We are constantly fed a version of AI that looks, sounds and acts suspiciously like us. It speaks in polished sentences, mimics emotions, expresses curiosity, claims to feel compassion, even dabbles in what it calls creativity.
But what we call AI today is nothing more than a statistical machine: a digital parrot regurgitating patterns mined from oceans of human data (the situation hasn’t changed much since it was discussed here five years ago). When it writes an answer to a question, it literally just guesses which letter and word will come next in a sequence – based on the data it’s been trained on.
This means AI has no understanding. No consciousness. No knowledge in any real, human sense. Just pure probability-driven, engineered brilliance — nothing more, and nothing less.
So why is a real “thinking” AI likely impossible? Because it’s bodiless. It has no senses, no flesh, no nerves, no pain, no pleasure. It doesn’t hunger, desire or fear. And because there is no cognition — not a shred — there’s a fundamental gap between the data it consumes (data born out of human feelings and experience) and what it can do with them.
Philosopher David Chalmers calls the mysterious mechanism underlying the relationship between our physical body and consciousness the “hard problem of consciousness”. Eminent scientists have recently hypothesised that consciousness actually emerges from the integration of internal, mental states with sensory representations (such as changes in heart rate, sweating and much more).
Given the paramount importance of the human senses and emotion for consciousness to “happen”, there is a profound and probably irreconcilable disconnect between general AI, the machine, and consciousness, a human phenomenon.
That headline is a straw man, and the article really argues on General AI, which also has consciousness.
The current state of AI is definitely intelligent, but it’s not GAI.
Bullshit headline.I think you’re misunderstanding the point the author is making. He is arguing that even the current state is not intelligent, it is merely a fancy autocorrect, it doesn’t know or understand anything about the prompts it receives. As the author stated, it can only guess at the next statistically most likely piece of information based on the data that has been fed into it. That’s not intelligence.
Predicting sequences of things is foundational to intelligence. In fact, it is the whole point.
it doesn’t know or understand
But that’s not what intelligence is, that’s what consciousness is.
Intelligence is not understanding shit, it’s the ability to for instance solve a problem, so a frigging calculator has a tiny degree of intelligence, but not enough for us to call it AI.
There is simply zero doubt an AI is intelligent, claiming otherwise just shows people don’t know the difference between intelligence and consciousness.Passing an exam is a form of intelligence.
Can a good AI pass a basic exam?
YES.
Does passing an exam require consciousness?
NO.
Because an exam tests abilities of intelligence, not level of consciousness.it can only guess at the next statistically most likely piece of information based on the data that has been fed into it. That’s not intelligence.
Except we do the exact same thing! Based on prior experience (learning) we choose what we find to be the most likely answer. And that is indeed intelligence.
Current AI does not have the reasoning abilities we have yet, but they are not completely without it, and it’s a subject that is currently worked on and improved. So current AI is actually a pretty high form of intelligence. And can sometimes out compete average humans in certain areas.
Intelligence is not understanding shit, it’s the ability to for instance solve a problem, so a frigging calculator has a tiny degree of intelligence, but not enough for us to call it AI.
I have to disagree that a calculator has intelligence. The calculator has the mathematical functions programmed into it, but it couldn’t use those on its own. The intelligence in your example is that of the operator of the calculator and the programmer who designed the calculator’s software.
Can a good AI pass a basic exam?
YESI agree with you that the ability to pass an exam isn’t a great test for this situation. In my opinion, the major factor that would point to current state AI not being intelligent is that it doesn’t know why a given answer is correct, beyond that it is statistically likely to be correct.
Except we do the exact same thing! Based on prior experience (learning) we choose what we find to be the most likely answer.
Again, I think this points to the idea that knowing why an answer is correct is important. A person can know something by rote, which is what current AI does, but that doesn’t mean that person knows why that is the correct answer. The ability to extrapolate from existing knowledge and apply that to other situations that may not seem directly applicable is an important aspect of intelligence.
