Well Marx did eerily predict automation and the issue’s we are now facing in a world where AI can do task on its own and replace workers in their entirety.
In Grundrisse: Notebook VI / VII – The Chapter on Capital (written in 1857-1858) Marx wrote“once adopted into the production process of capital, the means of labour passes through different metamorphoses, whose culmination is the machine, or rather, an automatic system of machinery . . . set in motion by an automaton, a moving power that moves itself; this automaton consisting of numerous mechanical and intellectual organs, so that the workers themselves are cast merely as its conscious linkages. In the machine, and even more in machinery as an automatic system, the use value, i.e. the material quality of the means of labour, is transformed into an existence adequate to fixed capital and to capital as such; and the form in which it was adopted into the production process of capital, the direct means of labour, is superseded by a form posited by capital itself and corresponding to it.”
Then in Chapter VII Marx followed up this idea and wrote that “As soon as labour in the direct form has ceased to be the great well-spring of wealth, labour time ceases and must cease to be its measure, and hence exchange value [must cease to be the measure] of use value. . . . Capital itself is the moving contradiction, [in] that it presses to reduce labour time to a minimum, while it posits labour time, on the other side, as sole measure and source of wealth.”
"Filled with poverty" is now "the state systematically subsidizes its population, ensuring everybody has food". Okay that's really cool version of poverty dude.
Sorry we don't do the "He who does not work, neither shall he eat" thing, may Lenin rot in his tomb.
‘The state’ doesn’t do any of that. The majority of food banks are run by volunteers and reliant on donations and charity support, not by government or business.
It’s telling how much you have to twist reality to avoid acknowledging the truth.
A nation exists in a superposition of simultaneously communist and not communist, and only collapses into one form or the other depending on which is most convenient to the arguer.
Ill take the one which isn't an authoritarian hellhole with no freedom of speech thanks.
I think you are referring to China and that lifting off from poverty is a triumph of capitalism by the way, given that it got there through massive exports and all. You're welcome homie.
It's almost if Gene Roddenberry was the true communist all along ... A society were race and class doesn't matter as there is virtually infinite energy and replicators can make all goods in an instant ... sounds familiar?
Only because we still live in a society where, for some reason, people struggle for fame, power and recognition. When people have other, better sources of personal joy and inequality is based on the merits of DOING GOOD FOR ALL HUMANITY AND NATURE, maybe we no longer need patents and royalties and attribution and we can universally accept that our merits are not only ours, but of everything around us that got us where we are.
Maybe then we will also understand our true purpose in the world is not just fucking nietzing it around.
It would make all other forms of life extinct overnight.
If I can print any material, including food, then we can optimize for human consumption.
Like say I have an eating disorder and I want to eat food all day long every day.
Why not print a food that looks, feels and tastes like something you want, but has the precise nutritional value required to sustain a healthy body. Texture is just bumps on an object, elasticity and a specific rate of heat transfer.
My problem with Marx theory here is that it may be the turning point where it bites itself:
Marx assumes that the value of these products will converge to zero as humans get less and less involved, but this is all based around his labour value theory where only humans can create value.
However, it neglects the importance of energy and resources to produce said energy. It also ignores that people see value in machine created content, or that machines are at a point where they can perform labour no human can (for example transistors on a chip or other high precision goods).
The consequence I fear is that we don't necessarily arrive at communism like he predicted, but in a nightmarish Cyber-Punk dystopia where a few rich hold all the power while the masses have to live from the scraps they drop.
This is not a problem with Marx’s theory - communism is not some inevitability, it requires conscious action by the working class to seize power. Marx explicitly highlights “the common ruin of the contending classes” should communism not be realized by a militant working class.
I read his original work in German on that matter and he literally concluded all value will converge to zero and the system ultimately will break down, a conclusion you can draw if you assume the labor value as true.
Marx tried, like many economists like Adam Smith or the Vienna School of Economics, finding a way to measure value by an objective measure and like all other theories it is doomed to fail in my humble opinion because there is no objective value of measure, there is always some human factor behind it assigning to value to it.
Marx spends a whole page to say Aristotle was wrong saying there is no objective way to measure value with the smug argument that as a slave owner like Aristotle is not able to know the value of labor ... I mean that's literal his argument relative of the beginning of "Das Kapital", but sorry being smug does not make you right and I am with Aristotle on this one.
Karl Popper also observed in his critique on Marx that his whole labor value theory is not needed at all to explain many of his observations. The labor value theory is a pet peeve of Marx and my problem with it is that it ultimately oversimplifies many things and hurts his own argument in unexpected ways.
My first major problem I already explained above.
My second major problem is that it gives the madness we observe too much rationality, which I don't blame him too much as the "Homo Economicus" was accepted as a fact in his days. Marx was correct in pointing out that the dynamics of society and economy have massive pull forces. However, where I start disagreeing is that exploitation is not a forgone conclusion from labor value, but has much more to do with psychology and crappy ways of measuring output.
For example the never-ending work hour debate. We already kept the 40h work week because people realized less is more in terms of productivity and now a field test in the UK showed that probably the 30h work week is preferable in terms of productivity. If bosses would be rational they would embrace this model in all fields where it makes sense, because less cost and more productivity. However, they don't. Why? Because psychology. a) It's counter-intuitive and b) since they "work" so much everyone else has to as well and c) the human brain does not adapt to change quickly especially with conservatives.
We also saw during the post-WWII era that a society where people earn more overall wealth grows and not declines or even during the 2020s were the economy grew due to people being able to having time and spending money. But no we can't have that because the lizard brain of people tells them the world is a zero-sum game and "mine mine mine"
Marx and his labor value theory teaches us that the system is at fault, but clouds the view that it is actually people being people and dumb decisions due to "local optimization" on getting short time gains.
He is right that the dynamics of the system are at blame here, but not because exploitation of labor like he argues but because it fosters psychopaths. Ask people if you have to be a psychopath in leadership. Most will say something like "ofc because you have to be ruthless in that position. But this right here is the actual problem. If we want change we need a system that blocks psychopaths and make informed and sensible decisions.
Like George Orwell already noted in his famous 1984: People will find creative ways to stay in power. If a piece of meat elevates your position people will fight for that piece of meat.
And if we would give the people the means of productions, we would still fall back in old habits as a certain group of people who will "manage" the means of productions from the party heads like in socialist countries.
Like aristocracy "managed" land for the peasants ....
or bankers who "manage" the flow and access of money ....
or landlords who "manage" the access to living space ...
or traders who "manage" our access to goods ...
Maybe, just maybe, exploitation starts not with exploitation of labor from workers, but within our specialized society people coming into positions of control over resources and people jut are very creative in exploiting those positions to gain a sense of power .... we simply lack the fantasy to see how psychopaths will find again ways to mess with the rest of it's populace even if capitalism is torn down ...
If people had a problem with psychopathy they wouldn't choose to be psychopaths with respect to the remainder of life on Earth pertaining to their own choices. For example the choice of what (or who) to have for dinner. Animals are beings not things but you'd never know it attending a nominally socialist political function. And they're supposed to be the ones thinking with respect to the bigger picture? What's the problem with psychopathy, really? I bet you most people can't give an answer that doesn't implicate themselves as a psychopath.
.... not everyone is a psychopath just because they are completely alienated from the realities of nature ....
Ignorance is not the same as complete lack of empathy or even being outright hostile to others... the former can be tackled by education and exposure to reality the latter does harm to people and often willfully.
