r_lee

fair. but I think e.g. nitpicking about the FB rename to Muse distracts from the genuine points and makes the statements seem emotionally charged which makes it easier to dismiss

and I think there's also a point to be made that you shouldn't be too caring about issues that you can't influence, it simply causes harm toward yourself.

> building things yourself is just more fun

I can relate to this, but React definitely made things more fun for me at least.

Web Components on the other hand seemed to solve a problem that few actually were having, or for those that loathe React & modern frameworks because of bloat or whatever

the perfectionism and need for control is very true.

I think with the rejection and criticism it's not so much about the actual criticism but the idea of what the other person now thinks/operates on, while they might gloss over it, you might assume the worst case scenario and think that now they're against you and you have to win them over again

I'd say in general the concepts of "it's not a big deal" "it's probably nothing major" etc. don't seem to register

what's the point of saying all that?

you already know what will happen. if the model is good enough, people will use it.

you're asking for a lot of things here. I don't think most people take it that personally.

> where the personal data that does get out of the iPhone/ipad/Mac is encrypted with something Secure Enclave on processor, and “even Apple can’t read it”..

there already is technology for this, which is Confidential Computing which is implemented on Nvidia etc.

it requires transparency to audit it properly and if there's an exploit which weakens it then the security guarantees fall apart

but in terms of commercial viability, it's the most promising way to tackle this and it seems like Apple is already pushing for it with their Private Cloud Compute

why are you assuming it's not simply psychological?

"gets really nervous, makes a bunch of noise, annoys his brother, like he can't be canalised if he's not focused." could be autism or ADHD.

btw, ADHD doesn't mean "attention problems", it is much more complex than that and it can manifest in many different ways.

e.g. assuming he would have it, the listening provides a baseline level of stimulation, but as soon as it stops, there must be something to provide that stimulation, thus all those things.

if there's more behavioral stuff that you can recall, that would offer more clues.

I mean it'd be pretty easy for Apple to just make a special version of the OS/bootloader/whatever for their internal use

they deliberately do not want to enter the server market. it is literally an entirely different thing than what they're doing right now. they can't do any of their vertical integration there, there's no moat for them.

like Tim Cook said (quoting Steve I believe) - it's about knowing how to say no 1000 times so that you can say yes to the right thing once, and then put your all into that

I would have to agree with the guy, I mean:

"And so, while I overall very much enjoyed my time there and have freely recommended it to many people, the decision to pursue a PhD in Japan should always be considered carefully."

literally sounds like "as an AI model..." type of disclaimer or padding

like, who isn't going to carefully consider on how/where to do their PhD?

words can't describe how much I hate how everything is covered with white hospital LED lamps instead of the warm orange lamps that gave that 'night' feeling.

such a sad state of affairs. everything good must end I guess.

I would guess that if they do use such methods, it'd be to handle peak loads that go beyond their compute capacity, while they run the models at full capability when there's excess capacity

like before Anthropic signed the Colossus deal, the usage limits were insane and everyone was complaining, I wouldn't be surprised if they'd rather try to make inference faster that way than try to just limit people, at least for those on subscriptions

I noticed that a few weeks back 5.6 sol would regularly glitch out and start speeding random words or loop and then the next day it'd be fine

and there's sometimes just huge floods of complaints from people all of a sudden, which is pretty unlikely to be a coincidence

he's saying that he notices a difference between local (not nerfable) and hosted ones, so that it's not as likely to be just placebo

afaik https://ndstudio.gov/ (and GSA?) are now the ones responsible for such in-house engineering?

I think this means we need more cameras!

There's internet-connected industrial control panels online where you can just connect to them on your browser or some tool and start clicking buttons.

it's probably just convenient because there's no auth or VPNs required. yippee!

So what about like "mechanical engineers" or "process engineers" or "hardware engineer" or "biomedical engineers" etc?

> I don't know why that database was ever put into a computer, instead of filing cabinets in a vault, nor why it was ever connected to any other system, and not completely air-gapped in said vault, with manual controlled exfiltration of data upon request.

I suspect this is not because of the fake engineers but rather because other forces, e.g. management wanted more efficiency and security was the last item on the list. I don't see how changing job titles would change anything here

that's not a useful analogy, it's not like these models are CPUs where the cores just get increase in core count or clock speeds

they fundamentally change, architecture, the way they're trained etc.

with "run" I mean one agent that is misbehaving, as I'm guessing this is something that happens rarely

You're so right, he should've written:

"I think we as humans (I'm saying we as in probably the majority from my point of view and understanding of what the majority of people who visit this website likely think about this and how their ethical viewpoints likely align with my opinion that I am about to present however there may be a minority or some part of people that do not agree with what I am about to write therefore it is crucial that I make this clear before I go on with my point here)"

not to mention that ASML is likely majority owned by Americans

I think it wouldn't be too unreasonable for openAI to have a command center type of thing where they have people monitoring these runs where that wouldn't be such a problem. plus I feel like false positives aren't that likely if you'd actually run the reports by a capable model first which I'm guessing is they're doing

it's really interesting how they seemingly don't have a way to pause runs? like a P0 that would page an employee, shouldn't that pause the run and then make it into a decision on whether to let it continue vs that whole "run was killed" 2.5 hours later?

I'd say it's cool that they're openly writing about it and how they run their workloads. shows a nice window into how these things are actually deployed at scale.

it also shows how much density you can get easily from a single core if you wanted to replicate this

if it's agentic stuff, they likely aren't hammering a core constantly and they will maybe sit idle quite often between model requests, so it makes sense. I do wonder how much memory they allocate to each one though.

it's just very efficient use of shared cores that is required to make these kinds of workloads cost efficient

I've had bad experiences with gpt 6 luna, but 5.6 luna worked real nicely with e.g React, you can do a back n forth and it'll work like a teammate vs just a "done!" where you then have to clean everything up afterwards

it seems like most of the recent posts on Tao's blog are not actually Tao's?

and how will that pay rent?

like, how are people not getting this? is everyone here just living off their capital gains or something and that's it?

I feel like so many people here are part of that crowd. going on about how it's like humanities and how it's super valuable and those who are seeking to get paid are just missing the point

how is it even possible to think like that? or maybe it's some kind of coping mechanism

I don't think anyone hates all of AI except a minority

and even if you would hate AI, you don't really have a choice anymore since employers will expect you to be more productive due to it

but I definitely see a bunch of hate towards the dystopian aspects of it and as mentioned before, the whole slop thing and how you see AI slop everywhere now. it definitely evokes a very negative feeling in me when I see something like that