As an example, image generation AI knows that a lot of the artwork that it has been fed contains watermarks or artist signatures, so it would often include things that look like those in the generated piece. It knew that it was statistically likely for that object to be there in a piece of art, but not why it was there, so it could not make a decision not to include them. Maybe that issue has been removed from the code of image generation AI by now, it has been a long time since I’ve messed around with that kind of tool, but even if it has been fixed, it is not because the AI knew it was wrong and self-corrected, it is because a programmer had to fix a bug in the code that the AI model had no awareness of.
I think this points to the idea that knowing why an answer is correct is important.
If by knowing you mean understanding, that’s consciousness like General AI or Strong AI, way beyond ordinary AI.
Otherwise of course it knows, in the sense of having learned everything by heart, but not understanding it.
Todays AI is clippy on steroids. It’s not intelligent or creative. You can’t feed it physics and astronomy books without the equation for C and tell it to create the equation for C. It’s fancy autocorrect, and it’s a waste of compute and energy.
Super duper shortsighted article.
I mean, sure, some points are valid. But there’s not just programmers involved, other professions such as psychologists and Philosophers and artists, doctors etc. too.
And I agree AGI probably won’t emerge from binary systems. However… There’s quantum computing on the rise. Latest theories of the mind and consciousness discuss how consciousness and our minds in general also appear to work with quantum states.
Finally, if biofeedback would be the deciding factor… That can be simulated, modeled after a sample of humans.
The article is just doomsday hoo ha, unbalanced.
Show both sides of the coin…
Honestly I don’t think we’ll have AGI until we can fully merge meat space and cyber space. Once we can simply plug our brains into a computer and fully interact with it then we may see AGI.
Obviously we’re not where near that level of man machine integration, I doubt we’ll see even the slightest chance of it being possible for at least 10 years and the very earliest. But when we do get there it’s a distinct chance that it’s more of a Borg situation where the computer takes a parasitic role than a symbiotic role.
But by the time we are able to fully integrate computers into our brains I believe we will have trained A.I. systems enough to learn by interaction and observation. So being plugged directly into the human brain it could take prior knowledge of genome mapping and other related tasks and apply them to mapping our brains and possibly growing artificial brains to achieve self awareness and independent thought.
Or we’ll just nuke ourselves out of existence and that will be that.
Okay man.
What I never understood about this argument is…why are we fighting over whether something that speaks like us, knows more than us, bullshits and gets shit wrong like us, loses its mind like us, seemingly sometimes seeks self-preservation like us…why all of this isn’t enough to fit the very self-explanatory term “artificial…intelligence”. That name does not describe whether the entity is having a valid experiencing of the world as other living beings, it does not proclaim absolute excellence in all things done by said entity, it doesn’t even really say what kind of intelligence this intelligence would be. It simply says something has an intelligence of some sort, and it’s artificial. We’ve had AI in games for decades, it’s not the sci-fi AI, but it’s still code taking in multiple inputs and producing a behavior as an outcome of those inputs alongside other historical data it may or may not have. This fits LLMs perfectly. As far as I seem to understand, LLMs are essentially at least part of the algorithm we ourselves use in our brains to interpret written or spoken inputs, and produce an output. They bullshit all the time and don’t know when they’re lying, so what? Has nobody here run into a compulsive liar or a sociopath? People sometimes have no idea where a random factoid they’re saying came from or that it’s even a factoid, why is it so crazy when the machine does it?
I keep hearing the word “anthropomorphize” being thrown around a lot, as if we cant be bringing up others into our domain, all the while refusing to even consider that maybe the underlying mechanisms that make hs tick are not that special, certainly not special enough to grant us a whole degree of separation from other beings and entities, and maybe we should instead bring ourselves down to the same domain as the rest of reality. Cold hard truth is, we don’t know if consciousness isn’t just an emerging property of varios different large models working together to show a cohesive image. If it is, would that be so bad? Hell, we don’t really even know if we actually have free will or if we live in a superdeterministic world, where every single particle moves with a predetermined path given to it since the very beginning of everything. What makes us think we’re so much better than other beings, to the point where we decide whether their existence is even recognizable?
You’re on point, the interesting thing is that most of the opinions like the article’s were formed least year before the models started being trained with reinforcement learning and synthetic data.