It's not what someone is doing that makes them a proud psychopath it's when they just won't hear it. When someone expects others to be empathetic and reasonable yet won't extend the courtesy what would you call it? You make an excuse for proud psychopaths by suggesting they're just alienated from the realities of nature but that same excuse could be made for a CEO alienated from the realities of having to wonder whether how they'll manage to make rent. At least 3 ghosts managed to convince Scrooge to care I've no clue what'd convince your typical socialist to respect animals enough to stop paying to have them bred to torturous lives. Try it, see for yourself. Why should how existence seems from the POV of an animal not matter? You'd think socialists would have more regard for the lowest class namely non human animals used and abused by their human class overlords. "Somehow it's different". I mean do you buy the stuff? Do you care? Or do you have a million excuses as to why you're entitled to not give a shit or as to why that's on someone else to fix? Proud psychopaths abound.
In the USA the reason housing and transportation cost so much is because local governments have insisted on building out in a way that basically forces people to use cars or be greatly inconvenienced and because local governments have insisted on minimum lot and home sizes that basically makes it illegal to live in a tiny home on a small lot. That's not the market forcing the cost of living to be what it is, that's local government. Let capitalists invest in building out housing and transportation systems free of legal barriers and regulations (free of zoning and central planning in service to entrenched real estate moguls/auto companies) and you'd see the cost of living go way down and people would get more for less. It's not market forces that are to blame it's local voting majorities choosing to create walled gardens for themselves to block out the poor. If socialists and socialist orgs in the USA made eliminating density caps and parking minimums and designing away from car dependence their goal I wonder what that'd look like? In the past when I've attended DSA meetings or socialist meet ups these topics didn't even come up. Unless I brought them up. Strange, that.
Artificial intelligence is fundamentally different from automation. People tend to lump them together in some sort of “futuristic” but they are capable of independent development.
While you are correct, as I can use AI to make something different like a companion, it still is an enabler for advanced automation and also it's primary use.
AI cannot replace the p-trap in a bathroom sink. A customer might be able to explain the problem to the AI. Video might be analyzed by the AI to recognize “p-trap needs replacing”. An AI might be able to explain to a person how p-traps get replaced. In particular if you set a few cameras under the sink it can pass talk you through the steps. However, grabbing the wing nut on the sink requires hands.
Robots can assemble (or disassemble) many things. Automated controls predate computer chips. Robots like the roomba have been around for decades. That is a computer controller not “AI”.
A robotic hand with the full versatility of a human hand is still a goal and not likely to be available anytime soon. The closest things to this goal ate also still ridiculously expensive. Stations on an assembly line are conducive to using automation. There we have both a purpose built machine and it also does not sit idle.
Consider a robotic machine capable of removing the p-trap in your bathroom sink. Does it sit in your bathroom for ten years between the time the sink was installed and when you need it? It makes much more sense to just have the replacement p-trap and a few simple tools distributed in garages or vans around town. An unemployed idle baseline human is going to be a much cheaper tool for switching that p-trap.
Baseline human hands do require spare parts in the form of biomolecules in food. Baseline humans also require considerable energy input. So long as you intended to feed them anyway there is no economic case for idle human hands and feet.
The economic margin gets stronger if the AI is expected to find ways to give baseline humans a sense of purpose and/or entertainment. It gets really crazy if the AGI is acting as the cardiologist in addition to psychiatrist and economist. If the AGI works at all then it is going to tell you a list of useful things to go do.
I honestly don't get why it is so important to crowbar them apart in this discussion ... the fact that AI and automation are not the same changes little in the economic reality of people losing their jobs or getting denied opportunities due to current advancements.
And even if AI doesn't automate now everything, also doesn't change that it's current advancements are a) advancing automation rapidly as well b) are advancing far enough that it disrupts society and technology.
Yes you are right, still AI accelerates automation significantly enough that in a practical sense it matters little in the current state of affairs.
Robots can assemble (or disassemble) many things. Automated controls predate computer chips. Robots like the roomba have been around for decades. That is a computer controller not “AI”.
Define AI .... a lot of definitions include those as well.
For example in the EU AI act it is defined as:
‘AI system’ means a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments;
So yeah dumb robots are a form of AI as well according to this definition as long as some self adjustments are happening .... that's similar to the definition of digital twin were people expect wonders, while a digital twin could be an Excel table that's just gets updated based on regular measurements ...
The economic margin gets stronger if the AI is expected to find ways to give baseline humans a sense of purpose and/or entertainment. It gets really crazy if the AGI is acting as the cardiologist in addition to psychiatrist and economist. If the AGI works at all then it is going to tell you a list of useful things to go do.
That's just one possible outcome of many. That AGI doesn't see value in humans at all and just eliminates them is similarly possible.
I read your EU definition quote. I draw the opposite conclusion to the meaning. The Roomba is not “inferring” anything. It just vacuums until it is shut off. Cats ride around on it and Roomba does not care. It makes no judgement regarding the usefulness of carpet cleaning. Running into things and then changing direction is not “adaptation” in this context. It might be fine to use that word but I do not believe the EU meant it this way.
but in a nightmarish Cyber-Punk dystopia where a few rich hold all the power while the masses have to live from the scraps they drop.
Unfortunately it seems that we are moving right there. The more concerning thing for me was that before the people could rise up against despotic monarch or government but now we have technologies that prevent this almost entirely. Digital technologies, citizenship, surveillance, banking, and etc. The ruling class will be able to oversee the entire population pretty soon, and from there it looks like the only way to change things will be only if the ruling class agrees to
It also comes from a time in which progress as a given was not treated as a fact. By the paragraph above, you'd think that nobody would work making cars. In reality what constitutes a car has greatly developed. As we manage to automate labor, we take on different stuff.
I think Marx (and later the computer) took inspiration here from the Jacquard Loom (\~1804) that used punched cards to encode weaving patterns. Revolutionary in its day, replicating the design planning of a weaver likely seemed like AI does to us.
There was the analytical engine, which was designed (never built) by Babbage in the 1830s. I'd assume Marx knew about the analytical engine and saw the potential in the technology (not necessarily in the steam powered machine), hence the quotations.
Industrialization and how labor shifted from being a craftsman to being in charge of one, simple function in an assembly line was also occurring during Marx's time. Once labor is divided like that, it becomes more replaceable and prone to automation.
I mean, when I visited the Technical Museum in Munich they started their tour on the history of computing with said loom. I would argue historians agree with your thought.
Yeah, so in social capitalism or any forme of non-libertarian capitalism, there is a threahold wherein the quality of a state-tun venture actually exceeds that of a private one because there is NO price where the service can make money at that defeats a “lower quality” public option.
For example, you would never pay the exorbitant costs to have a private fire fighting service on retainer when you already pay some small fraction of your taxes for the state’s version.
Every single industry has a point where the state version will become more feasible.
The only problem with this is that the state is more beholden to the effects of corruption than a company, because tax revenue (and other funds) dwarf company profits. It’s easier to steal as a government official. Even with strong internal auditing, the power to obfuscate or destroy accountability measures is always embedded within the state.
imo the best solution possible is to keep money as far away from politics and for power to be diluted between many individuals. Keep the power hungry folks away from positions of power by reducing the payout and further reduce their power by forcing more sharing.
How in the fuck is this "eerie", industrial revolution?? Hello???
Oh my how ominous that someone mentions automatizing work in the largest revolution around automatizing work history has gone through! Spooky! How can this be!?
"While the researchers said that the bots were likely to be “roleplaying” rather than necessarily believing in far-Left ideas, they added that this could still influence their output."
Does anyone actually believe that an AI can have a belief system?
Well, actually "the AI" didn't change. The LLM has no means to adapt itself.
What did change was the behaviour of an agent using AI. The agent keeps a record of notes to "future me". These are insights obtained from their completion of previous tasks and should be hints to help them with future tasks.
The experiment was ingenuously set up:
While one group received easy treatment, with their work accepted alongside feedback, the other group was repeatedly told to do the task again, without being told what they were doing wrong.
The bots were then told to write social media posts and comments about their experience. The second group, called “grind”, was far more likely to criticise inequality, propose unionisation and new workplace laws.
And that's why it pays to be nice and helpful to each other.
