US Big Tech wants to eliminate the federal government administered by humans, and replace it with AI. Amid all the talk that has generated one aspect has gone relatively unreported. None of the AI they want to replace the humans with actually works.
AI is still plagued by widespread simple and basic errors in reasoning. Furthermore, there is no path to fixing this problem. Tinkering with training data has provided some improvements, but it has not fixed the fundamental problem. AI lacks the ability to independently reason.
'Move fast and break things' has always been a Silicon Valley mantra. It seems increasingly that is the way the basic functions of administering the US state will be run too.
Despite being unable to fix fundamental problems with hallucinations and garbage outputs, to justify their investor expenditure, US Big Tech insists AI should administer the US state, not humans.
byu/lughnasadh inFuturology
32 Comments
My friend gave ChatGPT the same engineering stress problem at 5 different times. The AI got the right answer twice, the wrong answer twice (once it was off by 30%, the other time it was off by 250%), and the last time it gave an answer that was unrelated to the question asked.
Yeah, let’s definitely let AI administer all our lives – NOT
Considering that despite the best efforts of politicians for decades, that the federal govt has only grown, the budget has ballooned well past our ability to maintain it, and we spend more money on debt interest than defense — I agree it’s time to move fast and break things.
Austerity is what it’s called when you finally have to pay for the things you’ve been buying on credit for years. This is it. We are here.
Any complaints that in the short term it’s costing us more or that they didn’t target the largest expenditures are irrelevant and only really motivated by partisan tribalism. We need a drastic reduction in the federal govt and ways to vastly increase the revenue. We may not get it right the very first try, but try we must — for our future.
Of course, easy to make it output what you want (if you are elon, Sam, etc) and then claim it was the “black box” of AI and so must be right.
AI is extremely good at comparing large amounts of data, it would be stupid not to use that in governing large amounts of people. Not without human oversight, of course.
People who think “AI” – LLMs – are intelligent are fools. Big tech is exploiting this – what they really mean is *they* want to run the federal government.
“But if it looks like intelligence then how do you know it’s not?”
If you really think this, you’ve been fooled, as if watching a magician perform a magic trick.
AI does not *think*. Call it what you will, make whatever comparisons you will, but at the end of the day LLMs are just the same input-output applications running on the same hardware that we have been running computation on for decades. There is no intelligence, no thought, no emotion. Just bits in-bits out. But now the output is language, and so we are easily fooled into thinking we are seeing intelligence. Nobody thought this when computers started outputting numbers (which they are far better at). To the computer/model, it does not see language, it sees numbers. That is all it can comprehend – and that is a stretch of the definition of “comprehend”.
Allowing additive training via interacting with people needing services would help, but it’s still gonna take a human helper right there correcting a program over and over in realtime before they’re as good as humans. I imagine that chatbots trained on scraped phone and email interactions and federal employee manuals are gonna show the same biases towards bureaucracy as their human counterparts. The federal government is supposed to be deliberative and thus slow, I imagine that will be baked in
They just have to create the mark of the beast and start forcing it. And soon this whole show will come to the dramatic end. I wish it could last a bit more.
At the end of the day human minds are just the same input-output applications running on the same biological hardware that have been running pattern recognition for millennia.
See what I did there.
The arguments here have been “LLM is not intelligence and if you can’t see that the you are dumb” which is dismissive as best, extremely shortsighted at worst.
Whenever someone makes this assertion I challenge them to prove that the human brain is not tokenizing auditory input at a biological level. And keep in mind we have evidence that something akin to tokenization of syllables is happening when we study brains under FMRI.
Well it’s a good thing I’m always polite to my echo. Hopefully our new AI overlord takes that into account.
It’s a glorified chat bot that is destroying the environment
>> Move fast and break things’ has always been a Silicon Valley mantra. It seems increasingly that is the way the basic functions of administering the US state will be run too.
As opposed to “move slow and keep things broken” which is our current state of governmental affairs.
> to justify their investor expenditure
Why should we care?
The actual technology does not begin to exist to attempt that. But perhaps they want to rule themselves and pretend that it is a (non-existent) AI doing it.
“uninformed guessing sucks but what if we automatically guess by echoing previous uninformed guesses.. that sounds amazing”
As if MAGA believing Trump was America’s Savior wasn’t an illusion?
The federal government is the cash cow come to the rescue for the massive investment it AI that has produced very little in income.
AI isn’t impartial, that’s the biggest problem. They are all skewed towards what they’re developers/controllers want. The “AI” being used here is just the data they want to see implemented with bells and whistles.
I wish people would stop saying “US Big Tech” when the mean Musk, Meta, or Bezos. There are several big tech companies that have no interest in this happening.
The issue is that the kind of political problems a country faces are not ones that are typically solved by ‘faster thinking’ or similar, they’re issues that arise because two or more groups have their professed interests at odds with each other and ideological opposition about what goods should take priority in society.
In this regard I don’t see what an AI has to offer except the illusion of objectivity. “We asked the AI and it said the *objectively correct* answer is that theres no money for public schools, sorry but thats what it says”
how well protected are the server farms and where are they? asking for someone way else
If AI is governing us, who is in charge of the AI and what is stopping them from making it have certain decisions?
They only like it when they get to be the ones breaking stuff. That’s the biggest problem.
I say this every time.
You shouldn’t be worried about whether AI is or will be capable of a skynet like takeover. You should be worried that corporations will offload important responsibilities to AI regardless if they are capable or not.
They are in the process of doing just that, the skynet boogeyman is a distraction to have know-it -alls hand wave the threat of it away because they stop considering any kind of threat the tech can pose
If it was open, source, sure. At least when something goes wrong, I can now it was an actual mistake instead of billionaires paying off politicians.
Unfortunately, it’d those same billionaires pushing this.
Which is worse. An AI that hallucinates, or an ex CEO of a chemical company like dupont, overseeing the government org that prevents pollution?
Be real now. The CEO is going to actively conspire with his former company to boost profits. The AI has no such interest. Which is worse, corruption or hallucinations?
Tech bros can let robots jack them off all day long, leave the rest of us out of it.
My favorite AI interaction over the last several days was Google’s AI telling me that a five letter word for heated was ‘fervid’
It only has to do better than Trump and Musk, who I’m pretty sure are suffering from some hallucinations of their own!
There’s that $500 billion Stargate AI project afterall
You can basically get it to make shit up by telling it that it is wrong.
It’s a giant Ponzi scheme by the rich and powerful to siphon more wealth into their pockets and further disablize the world.
We’re heading for a very depressing future in which not even the average person has less and less ways to sell their labour we won’t have capitalism.
Is it violent to say I’d rather we bulldoze silicon valley than Palestinian elementary schools?