Besides, Macbooks with 32GB RAM is consumer hardware, just maybe on the higher end.

That's my interpretation, disclaimer I'm not an expert in this stuff.

Page 55:

""" there are characters with no glyphs. glyphs that can correspond to a number of different characters according to context. Glyphs that correspond to multiple characters at the same time (with weightings assigned to each), and even more possibilities.

The problem of glyphs and characters is so complex that it has gone beyond the realm of computer specialists and has come to be of interest even to philosophers. For example, the Japanese philosopher Shigeki Moro, who has worked with ideographic characters in Buddhist documents, goes so far in his article Surface or Essence: Beyond Character Model Set [274] as to say that Unicode's approach is Aristotelian essentialist and to recommend supplanting it by an approach inspired by Jacques Derrida's theory of writing [114, 115]. The reader interested in the philosophical aspects of the issue is invited to consult [165,156], in addition to the works cited above. """

I think "essentialist" is probably a good description of the philosophy of how Unicode defines characters as opposed to fonts and glyphs, so I adopted it.

I think some characters got different code points, while others were merged. And apparently the Japanese complained bitterly over the ones that were merged. If you had read any articles about that, this is probably what you have in mind right now.

And I'm also a bit tilted by the ones that had different code points, because when processing CJK text now we have to deal with characters that are (in my native Cantonese) essentially the same, looks similar (to my undiscriminating eyes), yet having different code points so that things like text search sometimes don't work.

Of course I'm not "blaming" the Japanese, if anything the simplified vs traditional Chinese thing is much more of a practical problem, and the conflicting code points I deal with on a routine basis are more of a Hong Kong vs Taiwan thing, but I was told that the Unified CJK thing adopted a different "philosophy" from the rest of Unicode (which I think really is some kind of Aristotelian essentialism...) mostly due to vocal objections from the Japanese.

The peculiar properties of CJK characters and the philosophy (apparently the Japanese did not like Unicode's tendencies towards Aristotelian essentialism) under which they were implemented in Unicode probably singlehandedly forced unicode to expand beyond the BMP....

He made such a big fuss about ollama implementing their own kernels and felt slighted about the online comments saying ollama didn't properly credit llama.cpp and it kind of left a bad taste in the mouth among the local inference community.

For me personally, it was this that made me avoid them at all costs: https://github.com/ollama/ollama/issues/11714#issuecomment-3...

At this point, I am unaware of a better definition. I know the Chinese Room argument (and I disagree with it), but I'm not aware whether the proponents of that argument have a better definition of understanding other than "well, the Turing Test isn't enough"...

---

PS: Interestingly the issue of compression is highly relevant regarding the Chinese Room argument -- the essential element in the Chinese Room argument is that the information is not compressed...

I mean, my interpretation is that the question Turing tried to answer is equivalent to "How can we determine whether machines understand humans/human thought?"

This only works when both systems can talk about pretty much arbitrary things, but if you want a more general method for less complex systems, perhaps having one system simulate another system is sufficient. (Which is also another Turing invention)

Either the author knew of prior work that argues the same thing and they ignored it, or they didn't know. And if one writes a 1000+ word article premised on this idea, wouldn't one be presumed to know at least in which century the idea originated from?

Arguably these kind of blog posts should be more subject to such criticisms, because the blog posts purport to "teach" the general public about a concept in an authoritative tone (or at least the author seems to pose as knowledgeable in the subject), while for academic papers, everyone who actually reads the paper knows where the ideas came from anyway and it's mainly an issue of attribution (and maybe about fairly distributing the citation count...)

You can keep telling people to avoid flying and live a life of low carbon footprint, but if the issue is much more systematic, economical and technological, it will just become a distraction and make people dislike you (and the cause).

AI is definitely going to shape the industry, and while I agree it's hard to see how it can create more job opportunities for junior positions, since you're already decided to do CS, just learn as much as you can about anything that interests you.

It's better to be really good at an unpopular niche, than to have average skills in the hottest trend.

It's kind of sad that popular CS textbooks often focus on solving precise problems with lowest theoretical complexity bounds while ignoring more practical (but generally applicable) computation techniques.

In machine learning they call it "gradient descent", which in older days had analogies in techniques called "hill climbing", "local search" and "simulated annealing". Basically you have a function you need to optimize for, and you clumsily tweak the parameters so that you get the (locally) max/min value you wanted. These techniques were great at finding approximate, locally maximal solutions without trying all the possibilities at once (which is more akin to the kind of "brute force" in the traditional CS context).

I guess because these techniques were generally applicable yet the outputs were approximate and you couldn't analyze them much (no fancy O(n log n)), the theorists did not find them interesting and thus were not put into the spotlight of student's learning curricula.

In modern machine learning they do this gradient descent thing which is also tweaking the parameters bit by bit to optimize for the loss function, except that the parameters are now in the billions and trillions. The compute required is huge of course, but it's actually quite an "efficient" process, and it's not actually doing much of "brute forcing" at all. During training, the process is essentially, almost equivalent to, compressing the many many trillions of tokens of training data. To me it's quite amazing that they manage to complete such a process within a couple months of training, even if they have hundreds of thousands of GPUs...

The "best" thing the US government can do is to build a Great Firewall to wall off the "existential threat from China". I'm not an American so if you guys decide to do it, good luck.

It's my personal experience at least, and since then I've never used social media attention as a gauge of the quality of my writing.