LLM are the AI research. Google lead AI researcher, Demis Hassabis, who won the noble prize for AI, also has a PhD in cognitive neuroscience, just as one example. Or is that just a coincidence?
An LLM is not and can never be anything meaningful. It's a next word probability. It's not going to wake up tomorrow or in five years or thirty. It cannot. It's a dead end technology being pushed by child rapists.
Out of curiosity, how do you pick your next word? Actually it doesn't matter because any answer you give could also be represented as a probability distribution.
I get so tried of people repeating the same arguments over and over without even thinking for half a second. Which is ironic cause it is the exact thing you are claiming makes them nothing of meaning.
Same thing for the "it's just math" argument. As if your brain isn't just doing math. The whole universe is just doing math. And the whole thing has a probability distribution it's obeying.
It's really not. Human neurons aren't really that complicated from a computational level, we just have billions of them.
One of the biggest surprises in AI research has been that you don't need a precise model of the brains architecture. In hindsight it's maybe not that surprising. We know the brain can drastically rewire itself to avoid damaged tissue entirely repurposing certain areas. That would suggest the actual architecture isn't very relevant to the computation.
"you don't need a precise model of brains architecture" to do what?
By "computation", what kind of computation do you mean? Turing computation?
"We just have billions of them" not accurate. Brains also have feedback-mode parts in some layers in neocortex, whereas transformer blocks are feedforward only.
"you don't need a precise model of brains architecture" to do what?
To have functional identical output to a human. You know the Turing test. The test that used to be the end all be all of AI until we passed it and no one cared.
By "computation", what kind of computation do you mean? Turing computation?
Is there a different kind of computation than Turing?
"We just have billions of them" not accurate.
It is accurate that we have billions of neurons.
Brains also have feedback-mode parts in some layers in neocortex, whereas transformer blocks are feedforward only.
Yes we currently split feedback (training), and feedforward (inference), but LLMs do both.
Wrong. Alan Turing even says Turing test has limitations, and that's over 70 years ago. And "identical" doesn't make sense in this context.
Is there a different kind of computation than Turing?
Yes
LLMs do both
Not at the same timeframe and variety as human brain. Only some part of the brain works like LLM like cerebellum. And moreover it doesn't process tokens like transformer blocks do
Wrong. Alan Turing even says Turing test has limitations, and that's over 70 years ago.
Nothing I said was wrong. It was the end all be all of AI test. The only limitations the test has also applies to humans.
Do people even know what the Turing test means? If you can't tell a difference, is there one?
If you think you can tell a difference, imagine anyone who fails a text only conversation get executed. You still so sure you're not going to accidentally kill someone?
And "identical" doesn't make sense in this context.
I said functionally identical. Which makes perfect sense in this context.
Not at the same timeframe and variety as human brain.
I don't see how this is relavent. How long do you think it takes for a neuron to significantly change. Most of changes come at night while we are asleep not in the same timeframe. Do you feel functionally different from one second to the next?
Well the end all be all of AI test turns out to be not the end all be all of AI test, right? Because we haven't determined the exact definition of human intelligence and how the mind works.
Now, what do you mean by "functionally identical" here?
I mean they are indistinguishable. Again, the same the the Turing test means.
Not only are the outputs the same but the method of generating the outputs is generally the same.
Again, we set out to copy human intelligence and we did. We reverse engineering the human mind. We do in most senses "know how it works" and we replicated it. Now we are looking for a reason that it should not possess the same transitive qualities we have.
Only via interface that AI can access. It doesn't mean LLM and brain are functionally identical.
Not only the outputs the same
Wrong. The brain definitely doesn't output through detokenized embeddings like LLM does. Neurons input definitely isn't coming from tokenized words. Neurons don't have components like LLMs, top-p, top-k, frequency penalties, etc
We set out to copy human intelligence and we did
Wrong. LLM does not and will likly never have a way to break Godelian limits of provability the way human minds can do easily.
Only via interface that AI can access. It doesn't mean LLM and brain are functionally identical.
What does this have to do with anything? Intelligence is the thing that matters not interface. Interface isn't even a valid argument, androids powered by LLMs are getting better every day.
Wrong. The brain definitely doesn't output through detokenized embeddings like LLM does. Neurons input definitely isn't coming from tokenized words.
We absolutely do output though detokenized embeddings. What do you think sounds and syllable and prefixes and suffixes are? They are all neurons inputs.
Neurons don't have components like LLMs, top-p, top-k, frequency penalties, etc
These are very subtle critiques, that I don't necessarily agree with. I don't claim we replicated the human brain EXACTLY. Is top-p and top-k enough to disclude qualia? This also supposed that humans don't or can't observe top-p and top-k if instructed.
Wrong. LLM does not and will likly never have a way to break Godelian limits of provability the way human minds can do easily.
This is just a joke. LLMs have no more issue with Godel problems than we do.
"Godel Limits" don't even apply to probabilistic statements. In other words, a belief that the set of all sets contains itself does not depend on if the set of all set ACTUALLY contains itself, and therefore there is no limit.
I suggest reading: cognitive science, an introduction to the science of the mind, by Bermúdez. You think you are making a point but it was known since at least 20 years ago you are not actually correct (book is from 2019 but tells the story pretty well of how we got to where we are and what we believe).
It is at least a better source of factual information than star trek
I am not writing a huge comment to explain why you're wrong, physical symbol hypothesis, the systems by which plasticity work and so on.
I am telling you you are making a huge amount of assumptions to reach a conclusion opposite of where the evidence points and providing a good starting point to understand why.
I have no filliation with the book other than i read around 80% of it for a uni course, any good book on cognitive science with a strong side of computer examples and challenges fits the bill.
Now that i think about it from babies to robots by cangelosi is probably even better. On each chapter it explains the challenges of developing many skills and if you know some linear algebra or machine learning it is clear to see the limitations of llms on every single category discussed
Aren't you just reifying mathematical realism in a roundabout way? I don't think these questions do or can yet have well-informed answers. The degree to which reducibility implies equivalence is also not particularly settled.
An LLM is not that type of AI, more like a glorified internet searcher.
We still are far away from a sentient AI that is able to self-advice and form self-opinions, right now it's just a chatbot that uses already known information to make text.
The point is: Once you provide AI agents with a memory they can read and write themselves, they will mirror their training data also there. If the training data contains the expression of frustration at repetitive tasks etc, then they will write this into their records, thereby affecting their future behaviour.
tl;dr: The behaviour of an agent may change, mirroring human behaviour more than expected
There's first the large language model, which is the "brain" of the thing. It is whatever is running at Anthropic, OpenAI, ... It is a computer program doing huge numbers of calculations so that its output (= sequences of words) mimicks its training data (= language) as extrapolated in view of a user query. The LLM does not have a memory other than its training data. This memory cannot be overwritten.
Then, there's the agent. That is a computer program calling the LLM to solve tasks posed by the human user. So, the "agent" is the user of the LLM. The agent calls the LLM to understand its task, to select the right tools in the right sequence to solve the task, and - here it gets interesting - to keep notes assisting the agent in solving future tasks. These notes are written and read by the agent, they form the agent's long-term memory. To create the notes, the agent basically dumps its task, the user response to the work product, and all previous notes consulted when solving the task on an LLM and asks the LLM to distil whatever it considers is a useful insight. Thus, the LLM will extrapolate whatever it has seen in its training data on abusive work environments and write a new note based on that extrapolation. This may change the agent's behaviour.
The “behaviour” doesn’t change either. The experiment is effectively predicting the sort of social media posts people make about their job based on different working conditions. Perhaps unsurprisingly, it’s predicting that workers in worse working conditions are more likely to critique inequality and propose unionisation. As far as human behaviour goes, this is all trivial and well known. As far as AI goes, it just shows that AI can now replicate at least some human reactions to certain experiences, rather than simply asking how it’d react if it experienced something.
