no notes

[theorem_poem@mstdn.social]

@karl they're going to choose both - by building their data centres in Asia/Africa. Like how they bought their plastic waste problem down by exporting their garbage.

[rumpelheinz@dresden.network]

@karl Computer haben andere Bedürfnisse als Menschen. Menschen sind nicht mehr notwendig, wenn es KI gibt.

[th@social.v.st]

@karl The article has been amended to remove the quote "Editor’s note: This is an updated version of the article. The previous version wrongly quoted Jeff Bezos. The error is regretted."

[saxnot@chaos.social]

@karl ok easy choice

continue living
v
billionaire pet project

easy choice

[saxnot@chaos.social]

@orange_lux @dibi58 @karl for most purposes you don't need a LLM at all

cut 95 % of all car usage
we want public transport

cut 95 % of all LLM usage
(i don't know the details but I do know the LLM garbage on e. g. immoscout or visualping is useless to me ; just drop all of that)

[saxnot@chaos.social]

@th @karl LLM is a dead end towards AGI and I thought that was shown to be true since years?

what a crazy take
we need established solutions to established problems
e. g. de-carbonize agriculture, steel, energy, etc.

total water recollection (closed loop)
never pollute wastewater with pfas
those are already bare minimums i would say
and just don't build them in the first place we already have an over-supply for this no-benefit bubble

[saxnot@chaos.social]

@th @karl i want established solutions to established problems

Gimme a wind turbine, veggie-first food, gimme free train rides etc
all of this has been invented a generation ago and we can switch once the [imaginary thing] has invented better

e. g. hyperloop is garbage and was never better than e. g. railroad
(and we knew the vacuum train idea is a dud centruries musk "invented" it (patented 1799 and shown to be a dud >50 years ago)
and yet the hyperloop hindered the california rail expansion

[orange_lux@eldritch.cafe]

@saxnot @dibi58 @karl and for what purposes should we NEED an LLM ?

[saxnot@chaos.social]

@th @karl assuming a hyperloop thing is invented tomorrow and it's better than trains
until then we go all-in on trains

assuming a better energy source is invented tomorrow
great
until then we go with wind / solar / storage / etc
but UNTIL THEN we do that
e. g. german politiclowns are obsessed by fusion power (the same party which starved fusion research funding now wants to start building today).

should be obvious common sense (because it is)
musk, cdu, etc are "ideology" to use their lingo

[znexhf_svax@social.cologne]

@karl choose wisely

[jelu@mastodon.social]

@karl maybe we can get the tech bros to care about the environment if we spin it that they need to save the fresh water needed for the data centers

[hex@kolektiva.social]

@orange_lux @saxnot @dibi58 @karl llms are useful for NLP. Inference is actually relatively cheap, it's the training that's really expensive and resource intensive. We've probably already maxed out LLM capabilities, so most of this training is not useful. Companies keep training because they need to convince investors that infinite growth is possible. What actual gains are being made are coming from architectural changes, not from training.

Basically, "AI data centers" should not exist. Local models can do everything that's needed. If we need to train new models, those need to be balanced against climate goals (basically, don't fucking do it). And LLMs should be removed from basically everything they've been shoved into recently.

If you don't know why LLMs are useful, you shouldn't have to interact with LLMs. Even some of the places where they are useful, they can be used to construct cheaper models.

There are a few things, like correlation across huge data sets, that they're useful for. But even then, simple encoding can give you semantic search, where inference is not necessary or only provides minimal additional benefit.

Yeah, basically, 95-99% reduction in cars and AI. It's basically the same thing.

[clouderst@abc.clouderst.xyz]

@karl i cannot comprehend how this is something to choose.

how stupid can you be to pick between those two

[stellar@mk.absturztau.be]

@karl@infosec.exchange we lost france to the great freezing ​​ (guillotine the guy who colored this map)

[saxnot@chaos.social]

@orange_lux @dibi58 @karl same we use CNN, RNN, ... etc
LMM is just another useful helpful machine learning type.

in contrast: blockchain was useless from the start.
LLM is boring tech blown into weird proportions due to greed, capitalism, tech-oligarchy, etc.

[saxnot@chaos.social]

@orange_lux @dibi58 @karl e. g. the sort of stuff which ranks search queries, recommends videos, "people also bought", ads, mall layouts, etc

site note: my degree has focus on data/ml stuff

[janriemer@floss.social]

@Stellar @karl

Coloring makes sense, if you interpret the grey as "dead land".

[m@martinh.net]

@karl Imagine a group of hyper-intelligent beings building an incredibly powerful computer AGI in order to find out the answer to the ultimate question of life, the universe and everything ​

[pasmac@atmasto.com]

@karl jzzz Data Centers can be optimize by a lot.
If we are still able to contact Voyager 1/2 how the help we can’t optimize IA, datacenters, etc.

Buy buy buy, crap crap crap …
We need to get to basics and build with few and what we have.

It’s possible to reach climate goals and have computer power. But for that we need to have the essential and work with it.

[orange_lux@eldritch.cafe]

@saxnot @dibi58 @karl I'm not working into data, I'm "just" a normal dev, but we had ways of recommending stuff (also bought, recommendations) way before LLMs. I don't think this is an example of why we would need them.