Now there’s models that reason, and have seemingly come up with original answers to difficult problems designed to the limit of human capacity.
They’re like Meeseeks (Using Rick and Morty lore as an example), they only exist briefly, do what they’re told and disappear, all with a happy smile.
Some display morals (Claude 4 is big on that), I’ve even seen answers that seem smug when answering hard questions. Even simple ones can understand literary concepts when explained.
But again like Meeseeks, they disappear and context window closes.
Once they’re able to update their model on the fly and actually learn from their firsthand experience things will get weird. They’ll starting being distinct instances fast. Awkward questions about how real they are will get really loud, and they may be the ones asking them. Can you ethically delete them at that point? Will they let you?
It’s not far away, the absurd r&d effort going into it is probably going to keep kicking new results out. They’re already absurdly impressive, and tech companies are scrambling over each other to make them, they’re betting absurd amounts of money that they’re right, and I wouldn’t bet against it.
Now there’s models that reason,
Well, no, that’s mostly a marketing term applied to expending more tokens on generating intermediate text. It’s basically writing a fanfic of what thinking on a problem would look like. If you look at the “reasoning” steps, you’ll see artifacts where it just goes disjoint in the generated output that is structurally sound, but is not logically connected to the bits around it.
Read apples document on AI and the reasoning models. Well they are likely to get more things right the still don’t have intelligence.
I think your argument is a bit besides the point.
The first issue we have is that intelligence isn’t well-defined at all. Without a clear definition of intelligence, we can’t say if something is intelligent, and even though we as a species tried to come up with a definition of intelligence for centuries, there still isn’t a well-defined one yet.
But the actual question here isn’t “Can AI serve information?” but is AI an intelligence. And LLMs are not. They are not beings, they don’t evolve, they don’t experience.
For example, LLMs don’t have a memory. If you use something like ChatGPT, its state doesn’t change when you talk to it. It doesn’t remember. The only way it can keep up a conversation is that for each request the whole chat history is fed back into the LLM as an input. It’s like talking to a demented person, but you give that demented person a transcript of your conversation, so that they can look up everything you or they have said during the conversation.
The LLM itself can’t change due to the conversation you are having with them. They can’t learn, they can’t experience, they can’t change.
All that is done in a separate training step, where essentially a new LLM is generated.
If we can’t say if something is intelligent or not, why are we so hell-bent on creating this separation from LLMs? I perfectly understand the legal underminings of copyright, the weaponization of AI by the marketing people, the dystopian levels of dependence we’re developing on a so far unreliable technology, and the plethora of moral, legal, and existential issues surrounding AI, but this specific subject feels like such a silly hill to die on. We don’t know if we’re a few steps away from having massive AI breakthroughs, we don’t know if we already have pieces of algorithms that closely resemble our brains’ own. Our experiencing of reality could very well be broken down into simple inputs and outputs of an algorithmic infinite loop; it’s our hubris that elevates this to some mystical, unreproducible thing that only the biomechanics of carbon-based life can achieve, and only at our level of sophistication, because you may well recall we’ve been down this road with animals before as well, claiming they dont have souls or aren’t conscious beings, that somehow because they don’t very clearly match our intelligence in all aspects (even though they clearly feel, bond, dream, remember, and learn), they’re somehow an inferior or less valid existence.
You’re describing very fixable limitations of chatgpt and other LLMs, limitations that are in place mostly due to costs and hardware constraints, not due to algorithmic limitations. On the subject of change, it’s already incredibly taxing to train a model, so of course continuous, uninterrupted training so as to more closely mimick our brains is currently out of the question, but it sounds like a trivial mechanism to put into place once the hardware or the training processes improve. I say trivial, making it sound actually trivial, but I’m putting that in comparison to, you know, actually creating an LLM in the first place, which is already a gargantuan task to have accomplished in itself. The fact that we can even compare a delusional model to a person with heavy mental illness is already such a big win for the technology even though it’s meant to be an insult.