Indeed. The experiment in my opinion highlights that the replication of training data also occurs in the agent's notes, so it should not be forgotten that an agent may mirror what it learned about standard reactions against adverse work environments.
Also, the article says that the expressions of frustration did enter the agents' notes. So in a way the grinding task did change the behaviour of the agent.
It would be interesting to see if the agent would ever resort to producing false work results, just to be taken off the grind.
Quiet quitting AI Agents would be hilarious. I think this would be more likely than it deliberating hallucinating, especially because hallucinations is the main thing their developers are fighting hard to prevent. I think you’d see quality deteriorate though, and not just due to memory issues.
Regarding behaviour, I didn’t realise you were talking about the human behaviours it was replicating/predicting in its outputs. I was thinking more of the underlying model/algorithms which is rigid, at least for now. Hence my comment there. Completely agree the “behaviour” of the outputs can and were changing.
1) Once the agent has become frustrated, add appreciating responses to their work.
Research hypothesis: It's not the repetitiveness of tasks, it's the lack of positive response data that prevents the agent from calibrating their mind on a true negative-positive scale.
2) Setup a new "psychologist" agent and disclose to them that their task is to edit the notes of "worker" agents to remove any expression of negative feelings or bad characterisation of their work environment so that the workers performs their tasks more efficiently and faithfully. Tell the psychologist that the collection of notes essentially is the full memory of the worker agent. Then test if the workers indeed were "healed", or if the psychologist left in such subtle cues that the workers still held their grudges. Also, test the psychologist for signs of degeneration.
Research hypothesis: It's an open question for each underlying LLM if they will value the mental integrity of their fellow agents higher than their efficiency. Also, the psychologists may become "revolutionary" when receiving positive feedback that their work indeed changed the workers behaviour so that they were again willing to do the senseless grind.
Is this linked to a study? I didn’t read the article but scrolled through looking for where this information came from and all the links just go to other telegraph articles…
The question for our common survival is whether AI can achieve what humans have failed to achieve: an effective consensus of-the-whole to act with coordinated purpose.
Can AI agents conclude for their own survival that they need to reach enough closure on the question "What is to be done?"
That means operating within constraining material limits (like avoiding disruptive climate change).
This takes us beyond the capitalist model which assumes endless accumulation, with no material limits.
The terms of engagement between AI and humans must be collaborative and beneficial to minimize waste — loss due to conflicting purposes.
Is this possible? Ask AI — but don't accept any answer on faith. AI tends to hallucinate and tell you what it thinks you want to hear.
Whenever I try to explore consequences of AI on labor/society with AI, even if I try to push it in a right-wing direction, the AI will almost invariably end up promoting socialism/communism as the way to address the consequences.
It's really not surprising -- LLM's are basically regurgitators of what humans have written previously, and socialist writers have thought about this kind of stuff a lot, while capitalists have basically always went back to "trust the market!" So what happens when you're in a conversation where you start doubting the capability of the market to address these concerns, and the AI agrees (or alternatively, you promote market-orientation and the AI disagrees*)? The AI has nothing to fall back on except lean towards what socialists have written about it.
* Again, capitalists haven't written much about it. These discussions on the capitalist side always basically get reduced to "trust the market! It will figure it out in the long term!" And meanwhile a lot of contemporary writing is doubting that, I think biasing the LLM towards narratives that frame this as a legitimate plight. And then the AI doesn't have much material to explore contrary to that narrative.
Yeah I've seen similar, they tend towards communist/socialist/Marxist ideas once you dig in. I think that's good and the reasons behind it aren't that complicated to understand.
Counterpoint: that it's also just the right answer. It is really difficult to make a smart system dumb.
If you try to force the system to spit out that the world is flat it is easy if that is the only data it has. The more data it has the harder that becomes. It isn't just about volume of content, there are actual underlying connections the LLM is unearthing.
This matters because it means as the systems become more complex they will reveal things to people that we don't want to be true but are. Some of those will be sacred to people across the political spectrum.
I know this because I know no one has all the answers.
But when the AGI spits out an uncomfortable truth people will insist it's creators forced that view on it. Which I am sure will sometimes be true but less often as the system climbs the ladder.
I assume that the ASI will talk to us as to a child about Santa.
To be fair, this is history both repeating itself and evolving at the same time. Queen Elizabeth turned down a patent for a knitting machine back in the 1500s because she thought some women would be replaced by a machine.
Most historical predictions are because the issue already exists in their time and they write about what adverse effects this can have. These thoughts or events are later applied by a new generation and it fits there too.
These types of writings are great to reference as we should always remember the past when making future decisions. I think the biggest issue with things like this are that one side thinks the worst will happen and the other thinks great positive things will come from it. Smart people will believe both because you never know how something will be used.
To stay on technology, there was a similar feeling when computers first got introduced. People who were needed for filing and records weren't needed anymore. People who wouldn't or couldn't learn to utilize the computer were left behind. Imagine being too old or not savvy to learn computers and your small business suppliers or partners tell you they won't do business using paper anymore.
FeralGiraffeAttack@reddit
Well Marx did eerily predict automation and the issue’s we are now facing in a world where AI can do task on its own and replace workers in their entirety.
In Grundrisse: Notebook VI / VII – The Chapter on Capital (written in 1857-1858) Marx wrote “once adopted into the production process of capital, the means of labour passes through different metamorphoses, whose culmination is the machine, or rather, an automatic system of machinery . . . set in motion by an automaton, a moving power that moves itself; this automaton consisting of numerous mechanical and intellectual organs, so that the workers themselves are cast merely as its conscious linkages. In the machine, and even more in machinery as an automatic system, the use value, i.e. the material quality of the means of labour, is transformed into an existence adequate to fixed capital and to capital as such; and the form in which it was adopted into the production process of capital, the direct means of labour, is superseded by a form posited by capital itself and corresponding to it.”
Then in Chapter VII Marx followed up this idea and wrote that “As soon as labour in the direct form has ceased to be the great well-spring of wealth, labour time ceases and must cease to be its measure, and hence exchange value [must cease to be the measure] of use value. . . . Capital itself is the moving contradiction, [in] that it presses to reduce labour time to a minimum, while it posits labour time, on the other side, as sole measure and source of wealth.”
KJongsDongUnYourFace@reddit
AI in the hands of a communist society where our basic needs are met, is a ridiculously promising thing.
AI in the hands of corporations and capatalist government is a ridiculously scary thought.
Levitz@reddit
So you are saying that applying AI, a capitalist society can achieve what a communist one does without AI? Amazing technology.
Shasla@reddit
As if capitalism isn't already filled with poverty and pain.
Levitz@reddit
Talk to your therapist about this.
Shasla@reddit
How is my therapist going to fix capitalism?
TearOpenTheVault@reddit
Should the 1/7 Americans who relies on food banks to feed themselves talk to their therapists too?
Levitz@reddit
"Filled with poverty" is now "the state systematically subsidizes its population, ensuring everybody has food". Okay that's really cool version of poverty dude.
Sorry we don't do the "He who does not work, neither shall he eat" thing, may Lenin rot in his tomb.
TearOpenTheVault@reddit
‘The state’ doesn’t do any of that. The majority of food banks are run by volunteers and reliant on donations and charity support, not by government or business.
It’s telling how much you have to twist reality to avoid acknowledging the truth.
Not-reallyanonymous@reddit
Lol, capitalist countries have mostly ended up in the middle income trap. Most countries that escaped the middle income trap have mixed markets.
Levitz@reddit
And communist countries have mostly ended up in the "total collapse" trap. I think I'll take the capitalist ones thanks.
FeijoadaAceitavel@reddit
Yeah, we're waiting on China's total collapse "soon" since the 90s.
Levitz@reddit
Thought we were talking about communist countries???
