Its already here, it’s called the average Redditor.
This isn’t a joke. You already know what a low power LLM is, its a 40W human brain with high school level knowledge and a confident human tone writing on Reddit. But not the entire brain, only the natural language and superficial knowledge part… Meaning having the human is better anyway.
I love your optimism. These are the people who add an S to mass nouns like ‘email’, and who pluralize with an apostrophe. A successful high-school education isn’t being shown, there, at least in terms of literacy (admittedly, we are represented by how we write).
Is it correlative, in that many Redditors live in a country where 54% of them do not communicate above a 6th-grade level?
That’s a massive assumption on your part. Exactly how do you propose that the entire architecture of LLM’s be restructured to not be the massive resource hog desperately seeking a profit that it is? Mind you, it’s shit now and you’re basing your argument on a hypothetical “maybe”, like maybe you’ll win a massive lottery payout tomorrow.
You make a compelling argument, it will never be achieved. I am convinced. Not one human is capable, nor are one thousand humans. It is unsurmountable mountain.
It is the year 5000. Aliens look at the wreckage of earth. They see what little remains of our meager existence. The goal that haunted humanity for all eternity, a doomed plan from the start, only ever resulting in certain failure? making chatbots use as much energy as a gaming rig. It never happened. It was… impossible.
Okay, since its obviously hyperbole that it isnt possible, its a potential outcome so lets return to my original question (deemed a massive assumption on my part by you):
once it no longer needs massive amounts of expensive resources, what then?
If you need a new type of nuclear reactor to power your shit it means your shit is too complicated.
AI is not complicated (hard but solvable with time and effort), it’s complex(easy but numerous and their interactions make it hard)?
sigmoid(aX+b) is not that complicated
That wasn‘t even the argument. The argument is that AI is a resource waste; it uses too much power and water for what is a shit return.
Who cares how complicated it is.
once it no longer needs massive amounts of expensive resources, what then?
Its already here, it’s called the average Redditor.
This isn’t a joke. You already know what a low power LLM is, its a 40W human brain with high school level knowledge and a confident human tone writing on Reddit. But not the entire brain, only the natural language and superficial knowledge part… Meaning having the human is better anyway.
I love your optimism. These are the people who add an S to mass nouns like ‘email’, and who pluralize with an apostrophe. A successful high-school education isn’t being shown, there, at least in terms of literacy (admittedly, we are represented by how we write).
Is it correlative, in that many Redditors live in a country where 54% of them do not communicate above a 6th-grade level?
That’s a massive assumption on your part. Exactly how do you propose that the entire architecture of LLM’s be restructured to not be the massive resource hog desperately seeking a profit that it is? Mind you, it’s shit now and you’re basing your argument on a hypothetical “maybe”, like maybe you’ll win a massive lottery payout tomorrow.
You make a compelling argument, it will never be achieved. I am convinced. Not one human is capable, nor are one thousand humans. It is unsurmountable mountain.
It is the year 5000. Aliens look at the wreckage of earth. They see what little remains of our meager existence. The goal that haunted humanity for all eternity, a doomed plan from the start, only ever resulting in certain failure? making chatbots use as much energy as a gaming rig. It never happened. It was… impossible.
Oh good. If you resort to hyperbole to prove your argument you’ve already lost.
Okay, since its obviously hyperbole that it isnt possible, its a potential outcome so lets return to my original question (deemed a massive assumption on my part by you):
There are plenty of things that are complicated and could use a new type of nuclear reactor.
Training LLMs just didn’t seem to be one of them
But thanks to Satya I can now right-click a file and select “Summarise”, how is that not a massive MASSIVE productivity boost???
IDK about you, but the time it takes for me to actually interface with AI and then verify the AI did things correctly are basically equal.
What kind of shits require nuclear power to be pushed out, mate? You’re getting me worried. Been having your fibers?