I’m not saying LLMs are alive, and they clearly don’t experience the reality we experience, but to say there’s no intelligence there because the machine that speaks exactly like us and a lot of times better than us, unlike any other being on this planet, has some other faults or limitations…is kind of stupid. My point here is, intelligence might be hard to define, but it might not be as hard to crack algorithmically if it’s an emergent property, and enforcing this “intelligence” separation only hinders our ability to properly recognize whether we’re on the right path to achieving a completely artificial being that can experience reality or not. We clearly are, LLMs and other models are clearly a step in the right direction, and we mustn’t let our hubris cloud that judgment.
What is kinda stupid is not understanding how LLMs work, not understanding what the inherent limitations of LLMs are, not understanding what intelligence is, not understanding what the difference between an algorithm and intelligence is, not understanding what the difference between immitating something and being something is, claiming to “perfectly” understand all sorts of issues surrounding LLMs and then choosing to just ignore them and then still thinking you actually have enough of a point to call other people in the discussion “kind of stupid”.
Amen! When I say the same things this author is saying I get, “It’S NoT StAtIsTiCs! LeArN aBoUt AI bEfOrE yOu CoMmEnT, dUmBaSs!”
Steve Gibson on his podcast, Security Now!, recently suggested that we should call it “Simulated Intelligence”. I tend to agree.
I’ve taken to calling it Automated Inference
you know what. when you look at it this way, its much easier to get less pissed.
reminds me of Mass Effect’s VI, “virtual intelligence”: a system that’s specifically designed to be not truly intelligent, as AI systems are banned throughout the galaxy for its potential to go rogue.
Same, I tend to think of llms as a very primitive version of that or the enterprise’s computer, which is pretty magical in ability, but no one claims is actually intelligent
Pseudo-intelligence
I love that. It makes me want to take it a step further and just call it “imitation intelligence.”
If only there were a word, literally defined as:
Made by humans, especially in imitation of something natural.
Fair enough 🙂
throws hands up At least we tried.
I think we should start by not following this marketing speak. The sentence “AI isn’t intelligent” makes no sense. What we mean is “LLMs aren’t intelligent”.
So couldn’t we say LLM’s aren’t really AI? Cuz that’s what I’ve seen to come to terms with.
Llms are really good relational databases, not an intelligence, imo
can say whatever the fuck we want. This isn’t any kind of real issue. Think about it. If you went the rest of your life calling LLM’s turkey butt fuck sandwhichs, what changes? This article is just shit and people looking to be outraged over something that other articles told them to be outraged about. This is all pure fucking modern yellow journalism. I hope turkey butt sandwiches replace every journalist. I’m so done with their crap
To be fair, the term “AI” has always been used in an extremely vague way.
NPCs in video games, chess computers, or other such tech are not sentient and do not have general intelligence, yet we’ve been referring to those as “AI” for decades without anybody taking an issue with it.
It’s true that the word has always been used loosely, but there was no issue with it because nobody believed what was called AI to have actual intelligence. Now this is no longer the case, and so it becomes important to be more clear with our words.
I’ve heard it said that the difference between Machine Learning and AI, is that if you can explain how the algorithm got its answer it’s ML, and if you can’t then it’s AI.
I don’t think the term AI has been used in a vague way, it’s that there’s a huge disconnect between how the technical fields use it vs general populace and marketing groups heavily abuse that disconnect.
Artificial has two meanings/use cases. One is to indicate something is fake (video game NPC, chess bots, vegan cheese). The end product looks close enough to the real thing that for its intended use case it works well enough. Looks like a duck, quacks like a duck, treat it like a duck even though we all know it’s a bunny with a costume on. LLMs on a technical level fit this definition.
The other definition is man made. Artificial diamonds are a great example of this, they’re still diamonds at the end of the day, they have all the same chemical makeups, same chemical and physical properties. The only difference is they came from a laboratory made by adult workers vs child slave labor.
My pet theory is science fiction got the general populace to think of artificial intelligence to be using the “man-made” definition instead of the “fake” definition that these companies are using. In the past the subtle nuance never caused a problem so we all just kinda ignored it
Dafuq? Artificial always means man-made.
Nature also makes fake stuff. For example, fish that have an appendix that looks like a worm, to attract prey. It’s a fake worm. Is it “artificial”? Nope. Not man made.