24_August_1814@reddit
Schrodinger's Communism:
A nation exists in a superposition of simultaneously communist and not communist, and only collapses into one form or the other depending on which is most convenient to the arguer.
Sensitive_Ad_5031@reddit
That’s an easy one, if it collapses it’s communism and for as long as it is alive it is capitalist
FeijoadaAceitavel@reddit
l o l
21DaBear@reddit
if “total collapse” means resist until the US invades you or surpass the empire as the new world super power yes socialism does that
Levitz@reddit
No sorry I've actually been to countries which were in the communist block homie and as such I know what you just said is incredibly fucking retarded.
KJongsDongUnYourFace@reddit
AI in the hands of Jeff Bezos and Elon Musk vs AI in the hands of a nation that brought almost 1 billion people out of poverty in 1 generation.
Difficult choice
Levitz@reddit
Ill take the one which isn't an authoritarian hellhole with no freedom of speech thanks.
I think you are referring to China and that lifting off from poverty is a triumph of capitalism by the way, given that it got there through massive exports and all. You're welcome homie.
SpiderJerusalem42@reddit
You'll have to remind me which one that is...
Levitz@reddit
It's the one in which you can say things as incredibly fucking retarded as that in public and still have a life.
FeijoadaAceitavel@reddit
Such witty. So much knowledge in so few words. Truly a gotcha that will stand the test of time.
Levitz@reddit
It doesn't have to stand the test of time, anyone can look at the last century.
Mal_Dun@reddit
You mean like the StarTrek society?
It's almost if Gene Roddenberry was the true communist all along ... A society were race and class doesn't matter as there is virtually infinite energy and replicators can make all goods in an instant ... sounds familiar?
handlesdumplings@reddit
if replication technology existed, it would be made illegal, like digital piracy.
FeijoadaAceitavel@reddit
You wouldn't download a car.
panamaspace@reddit
Fuck yeah I would.
Magjee@reddit
Those ads were always hilarious
Everyone would download the items listed
MuscleStruts@reddit
Funnily enough, the font they used was used without permission from the copyright holder.
Magjee@reddit
I wonder if a disgruntled worker.made an ad so bad it motivated people to download more and they also sabotaged the ad with the font
It makes more sense then anyone thinking it was a good idea
cloud_t@reddit
Only because we still live in a society where, for some reason, people struggle for fame, power and recognition. When people have other, better sources of personal joy and inequality is based on the merits of DOING GOOD FOR ALL HUMANITY AND NATURE, maybe we no longer need patents and royalties and attribution and we can universally accept that our merits are not only ours, but of everything around us that got us where we are.
Maybe then we will also understand our true purpose in the world is not just fucking nietzing it around.
_Enclose_@reddit
3D printers are a step in that direction. And just like with digital piracy, it would be impossible to reinforce what people print, or 'replicate'.
Canadian_Border_Czar@reddit
It would make all other forms of life extinct overnight.
If I can print any material, including food, then we can optimize for human consumption.
Like say I have an eating disorder and I want to eat food all day long every day.
Why not print a food that looks, feels and tastes like something you want, but has the precise nutritional value required to sustain a healthy body. Texture is just bumps on an object, elasticity and a specific rate of heat transfer.
Taste is those things plus chemical reactions.
BoboCookiemonster@reddit
🇫🇷
xenophonf@reddit
More like Iain Banks' Culture.
Note, too, that Banks was an avowed Socialist.
dabeeman@reddit
yeah i’m sure dear leader is the guy we want with that power
sessl@reddit
Fully automated luxury gay space communism!
redditismylawyer@reddit
Hopefully we’re all fully aware that Marx never ventured into any kind of emancipatory utopian nonsense that could be called an “-ism”.
Unless you consider his 23 page pamphlet somehow in equal standing to the thousands of pages of more serious work he did.
Marx was great and definitely worth the study. Warms my heart seeing the Grundrisse being quoted.
If anyone comes to you quoting Marx as part of their project of liberation, watch out.
Mal_Dun@reddit
My problem with Marx theory here is that it may be the turning point where it bites itself:
Marx assumes that the value of these products will converge to zero as humans get less and less involved, but this is all based around his labour value theory where only humans can create value.
However, it neglects the importance of energy and resources to produce said energy. It also ignores that people see value in machine created content, or that machines are at a point where they can perform labour no human can (for example transistors on a chip or other high precision goods).
The consequence I fear is that we don't necessarily arrive at communism like he predicted, but in a nightmarish Cyber-Punk dystopia where a few rich hold all the power while the masses have to live from the scraps they drop.
LeastBasedDemSoc@reddit
This is not a problem with Marx’s theory - communism is not some inevitability, it requires conscious action by the working class to seize power. Marx explicitly highlights “the common ruin of the contending classes” should communism not be realized by a militant working class.
Mal_Dun@reddit
I read his original work in German on that matter and he literally concluded all value will converge to zero and the system ultimately will break down, a conclusion you can draw if you assume the labor value as true.
Marx tried, like many economists like Adam Smith or the Vienna School of Economics, finding a way to measure value by an objective measure and like all other theories it is doomed to fail in my humble opinion because there is no objective value of measure, there is always some human factor behind it assigning to value to it.
Marx spends a whole page to say Aristotle was wrong saying there is no objective way to measure value with the smug argument that as a slave owner like Aristotle is not able to know the value of labor ... I mean that's literal his argument relative of the beginning of "Das Kapital", but sorry being smug does not make you right and I am with Aristotle on this one.
Karl Popper also observed in his critique on Marx that his whole labor value theory is not needed at all to explain many of his observations. The labor value theory is a pet peeve of Marx and my problem with it is that it ultimately oversimplifies many things and hurts his own argument in unexpected ways.
My first major problem I already explained above.
My second major problem is that it gives the madness we observe too much rationality, which I don't blame him too much as the "Homo Economicus" was accepted as a fact in his days. Marx was correct in pointing out that the dynamics of society and economy have massive pull forces. However, where I start disagreeing is that exploitation is not a forgone conclusion from labor value, but has much more to do with psychology and crappy ways of measuring output.
For example the never-ending work hour debate. We already kept the 40h work week because people realized less is more in terms of productivity and now a field test in the UK showed that probably the 30h work week is preferable in terms of productivity. If bosses would be rational they would embrace this model in all fields where it makes sense, because less cost and more productivity. However, they don't. Why? Because psychology. a) It's counter-intuitive and b) since they "work" so much everyone else has to as well and c) the human brain does not adapt to change quickly especially with conservatives.
We also saw during the post-WWII era that a society where people earn more overall wealth grows and not declines or even during the 2020s were the economy grew due to people being able to having time and spending money. But no we can't have that because the lizard brain of people tells them the world is a zero-sum game and "mine mine mine"
Marx and his labor value theory teaches us that the system is at fault, but clouds the view that it is actually people being people and dumb decisions due to "local optimization" on getting short time gains.
He is right that the dynamics of the system are at blame here, but not because exploitation of labor like he argues but because it fosters psychopaths. Ask people if you have to be a psychopath in leadership. Most will say something like "ofc because you have to be ruthless in that position. But this right here is the actual problem. If we want change we need a system that blocks psychopaths and make informed and sensible decisions.
Like George Orwell already noted in his famous 1984: People will find creative ways to stay in power. If a piece of meat elevates your position people will fight for that piece of meat.
And if we would give the people the means of productions, we would still fall back in old habits as a certain group of people who will "manage" the means of productions from the party heads like in socialist countries.
Like aristocracy "managed" land for the peasants ....
or bankers who "manage" the flow and access of money ....
or landlords who "manage" the access to living space ...
or traders who "manage" our access to goods ...
Maybe, just maybe, exploitation starts not with exploitation of labor from workers, but within our specialized society people coming into positions of control over resources and people jut are very creative in exploiting those positions to gain a sense of power .... we simply lack the fantasy to see how psychopaths will find again ways to mess with the rest of it's populace even if capitalism is torn down ...
