1,377 karma · joined August 8, 2022
An LLM cannot possibly have any concept of even what a proof is, much less whether it is true or not, even if we're not talking about math. The lower training data amount and the fact that math uses tokens that are largely field-specific, as well as the fact that a single-token error is fatal to truth in math means even output that resembles training data is unlikely to be close to factual.
Your last point also highlights a real issue that affects real humans: just because someone (or something) cannot talk doesn't mean that they are not intelligent. This is a very current subject in disability spaces, as someone could be actually intelligent, but not able to express their thoughts in a manner that is effective in sharing them due to a disability (or even simply language barriers!), and be considered to be unintelligent.
In this way, you could say LLMs are "dumb" (to use the actual definition of the word, ie nonverbal) in some modes like speech, body language or visual art. Some of these modes are fixed in LLMs by using what are basically disability aids, like text to speech or text to image, but the point still stands just the same, and in fact these aids can be and are used by disabled people to achieve the exact same goals.
If your chip needs very basic processing, and you're going to sell it for less than a cent per chip, you need your core to be basically free.
Sure RISC-V is very inexpensive, but for parts like these risc is massive overkill and as a modern core, you'll be spending a lot of die area on a core that should really be an ASIC, but designing an ASIC is expensive so you can just slap a 6502 in that bad boy and call it a day.
On the other end of the scale, you'd be surprised to learn that many chips you don't ever interface with as a user contain even ARM cores. These are usually ASICs in high-end products.
Although there are likely an overwhelming majority of my peers who do have an interruption device phone, I refuse my phone not be My device, and I explicitly buy phones that I know I can customize (unlock bootloaders, flash roms, or simply just know it has the interop capabilities I need to make it my own). Where this is in opposition with your statement is that I do have social media, mail and even work email on my phone, I just chose to customize it to not be interrupting to work and personal life (though obviously that line is different for everyone).
Of course I'm not saying your way is wrong at all, just sharing my experience.
There are very valid reasons to want to send traffic to a node so called "out of the blue", and suggesting otherwise suggests you have never looked deeper than HTTP.
If a server has no AAAA record, machines only connected to IPV6 simply cannot contact that server. This is fine in theory (just buy an ipv4 lolz) but in practices this means additional costs that will only go up, as the ipv4 space becomes more and more crowded. As a non-network admin, I care that services I'm a business costumer of are available to my servers, the biggest offender being GitHub. GitHub still does not have an ipv6 despite "looking into the issue" for several years, and this means that I am unable to use their services or indeed any software that requires access to GitHub for, say, updates, pulling artifacts, container images or whatever else from their servers.
It does sound like it's a bit presumptuous of me as a random IT person to demand an IPV6 out of them, but in the year of the Linux desktop, there's no reason for them not to have the infrastructure in place to simply flip a switch and be done with it, for basically free.
Of course, the big players own swathes of IPV4 space that they are able to rent out at a premium, so they are heavily disincentivized from ever supporting ipv6, as that means obsoleting their investment with no replacement, but it's bullshit that they are able to do this in the first place.
Funnily enough we did end up saying "implement a full browser stack" :/
The alternative, however is death by starvation, so I'd still pick the coconuts.