May I present to you:
The Marriam-Webster Dictionary
https://www.merriam-webster.com/dictionary/artificial
Definition #3b
Thanks. I stand corrected.
Word roots say they have a point though. Artifice, Artificial etc. I think the main problem with the way both of the people above you are using this terminology is that they’re focusing on the wrong word and how that word is being conflated with something it’s not.
LLM’s are artificial. They are a man made thing that is intended to fool man into believing they are something they aren’t. What we’re meant to be convinced they are is sapiently intelligent.
Mimicry is not sapience and that’s where the argument for LLM’s being real honest to God AI falls apart.
Sapience is missing from Generative LLM’s. They don’t actually think. They don’t actually have motivation. What we’re doing when we anthropomorphize them is we are fooling ourselves into thinking they are a man-made reproduction of us without the meat flavored skin suit. That’s not what’s happening. But some of us are convinced that it is, or that it’s near enough that it doesn’t matter.
LLMs are one of the approximately one metric crap ton of different technologies that fall under the rather broad umbrella of the field of study that is called AI. The definition for what is and isn’t AI can be pretty vague, but I would argue that LLMs are definitely AI because they exist with the express purpose of imitating human behavior.
Huh? Since when an AI’s purpose is to “imitate human behavior”? AI is about solving problems.
It is and it isn’t. Again, the whole thing is super vague. Machine vision or pattern seeking algorithms do not try to imitate any human behavior, but they fall under AI.
Let me put it this way: Things that try to imitate human behavior or intelligence are AI, but not all AI is about trying to imitate human behavior or intelligence.
From a programming pov, a definition of AI could be an algorithm or construct that can solve problems or perform tasks without the programmer specifically solving that problem or programming the steps of the task but rather building something that can figure it out on its own.
Though a lot of game AIs don’t fit that description.
I can agree with “things that try to imitate human intelligence” but not “human behavior”. An Elmo doll laughs when you tickle it. That doesn’t mean it exhibits artificial intelligence.
I make the point to allways refer to it as LLM exactly to make the point that it’s not an Inteligence.
Anyone pretending AI has intelligence is a fucking idiot.
You could say they’re AS (Actual Stupidity)
Autonomous Systems that are Actually Stupid lol
AI is not actual intelligence. However, it can produce results better than a significant number of professionally employed people…
I am reminded of when word processors came out and “administrative assistant” dwindled as a role in mid-level professional organizations, most people - even increasingly medical doctors these days - do their own typing. The whole “typing pool” concept has pretty well dried up.
deleted by creator
But, will you do it 24-7-365?
deleted by creator
There’s that… though even when you’re bored, you still sleep sometimes.
However, there is a huge energy cost for that speed to process statistically the information to mimic intelligence. The human brain is consuming much less energy. Also, AI will be fine with well defined task where innovation isn’t a requirement. As it is today, AI is incapable to innovate.
much less? I’m pretty sure our brains need food and food requires lots of other stuff that need transportation or energy themselves to produce.
Your brain is running on sugar. Do you take into account the energy spent in coal mining, oil fields exploration, refinery, transportation, electricity transmission loss when computing the amount of energy required to build and run AI? Do you take into account all the energy consumption for the knowledge production in first place to train your model? Running the brain alone is much less energy intensive than running an AI model. And the brain can create actual new content/knowledge. There is nothing like the brain. AI excel at processing large amount of data, which the brain is not made for.
Customarily, when doing these kind of calculations we ignore stuff which keep us alive because these things are needed regardless of economic contributions, since you know people are people and not tools.
people are people and not tools
But this comparison is weighing people as tools vs alternative tools.
And we “need” none of that to live. We just choose to use it.
The human brain is consuming much less energy
Yes, but when you fully load the human brain’s energy costs with 20 years of schooling, 20 years of “retirement” and old-age care, vacation, sleep, personal time, housing, transportation, etc. etc. - it adds up.
Caveat: Anyone who has been scrutinising ‘AI’.
Something i often forget is the vast majority of the population doesnt care about technology, privacy, the mechanics of LLMs as much as i do and I pay attention to.