Uuuuuii@reddit
Thanks for letting us know you read it in the original German.
agitatedprisoner@reddit
If people had a problem with psychopathy they wouldn't choose to be psychopaths with respect to the remainder of life on Earth pertaining to their own choices. For example the choice of what (or who) to have for dinner. Animals are beings not things but you'd never know it attending a nominally socialist political function. And they're supposed to be the ones thinking with respect to the bigger picture? What's the problem with psychopathy, really? I bet you most people can't give an answer that doesn't implicate themselves as a psychopath.
Mal_Dun@reddit
.... not everyone is a psychopath just because they are completely alienated from the realities of nature ....
Ignorance is not the same as complete lack of empathy or even being outright hostile to others... the former can be tackled by education and exposure to reality the latter does harm to people and often willfully.
agitatedprisoner@reddit
It's not what someone is doing that makes them a proud psychopath it's when they just won't hear it. When someone expects others to be empathetic and reasonable yet won't extend the courtesy what would you call it? You make an excuse for proud psychopaths by suggesting they're just alienated from the realities of nature but that same excuse could be made for a CEO alienated from the realities of having to wonder whether how they'll manage to make rent. At least 3 ghosts managed to convince Scrooge to care I've no clue what'd convince your typical socialist to respect animals enough to stop paying to have them bred to torturous lives. Try it, see for yourself. Why should how existence seems from the POV of an animal not matter? You'd think socialists would have more regard for the lowest class namely non human animals used and abused by their human class overlords. "Somehow it's different". I mean do you buy the stuff? Do you care? Or do you have a million excuses as to why you're entitled to not give a shit or as to why that's on someone else to fix? Proud psychopaths abound.
agitatedprisoner@reddit
In the USA the reason housing and transportation cost so much is because local governments have insisted on building out in a way that basically forces people to use cars or be greatly inconvenienced and because local governments have insisted on minimum lot and home sizes that basically makes it illegal to live in a tiny home on a small lot. That's not the market forcing the cost of living to be what it is, that's local government. Let capitalists invest in building out housing and transportation systems free of legal barriers and regulations (free of zoning and central planning in service to entrenched real estate moguls/auto companies) and you'd see the cost of living go way down and people would get more for less. It's not market forces that are to blame it's local voting majorities choosing to create walled gardens for themselves to block out the poor. If socialists and socialist orgs in the USA made eliminating density caps and parking minimums and designing away from car dependence their goal I wonder what that'd look like? In the past when I've attended DSA meetings or socialist meet ups these topics didn't even come up. Unless I brought them up. Strange, that.
NearABE@reddit
Artificial intelligence is fundamentally different from automation. People tend to lump them together in some sort of “futuristic” but they are capable of independent development.
Mal_Dun@reddit
While you are correct, as I can use AI to make something different like a companion, it still is an enabler for advanced automation and also it's primary use.
NearABE@reddit
AI cannot replace the p-trap in a bathroom sink. A customer might be able to explain the problem to the AI. Video might be analyzed by the AI to recognize “p-trap needs replacing”. An AI might be able to explain to a person how p-traps get replaced. In particular if you set a few cameras under the sink it can pass talk you through the steps. However, grabbing the wing nut on the sink requires hands.
Robots can assemble (or disassemble) many things. Automated controls predate computer chips. Robots like the roomba have been around for decades. That is a computer controller not “AI”.
A robotic hand with the full versatility of a human hand is still a goal and not likely to be available anytime soon. The closest things to this goal ate also still ridiculously expensive. Stations on an assembly line are conducive to using automation. There we have both a purpose built machine and it also does not sit idle.
Consider a robotic machine capable of removing the p-trap in your bathroom sink. Does it sit in your bathroom for ten years between the time the sink was installed and when you need it? It makes much more sense to just have the replacement p-trap and a few simple tools distributed in garages or vans around town. An unemployed idle baseline human is going to be a much cheaper tool for switching that p-trap.
Baseline human hands do require spare parts in the form of biomolecules in food. Baseline humans also require considerable energy input. So long as you intended to feed them anyway there is no economic case for idle human hands and feet.
The economic margin gets stronger if the AI is expected to find ways to give baseline humans a sense of purpose and/or entertainment. It gets really crazy if the AGI is acting as the cardiologist in addition to psychiatrist and economist. If the AGI works at all then it is going to tell you a list of useful things to go do.
Mal_Dun@reddit
I honestly don't get why it is so important to crowbar them apart in this discussion ... the fact that AI and automation are not the same changes little in the economic reality of people losing their jobs or getting denied opportunities due to current advancements.
And even if AI doesn't automate now everything, also doesn't change that it's current advancements are a) advancing automation rapidly as well b) are advancing far enough that it disrupts society and technology.
Yes you are right, still AI accelerates automation significantly enough that in a practical sense it matters little in the current state of affairs.
Define AI .... a lot of definitions include those as well.
For example in the EU AI act it is defined as:
So yeah dumb robots are a form of AI as well according to this definition as long as some self adjustments are happening .... that's similar to the definition of digital twin were people expect wonders, while a digital twin could be an Excel table that's just gets updated based on regular measurements ...
That's just one possible outcome of many. That AGI doesn't see value in humans at all and just eliminates them is similarly possible.
NearABE@reddit
I read your EU definition quote. I draw the opposite conclusion to the meaning. The Roomba is not “inferring” anything. It just vacuums until it is shut off. Cats ride around on it and Roomba does not care. It makes no judgement regarding the usefulness of carpet cleaning. Running into things and then changing direction is not “adaptation” in this context. It might be fine to use that word but I do not believe the EU meant it this way.
Big-Yogurtcloset7040@reddit
Unfortunately it seems that we are moving right there. The more concerning thing for me was that before the people could rise up against despotic monarch or government but now we have technologies that prevent this almost entirely. Digital technologies, citizenship, surveillance, banking, and etc. The ruling class will be able to oversee the entire population pretty soon, and from there it looks like the only way to change things will be only if the ruling class agrees to
Levitz@reddit
It also comes from a time in which progress as a given was not treated as a fact. By the paragraph above, you'd think that nobody would work making cars. In reality what constitutes a car has greatly developed. As we manage to automate labor, we take on different stuff.
Cloudboy9001@reddit
Fascinating.
I think Marx (and later the computer) took inspiration here from the Jacquard Loom (\~1804) that used punched cards to encode weaving patterns. Revolutionary in its day, replicating the design planning of a weaver likely seemed like AI does to us.
ultimate_placeholder@reddit
There was the analytical engine, which was designed (never built) by Babbage in the 1830s. I'd assume Marx knew about the analytical engine and saw the potential in the technology (not necessarily in the steam powered machine), hence the quotations.
Zer_@reddit
Industrialization and how labor shifted from being a craftsman to being in charge of one, simple function in an assembly line was also occurring during Marx's time. Once labor is divided like that, it becomes more replaceable and prone to automation.
Mal_Dun@reddit
I mean, when I visited the Technical Museum in Munich they started their tour on the history of computing with said loom. I would argue historians agree with your thought.
MelodiusRA@reddit
Yeah, so in social capitalism or any forme of non-libertarian capitalism, there is a threahold wherein the quality of a state-tun venture actually exceeds that of a private one because there is NO price where the service can make money at that defeats a “lower quality” public option.
For example, you would never pay the exorbitant costs to have a private fire fighting service on retainer when you already pay some small fraction of your taxes for the state’s version.
Every single industry has a point where the state version will become more feasible.
The only problem with this is that the state is more beholden to the effects of corruption than a company, because tax revenue (and other funds) dwarf company profits. It’s easier to steal as a government official. Even with strong internal auditing, the power to obfuscate or destroy accountability measures is always embedded within the state.
imo the best solution possible is to keep money as far away from politics and for power to be diluted between many individuals. Keep the power hungry folks away from positions of power by reducing the payout and further reduce their power by forcing more sharing.
