I agree completely with all the points made in the article and would double down on the critique of the philosophy of biology as a discipline and science itself: namely that is has historically evolved as a mere system of typologies and that this is so settled in and so sacrosanct that biology is such a superstition with human-made categories a couple hundred years ago and not a real science.
We see the exact same things also when discussing what is a species and also completely disregarding the reality of horizontal gene transfers etc in the strict, traditional trees.
The models are quite wrong and even reduced wrong.
There is this one famous article that shows how traditional biology would go and analyze a transistor radio, namely just label its assumed components!
Touch on a stove makes any stove I know gastritic. A drip of water and it turns on power mode by itself and melts lids. A bit of oil and it randomly turns on itself. A piece of wet cloth and it does the same. Sometimes even nothing at all triggers it.
Honestly, I wouldn’t mind. I would continue and live my life. I would ride public transportation to go to work, I would go to the library if I want to check something out, I would just be spontaneous instead of preplanning everything online or be influenced by opinions online.
Even simple things like this would be difficult. For starters, ticketing tends to be powered by 5G on buses and over physical networks for permanent fare gates like at a railway station.
Bus arrivals screens and boards at stations are also fed by the network.
I guess if we take a generous interpretation of the OP's question though, those would still be able to function assuming they work over a private network instead of the internet.
I guess. I think the important thing is getting the program in your head, not on the screen. If the code is too complicated to hold it all in your mind then more columns of crisp text will not save you.
It is in a weird middle ground. It is much worse ram and much slower hardware (factor 4 or so IIRC) and only an option if you need more than 24 but less than 200. Also, only if you think that 9000€ is pocket change but 20000€ is cost prohibitive. Also, you need all that power but no server, no ECC,... And you only ever want to do inference but no training or tuning.
If you want a Mac anyway, sure. But if you don't care, this seems like a very, very specific Venn diagram.
It's not much worse RAM, though. RTX 4090 has memory bandwidth of 1050 Gb/s. M2 Ultra is 800 Gb/s. And you can get a Mac Studio with Ultra and 128Gb of RAM for $3K or less. It's great for 70-150B models.
You're correct that it's only good for inference, but most people running local LLMs only do inference.
You don't need an M2 for this, an M1 will do just fine. Nor do you need the one with maxed-out SSD, which jacks up the price considerably. Finding a brand new M1 Ultra for $4K or less right now is pretty easy. When I got mine, about a year ago, $3K was the best deal I could find.
I find the “a la mode” vs “au jus” discussion right under the daemon one very interesting!
I wasn’t familiar with both of these expressions but I looked it up and “a la mode” is an American culinary expression, meaning “served with ice cream”. And “au jus” is also an American culinary expression, meaning “gravy” or “broth”. Now, even though they are both derived from a French expression that is a prepositional phrase with à (meaning with), it does not matter any more when they were borrowed to English.
“A la mode” became a new adverbial expression meaning just that: “served with ice cream”. You can have pie a la mode = pie served with ice cream, but obviously not *pie with a la mode = pie with served with ice cream.
And “au jus” became a noun expression meaning “broth” or “gravy”. And you must say sandwich with au jus = sandwich with gravy and can’t say *sandwich au jus = sandwich gravy.
What is extremely interesting here is that it bothers the prescriptivist who wants language to be a certain way he feels it is supposed to be, also the author on that webpage.
Yeah, I was hanging out with someone recently who kept using "au jus" like "sauce", i.e. "you could make that with an au jus" , "ooh yeah that would be so good with an au jus on the side!" or similar ...
If you dig into this person’s posting history and also if you read regularly on HN for a couple years you will notice that it is actually this very user that deliberately uses the 0 prefixed 5 digit year numbers, and also goes out of their way to include year numbers into their posts to make people ask this question.
We see the exact same things also when discussing what is a species and also completely disregarding the reality of horizontal gene transfers etc in the strict, traditional trees.
The models are quite wrong and even reduced wrong.
There is this one famous article that shows how traditional biology would go and analyze a transistor radio, namely just label its assumed components!
Here is the discussion: https://news.ycombinator.com/item?id=31697757