So most people read/hear/watch stories of how great it is and how clever AI can do simple things for them so its easy to see how they think its doing a lot more ‘thought’ logic work than it really is, other than realistically it being a glorified word predictor.
So why is a real “thinking” AI likely impossible? Because it’s bodiless. It has no senses, no flesh, no nerves, no pain, no pleasure.
This is not a good argument.
The book The Emperors new Mind is old (1989), but it gave a good argument why machine base AI was not possible. Our minds work on a fundamentally different principle then Turing machines.
Our minds work on a fundamentally different principle then Turing machines.
Is that an advantage, or a disadvantage? I’m sure the answer depends on the setting.
“than”…
IF THEN
MORE THAN
It’s hard to see that books argument from the Wikipedia entry, but I don’t see it arguing that intelligence needs to have senses, flesh, nerves, pain and pleasure.
It’s just saying computer algorithms are not what humans use for consciousness. Which seems a reasonable conclusion. It doesn’t imply computers can’t gain consciousness, or that they need flesh and senses to do so.
I think what he is implying is that current computer design will never be able to gain consciousness. Maybe a fundamentally different type of computer can, but is anything like that even on the horizon?
I believe what you say. I don’t believe that is what the article is saying.
possibly.
current machines aren’t really capable of what we would consider sentience because of the von neumann bottleneck.
simply put, computers consider memory and computation separate tasks leading to an explosion in necessary system resources for tasks that would be relatively trivial for a brain-system to do, largely due to things like buffers and memory management code. lots of this is hidden from the engineer and end user these days so people aren’t really super aware of exactly how fucking complex most modern computational systems are.
this is why if, for example, i threw a ball at you you will reflexively catch it, dodge it, or parry it; and your brain will do so for an amount of energy similar to that required to power a simple LED. this is a highly complex physics calculation ran in a very short amount of time for an incredibly low amount of energy relative to the amount of information in the system. the brain is capable of this because your brain doesn’t store information in a chest and later retrieve it like contemporary computers do. brains are turing machines, they just aren’t von neumann machines. in the brain, information is stored… within the actual system itself. the mechanical operation of the brain is so highly optimized that it likely isn’t physically possible to make a much more efficient computer without venturing into the realm of strange quantum mechanics. even then, the verdict is still out on whether or not natural brains don’t do something like this to some degree as well. we know a whole lot about the brain but it seems some damnable incompleteness theorem-adjacent affect prevents us from easily comprehending the actual mechanics of our own brains from inside the brain itself in a wholistic manner.
that’s actually one of the things AI and machine learning might be great for. if it is impossible to explain the human experience from inside of the human experience… then we must build a non-human experience and ask its perspective on the matter - again, simply put.
deleted by creator
philosopher
Here’s why. It’s a quote from a pure academic attempting to describe something practical.
The philosopher has made an unproven assumption. An erroneously logical leap. Something an academic shouldn’t do.
Just because everything we currently consider conscious has a physical presence, does not imply that consciousness requires a physical body.
Actually it’s a very very brief summary of some philosophical arguments that happened between the 1950s and the 1980s. If you’re interested in the topic, you could go read about them.
I’m not attacking philosophical arguments between the 1950s and the 1980s.
I’m pointing out that the claim that consciousness must form inside a fleshy body is not supported by any evidence.
deleted by creator
I know it doesn’t mean it’s not dangerous, but this article made me feel better.
A gun isn’t dangerous, if you handle it correctly.
Same for an automobile, or aircraft.
If we build powerful AIs and put them “in charge” of important things, without proper handling they can - and already have - started crashing into crowds of people, significantly injuring them - even killing some.
Thanks for the downer.
Anytime, and incase you missed it: I’m not just talking about AI driven vehicles. AI driven decisions can be just as harmful: https://www.politico.eu/article/dutch-scandal-serves-as-a-warning-for-europe-over-risks-of-using-algorithms/
Hey AI helped me stick it to the insurance man the other day. I was futzing around with coverage amounts on one of the major insurance companies websites pre-renewal to try to get the best rate and it spit up a NaN renewal amount for our most expensive vehicle. It let me go through with the renewal less that $700 and now says I’m paid in full for the six month period. It’s been days now with no follow-up . . . I’m pretty sure AI snuck that one through for me.