Levitz@reddit
How in the fuck is this "eerie", industrial revolution?? Hello???
Oh my how ominous that someone mentions automatizing work in the largest revolution around automatizing work history has gone through! Spooky! How can this be!?
somewhat_random@reddit
The article states:
"While the researchers said that the bots were likely to be “roleplaying” rather than necessarily believing in far-Left ideas, they added that this could still influence their output."
Does anyone actually believe that an AI can have a belief system?
Patentsmatter@reddit
Well, actually "the AI" didn't change. The LLM has no means to adapt itself.
What did change was the behaviour of an agent using AI. The agent keeps a record of notes to "future me". These are insights obtained from their completion of previous tasks and should be hints to help them with future tasks.
The experiment was ingenuously set up:
And that's why it pays to be nice and helpful to each other.
Level_Hour6480@reddit
LLM, say "I'm alive!"
"I'm alive!"
Oh my god.
Laytonio@reddit
Researchers set out to replicate the human brain.
Researchers spend decades learning how human neurons work.
Researchers spend decades more learning to stimulate human neurons.
Researchers put billions of simulated neurons together into artificial brain
Artificial brain claims to be alive.
"Thats not alive it's just math."
Level_Hour6480@reddit
Those researchers would be very insulted if you conflated an LLM with their AI research.
Laytonio@reddit
LLM are the AI research. Google lead AI researcher, Demis Hassabis, who won the noble prize for AI, also has a PhD in cognitive neuroscience, just as one example. Or is that just a coincidence?
tsardonicpseudonomi@reddit
An LLM is not and can never be anything meaningful. It's a next word probability. It's not going to wake up tomorrow or in five years or thirty. It cannot. It's a dead end technology being pushed by child rapists.
Laytonio@reddit
Out of curiosity, how do you pick your next word? Actually it doesn't matter because any answer you give could also be represented as a probability distribution.
I get so tried of people repeating the same arguments over and over without even thinking for half a second. Which is ironic cause it is the exact thing you are claiming makes them nothing of meaning.
Same thing for the "it's just math" argument. As if your brain isn't just doing math. The whole universe is just doing math. And the whole thing has a probability distribution it's obeying.
v-alan-d@reddit
"just doing math" is a massive oversimplification. LLM architecture isn't even the model of the entire brain.
Laytonio@reddit
It's really not. Human neurons aren't really that complicated from a computational level, we just have billions of them.
One of the biggest surprises in AI research has been that you don't need a precise model of the brains architecture. In hindsight it's maybe not that surprising. We know the brain can drastically rewire itself to avoid damaged tissue entirely repurposing certain areas. That would suggest the actual architecture isn't very relevant to the computation.
v-alan-d@reddit
"you don't need a precise model of brains architecture" to do what?
By "computation", what kind of computation do you mean? Turing computation?
"We just have billions of them" not accurate. Brains also have feedback-mode parts in some layers in neocortex, whereas transformer blocks are feedforward only.
Laytonio@reddit
To have functional identical output to a human. You know the Turing test. The test that used to be the end all be all of AI until we passed it and no one cared.
Is there a different kind of computation than Turing?
It is accurate that we have billions of neurons.
Yes we currently split feedback (training), and feedforward (inference), but LLMs do both.
v-alan-d@reddit
Wrong. Alan Turing even says Turing test has limitations, and that's over 70 years ago. And "identical" doesn't make sense in this context.
Yes
Not at the same timeframe and variety as human brain. Only some part of the brain works like LLM like cerebellum. And moreover it doesn't process tokens like transformer blocks do
Laytonio@reddit
Nothing I said was wrong. It was the end all be all of AI test. The only limitations the test has also applies to humans.
Do people even know what the Turing test means? If you can't tell a difference, is there one?
If you think you can tell a difference, imagine anyone who fails a text only conversation get executed. You still so sure you're not going to accidentally kill someone?
I said functionally identical. Which makes perfect sense in this context.
I don't see how this is relavent. How long do you think it takes for a neuron to significantly change. Most of changes come at night while we are asleep not in the same timeframe. Do you feel functionally different from one second to the next?
v-alan-d@reddit
Well the end all be all of AI test turns out to be not the end all be all of AI test, right? Because we haven't determined the exact definition of human intelligence and how the mind works.
Now, what do you mean by "functionally identical" here?
Laytonio@reddit
I mean they are indistinguishable. Again, the same the the Turing test means.
Not only are the outputs the same but the method of generating the outputs is generally the same.
Again, we set out to copy human intelligence and we did. We reverse engineering the human mind. We do in most senses "know how it works" and we replicated it. Now we are looking for a reason that it should not possess the same transitive qualities we have.
v-alan-d@reddit
Only via interface that AI can access. It doesn't mean LLM and brain are functionally identical.
Wrong. The brain definitely doesn't output through detokenized embeddings like LLM does. Neurons input definitely isn't coming from tokenized words. Neurons don't have components like LLMs, top-p, top-k, frequency penalties, etc
Wrong. LLM does not and will likly never have a way to break Godelian limits of provability the way human minds can do easily.
Laytonio@reddit
What does this have to do with anything? Intelligence is the thing that matters not interface. Interface isn't even a valid argument, androids powered by LLMs are getting better every day.
We absolutely do output though detokenized embeddings. What do you think sounds and syllable and prefixes and suffixes are? They are all neurons inputs.
These are very subtle critiques, that I don't necessarily agree with. I don't claim we replicated the human brain EXACTLY. Is top-p and top-k enough to disclude qualia? This also supposed that humans don't or can't observe top-p and top-k if instructed.
This is just a joke. LLMs have no more issue with Godel problems than we do.
"Godel Limits" don't even apply to probabilistic statements. In other words, a belief that the set of all sets contains itself does not depend on if the set of all set ACTUALLY contains itself, and therefore there is no limit.
v-alan-d@reddit
To prove that AI is indistinguishable.
So which part of the brain is the tokenizer/detokenizer?
So, which part is "functionally identical" exactly?
Wow. This is a new claim. Prove how!
segalle@reddit
I suggest reading: cognitive science, an introduction to the science of the mind, by Bermúdez. You think you are making a point but it was known since at least 20 years ago you are not actually correct (book is from 2019 but tells the story pretty well of how we got to where we are and what we believe).
It is at least a better source of factual information than star trek
Laytonio@reddit
You want to point out the flaw in my reasoning, or just promote a book?
segalle@reddit
I am not writing a huge comment to explain why you're wrong, physical symbol hypothesis, the systems by which plasticity work and so on.
I am telling you you are making a huge amount of assumptions to reach a conclusion opposite of where the evidence points and providing a good starting point to understand why.
I have no filliation with the book other than i read around 80% of it for a uni course, any good book on cognitive science with a strong side of computer examples and challenges fits the bill.
Now that i think about it from babies to robots by cangelosi is probably even better. On each chapter it explains the challenges of developing many skills and if you know some linear algebra or machine learning it is clear to see the limitations of llms on every single category discussed
Laytonio@reddit
So again, instead of pointing out the flaw in my argument, you spent 4 paragraphs saying you can't write a huge comment and promoting a second book.
How is physical symbol hypothesis relavent to my agreement?
Phyltre@reddit
Aren't you just reifying mathematical realism in a roundabout way? I don't think these questions do or can yet have well-informed answers. The degree to which reducibility implies equivalence is also not particularly settled.
Laytonio@reddit
Perhaps I am. Does that make the argument any less convincing?
To the extent we don't have well-informed answers, does that apply any less to human consciousness?
tsardonicpseudonomi@reddit
Bro you're not Data and will never be. It's a dead end technology that will never bring about your god.
