Tenoke

That seems naive in a status quo bias way to me. Why and where do you expect AI progress to stop? It sounds like somewhere very close to where we are at in your eyes. Why do you think there won't be many further improvements?

$3k is already close to the price of just a 5090, so if you can have a reasonably fast but way more (equivalent to) VRAM (128gb) machine sounds great for a hobbyist assuming you can easily run and train llama, Stable Diffusion and similar vram-hungry projects.

I've always liked the story but if I was presented with that revelation I'd be questioning to what extent these other me's are really me and how really everything else (genetics, environment) is shaping behaviour and I'm what? A vessel that gets all the memories at the end?

>“Every time you victimized someone,” I said, “you were victimizing yourself. Every act of kindness you’ve done, you’ve done to yourself. Every happy and sad moment ever experienced by any human was, or will be, experienced by you.”

I mean, not really, since I wouldn't do quite all those things as I am now. In this framework I'm just forced into doing those things by the factors going into those particular lives.

>"let's migrate our infra from AWS to X"

We migrated to Scaleway pretty easily and the main argument was that it's easier to conform to EU regulations if all your servers and data are on an European cloud.

Hetzner has been the cheapest big option when you just want raw servers for at least a decade, and e.g. Scaleway is pretty similar in price to US based providers.

I usually don't remember details like that but I practice the ways to get them - e.g. for elementary school you should know how old you were for the year, and then you can check at what date elementary school starts in your area (where I'm from it's always 15th of September) so you can pinpoint it.

In large enough groups of people there'll always be some crazies. Are there really more of them to come out of LW circles than a comparably large other group?

That's such a counter-productive and frankly dumb thing to do. Just don't vote on them.

I was skeptical that Bitcoin still uses more than AI but it seems to indeed be the case (or close) - current Energy estimates for Bitcoin recently are at 175 TW/h[0] (though I don't know if those are extra pessimistic or not).

For AI the total consumption is perhaps a little less currently (the figures I find quickly differ) but mostly projected to go over twice that in the next couple of years[1].

0. https://digiconomist.net/bitcoin-energy-consumption

1. https://www.windindustrie-in-deutschland.de/meldungen/electr...

They wouldn't needed to be 'literate' just how many programmers today don't need to know assembly or low level languages or compilers for their jobs.

I guess with this they can also record user-browser interactions to use as training data, which is one way I was envisioning for creating a human-like AGI back in the day (2019)[0]. Of course, the current paradigm has went in a different direction and training directly from all the inputs/outputs of computer usage isn't quite how this data would be used, but still.

0. https://svilentodorov.xyz/blog/human-imitating-task/

Some reports[0] paint this as something Trump announced and that the US Government is heavily involved with but the announcement only mentions private sector (and lead by Japan's Softbank at that). Is the US also putting in money? How much control of the venture is private vs public here?

0. https://www.thewrap.com/trump-open-ai-oracle-stargate-ai-inf...

1. https://www.cbsnews.com/news/trump-announces-private-sector-...

A bowl of yogurt would help but is not really that much extra protein. Definitely not near the extra protein serious gym people attempt to consume.

And that's if you can even process a high joghurt diet without issues which isn't the case for plenty of people.

Maybe not on or only on their class subjects, but kids also read random things including Wikipedia articles on topics they get curious about.

>If you always have 3% more working memory you might accumulate more knowledge after a while.

Or you might accumulate way less knowledge because you don't have a phone to get information from and are getting say 300% less information overall despite retaining 3% more of what you get.

Driving performance after 24+ hours of being awake is pretty irrelevant to phones (even if they cause you to sleep less they don't cause you to not sleep at all), and it's just some brief effect with no reason to think otherwise from looking at this.

It's definitely the best contender, and I've been trying to use it but so far there's less people, my feed there is much blander, there's missing features that I miss from Twitter, and everyone seems to be posting about Twitter itself all the time.

The take-home challenges at my last jobs have been close enough to the work to be useful, and much more relevant than anything else I've seen in any interview.

I do wonder how much of this is the prompt. Gpt-4 was doing really well against me when it launched but my prompt was quite different

I'm fairly sure it's relatively true (except for occasional extra purchases on top) unless he's been keeping it up in places I wouldn't expect him to.

I don't like that now people might pigeonhole him a bit by thinking about his effective frugality but I do hope he gets a ton of donations (either directly or via patreon.com/gwern ) to make up for it.

Everything? He also writes things like this: "A list of unheralded improvements to ordinary quality-of-life since the 1990s going beyond computers."

https://gwern.net/improvement

Err, checking the line on a horse race is a much more reliable way to tell who has what chance of winning than going to the stables as a random person, and most likely for a professional as well.

If the issue is reusing names then they also shouldn't have named the radio lora with the same name as my aunt.

This is such an exaggeration. I fly with them, wizzair, easyjet etc. at least once a year each and I never have issues - I just press skip on every page that upsells me, same as I do on any other site including Amazon.

Maybe it's not the case, but I've been suspecting that at this point the worm, and even much more complex organism simulations can work well if enough work and compute is put on it, the development has just stalled and isn't especially well funded. This is only based on me quickly checking a while back how many commits they get, and how many new approaches are even tried and it didnt seem to be so much anymore.

I'm rooting for them though, it's a really cool project.

Yes, though it's not about taking precedence but about sampling frequency. So for example, if you have 1 GB emails, 10 GB external data, you can instead sample your emails twice as much and effectively change the ratio of what the model was trained on from 1:10 to 2:10.

Just checkin for free on your phone and use your phone as a ticket. I fly with Ryanair often enough and honestly wasn't even aware there's still a fee like that.

I just flew with Ryanair and yes, the fees for luggage etc are expensive, but if you don't take luggage and any extras they offer, they are amazing value. I hope the cheap airlines don't get overrun with fines because if not for them I'd barely see my parents.

Quality suffers at very large contexts, and you might want to use your large context for something else (e.g. a long current conversation or a lot of recent data).

In all fairness, most things did move to the cloud.