I found that same fusion super-jarring, to the point that a part of me actually started feeling embarrassed for the author. Even though I know it's a satire, and done in that way purposefully - it was visceral, I just couldn't help it.
I think it's hard to know whether you're interested in the truth. Even if you think you've scrupulously cleared away all the thickets of biases and prejudices, there are probably more still there, hiding in all the corners you've turned away from.
I think it's work either way. You either spend brain-time on memorizing the intricacies of a lot of prebuilt tools, and how / when one might apply them (modern c++) - or you dispense with the many tools, and spend your brain time on memorizing relatively few, but with more effort and detail required on their application. (C). I prefer the latter, but as you say - it's probably linked to personality to some extent, and quite "baked in".
Man, I write all my programs in pretty much C. I'll use a class or two, but that's as far as it goes. I call myself a C++ programmer, but really I don't touch 95% of it with a barge-pole. It just seems like a load of unnecessary, complicated crap that will surely go wrong and confuse me.
I suppose it's whether you think of array indexing as an operation that offsets the read location relative to th start of the array. It's probably rather low level for the way people are used to working these days.
It is admittedly a mindset that is born of C coding. Pointers and memory offsets and so forth. But lots of languages that abstract such "close to the metal" factors away, still use 0 based indexing.
I was interviewing a candidate a couple of months ago, and as part of the assessment they were required to write a bat and ball game in Javascript, in advance of the interview.We would go through the code during the interview and I would ask them how it might be modified / extended in various ways.
This candidate produced a reasonable, functional bat and ball game in a couple of pages of Javascript code, and I had high hopes for the interview. But as soon as I tried to delve into the code with the candidate, it became clear they had no idea what half of it was doing. I suppose they got chatGPT to write it for them, or something. Was disappointed and vaguely annoyed to have wasted my time.
Differentiable in this context, means that it can be learned by gradient descent, which uses derivatives to adjust a neural network model's parameters to minimise some error measure. As for how that applies to a hash function, I think the lookup gives some sort of weighting for each possible output, for each input, with the largest weightings corresponding to a "match". But tbh I'm not certain on that last part...
Now here is an app that ought to be held high as having real, and potentially enormous benefit to the world. (Though I wont hold my breath for any online media sources to write about it.) :-)
Oh, I'm right there with you man. Modern pop is mostly dreadful. I think music producers have homed in on "the most appealing chord progression possible", and are reluctant to deviate from it, like an arms race. What boggles my mind though, is that some people don't seem to get bored by it. Tbh modern pop isn't even really music any more, it's much more a multimedia thing - with visuals (hot performers, cool videos etc) playing a massive part in its appeal.
I've been a UFO enthusiast since I was a kid, always excited to hear about the real, hidden truth which is surely just around the corner, about to be revealed. (I like the slightly sinister sense of mystery, I think). But I have to admit, it is suspicious that given the super high res. cameras we all carry around, every UFO video we see these days is just a few blurry pixels.
IQ isn't all that. Mine is 140+ and I'm just a somewhat well paid software engineer. It's TOO abstract a metric in my view - for sure it doesn't always translate into real world success.
In the article Richard shows us the compression stats for the png algorithm vs lz4, for the original lossless image. It would also be interesting to see how the png algorithm performs on the lossy, preconditioned versions of the image. Does it also get any benefit, or does lz4 pull ahead?
The cost of decompression is something I've run into, regarding 3D data. Draco, which is a highly efficient compression system for 3D mesh data, is primarily used with browser based webGL / gltf model rendering, and it achieves excellent compression ratios for online transmission. However, the decompression cost at the viewer's end often means that you will wait longer to see your 3D model than if you had used a simpler format, especially over fast links.
Lol. I clicked on the link, thinking "Oh come on, it can't be THAT bad", and was presented with a literal endless gallery of slightly Manga-ish women with unfathomably huge boobs.
Maybe, instead of turning the whole world into paperclips, the AI apocalypse will consist of the filling of every last speck of hard disk storage space with a trillion boringly similar synthetic jpegs of such charming ladies.
Before even deciding between those choices, we should at least stop the problem continuing, by preventing so many prescriptions for opioids being given out to people in pain. The government should prosecute the likes of Purdue for their aggressive targeting of people in pain.
It won't happen though, because of regulatory capture, and general lobbying / corruption.
I could be wrong, but I think the only probabilistic component of an LLM is the statistical word fragment selection at the end. Assuming this is true, one could theoretically run the program multiple times, making different fragment choices. This (while horribly inefficient) would allow a sort of backtracking.