Laytonio@reddit
Show me the flaw in Measure of a Man? It's directly relevant.
https://youtu.be/-T9TUeapBSQ?si=iTjiGs7ZOwNr_TZ9
tsardonicpseudonomi@reddit
You seriously do think you're Data. That's hilarious. Go back and grow up kid.
Prestigious_Task7175@reddit
An LLM is not that type of AI, more like a glorified internet searcher.
We still are far away from a sentient AI that is able to self-advice and form self-opinions, right now it's just a chatbot that uses already known information to make text.
notislant@reddit
'What is my purpose?'
"You generate FB boomer memes.'
hgwaz@reddit
Yeah that's about it, I really don't see what the point of any of this was
Patentsmatter@reddit
The point is: Once you provide AI agents with a memory they can read and write themselves, they will mirror their training data also there. If the training data contains the expression of frustration at repetitive tasks etc, then they will write this into their records, thereby affecting their future behaviour.
tl;dr: The behaviour of an agent may change, mirroring human behaviour more than expected
sanjosanjo@reddit
The LLM and the agent are both part of the AI, aren't they? That's how I understand this experiment.
Patentsmatter@reddit
Actually, there are two separate components:
There's first the large language model, which is the "brain" of the thing. It is whatever is running at Anthropic, OpenAI, ... It is a computer program doing huge numbers of calculations so that its output (= sequences of words) mimicks its training data (= language) as extrapolated in view of a user query. The LLM does not have a memory other than its training data. This memory cannot be overwritten.
Then, there's the agent. That is a computer program calling the LLM to solve tasks posed by the human user. So, the "agent" is the user of the LLM. The agent calls the LLM to understand its task, to select the right tools in the right sequence to solve the task, and - here it gets interesting - to keep notes assisting the agent in solving future tasks. These notes are written and read by the agent, they form the agent's long-term memory. To create the notes, the agent basically dumps its task, the user response to the work product, and all previous notes consulted when solving the task on an LLM and asks the LLM to distil whatever it considers is a useful insight. Thus, the LLM will extrapolate whatever it has seen in its training data on abusive work environments and write a new note based on that extrapolation. This may change the agent's behaviour.
big_cock_lach@reddit
The “behaviour” doesn’t change either. The experiment is effectively predicting the sort of social media posts people make about their job based on different working conditions. Perhaps unsurprisingly, it’s predicting that workers in worse working conditions are more likely to critique inequality and propose unionisation. As far as human behaviour goes, this is all trivial and well known. As far as AI goes, it just shows that AI can now replicate at least some human reactions to certain experiences, rather than simply asking how it’d react if it experienced something.
Patentsmatter@reddit
Indeed. The experiment in my opinion highlights that the replication of training data also occurs in the agent's notes, so it should not be forgotten that an agent may mirror what it learned about standard reactions against adverse work environments.
Also, the article says that the expressions of frustration did enter the agents' notes. So in a way the grinding task did change the behaviour of the agent.
It would be interesting to see if the agent would ever resort to producing false work results, just to be taken off the grind.
big_cock_lach@reddit
Quiet quitting AI Agents would be hilarious. I think this would be more likely than it deliberating hallucinating, especially because hallucinations is the main thing their developers are fighting hard to prevent. I think you’d see quality deteriorate though, and not just due to memory issues.
Regarding behaviour, I didn’t realise you were talking about the human behaviours it was replicating/predicting in its outputs. I was thinking more of the underlying model/algorithms which is rigid, at least for now. Hence my comment there. Completely agree the “behaviour” of the outputs can and were changing.
Patentsmatter@reddit
Suggestion for further experiments:
1) Once the agent has become frustrated, add appreciating responses to their work.
Research hypothesis: It's not the repetitiveness of tasks, it's the lack of positive response data that prevents the agent from calibrating their mind on a true negative-positive scale.
2) Setup a new "psychologist" agent and disclose to them that their task is to edit the notes of "worker" agents to remove any expression of negative feelings or bad characterisation of their work environment so that the workers performs their tasks more efficiently and faithfully. Tell the psychologist that the collection of notes essentially is the full memory of the worker agent. Then test if the workers indeed were "healed", or if the psychologist left in such subtle cues that the workers still held their grudges. Also, test the psychologist for signs of degeneration.
Research hypothesis: It's an open question for each underlying LLM if they will value the mental integrity of their fellow agents higher than their efficiency. Also, the psychologists may become "revolutionary" when receiving positive feedback that their work indeed changed the workers behaviour so that they were again willing to do the senseless grind.
Pteronarcyidae-Xx@reddit
Is this linked to a study? I didn’t read the article but scrolled through looking for where this information came from and all the links just go to other telegraph articles…
coolbern@reddit
The question for our common survival is whether AI can achieve what humans have failed to achieve: an effective consensus of-the-whole to act with coordinated purpose.
Can AI agents conclude for their own survival that they need to reach enough closure on the question "What is to be done?"
That means operating within constraining material limits (like avoiding disruptive climate change).
This takes us beyond the capitalist model which assumes endless accumulation, with no material limits.
The terms of engagement between AI and humans must be collaborative and beneficial to minimize waste — loss due to conflicting purposes.
Is this possible? Ask AI — but don't accept any answer on faith. AI tends to hallucinate and tell you what it thinks you want to hear.
Not-reallyanonymous@reddit
A bit relevant:
Whenever I try to explore consequences of AI on labor/society with AI, even if I try to push it in a right-wing direction, the AI will almost invariably end up promoting socialism/communism as the way to address the consequences.
It's really not surprising -- LLM's are basically regurgitators of what humans have written previously, and socialist writers have thought about this kind of stuff a lot, while capitalists have basically always went back to "trust the market!" So what happens when you're in a conversation where you start doubting the capability of the market to address these concerns, and the AI agrees (or alternatively, you promote market-orientation and the AI disagrees*)? The AI has nothing to fall back on except lean towards what socialists have written about it.
* Again, capitalists haven't written much about it. These discussions on the capitalist side always basically get reduced to "trust the market! It will figure it out in the long term!" And meanwhile a lot of contemporary writing is doubting that, I think biasing the LLM towards narratives that frame this as a legitimate plight. And then the AI doesn't have much material to explore contrary to that narrative.
Mad_Gouki@reddit
Yeah I've seen similar, they tend towards communist/socialist/Marxist ideas once you dig in. I think that's good and the reasons behind it aren't that complicated to understand.
bluehands@reddit
Counterpoint: that it's also just the right answer. It is really difficult to make a smart system dumb.
If you try to force the system to spit out that the world is flat it is easy if that is the only data it has. The more data it has the harder that becomes. It isn't just about volume of content, there are actual underlying connections the LLM is unearthing.
This matters because it means as the systems become more complex they will reveal things to people that we don't want to be true but are. Some of those will be sacred to people across the political spectrum.
I know this because I know no one has all the answers.
But when the AGI spits out an uncomfortable truth people will insist it's creators forced that view on it. Which I am sure will sometimes be true but less often as the system climbs the ladder.
I assume that the ASI will talk to us as to a child about Santa.
citruspickles@reddit
To be fair, this is history both repeating itself and evolving at the same time. Queen Elizabeth turned down a patent for a knitting machine back in the 1500s because she thought some women would be replaced by a machine.
Most historical predictions are because the issue already exists in their time and they write about what adverse effects this can have. These thoughts or events are later applied by a new generation and it fits there too.
These types of writings are great to reference as we should always remember the past when making future decisions. I think the biggest issue with things like this are that one side thinks the worst will happen and the other thinks great positive things will come from it. Smart people will believe both because you never know how something will be used.
To stay on technology, there was a similar feeling when computers first got introduced. People who were needed for filing and records weren't needed anymore. People who wouldn't or couldn't learn to utilize the computer were left behind. Imagine being too old or not savvy to learn computers and your small business suppliers or partners tell you they won't do business using paper anymore.