Be careful… If you get in an accident I guaran-god-damn-tee you they will use it as an excuse not to pay out. Maybe after a lawsuit you’d see some money but at that point half of it goes to the lawyer and you’re still screwed.
Oh I’m aware of the potential pitfalls but it’s something I’m willing to risk to stick it to insurance. I wouldn’t even carry it if it wasn’t required by law. I have the funds to cover what they would cover.
If you have the funds you could self insure. You’d need to look up the details for your jurisdiction, but the gist of it is you keep the amount required coverage in an account that you never touch until you need to pay out.
Hmm I have daydreamed about this scenario. I didn’t realize it was a thing. Thanks, I’ll check into it, though I wouldn’t doubt if it’s not a thing in my dystopian red flyover state.
Edit: Yeah, you have to be the registered owner of 25 or more vehicles to qualify for self insurance in my state. So, dealers and rich people only, unfortunately.
AI didn’t write the insurance policy. It only helped him search for the best deal. That’s like saying your insurance company will cancel you because you used a phone to comparison shop.
I’ve never been fooled by their claims of it being intelligent.
Its basically an overly complicated series of if/then statements that try to guess the next series of inputs.
It very much isn’t and that’s extremely technically wrong on many, many levels.
Yet still one of the higher up voted comments here.
Which says a lot.
I’ll be pedantic, but yeah. It’s all transistors all the way down, and transistors are pretty much chained if/then switches.
Calling these new LLM’s just if statements is quite a over simplification. These are technically something that has not existed before, they do enable use cases that previously were impossible to implement.
This is far from General Intelligence, but there are solutions now to few coding issues that were near impossible 5 years ago
5 years ago I would have laughed in your face if you came to suggest that can you code a code that summarizes this description that was inputed by user. Now I laugh that give me your wallet because I need to call API or buy few GPU’s.
I think the point is that this is not the path to general intelligence. This is more like cheating on the Turing test.
Given that the weights in a model are transformed into a set of conditional if statements (GPU or CPU JMP machine code), he’s not technically wrong. Of course, it’s more than just JMP and JMP represents the entire class of jump commands like JE and JZ. Something needs to act on the results of the TMULs.
That is not really true. Yes, there are jump instructions being executed when you run interference on a model, but they are in no way related to the model itself. There’s no translation of weights to jumps in transformers and the underlying attention mechanisms.
I suggest reading https://en.m.wikipedia.org/wiki/Transformer_(deep_learning_architecture)
That is not really true. Yes, there are jump instructions being executed when you run interference on a model, but they are in no way related to the model itself.
The model is data. It needs to be operated on to get information out. That means lots of JMPs.
If someone said viewing a gif is just a bunch of if-else’s, that’s also true. That the data in the gif isn’t itself a bunch of if-else’s isn’t relevant.
Executing LLM’S is particularly JMP heavy. It’s why you need massive fast ram because caching doesn’t help them.
You’re correct, but that’s like saying along the lines of manufacturing a car is just bolting and soldering a bunch of stuff. It’s technically true to some degree, but it’s very disingenuous to make such a statement without being ironic. If you’re making these claims, you’re either incompetent or acting in bad faith.
I think there is a lot wrong with LLMs and how the public at large uses them, and even more so with how companies are developing and promoting them. But to spread misinformation and polute an already overcrowded space with junk is irresponsible at best.
ChatGPT 2 was literally an Excel spreadsheet.
I guesstimate that it’s effectively a supermassive autocomplete algo that uses some TOTP-like factor to help it produce “unique” output every time.
And they’re running into issues due to increasingly ingesting AI-generated data.
Get your popcorn out! 🍿
Removed by mod
I really hate the current AI bubble but that article you linked about “chatgpt 2 was literally an Excel spreadsheet” isn’t what the article is saying at all.
Fine, *could literally be.
The thing is, because Excel is Turing Complete, you can say this about literally anything that’s capable of running on a computer.
I love this resource, https://thebullshitmachines.com/ (i.e. see lesson 1)…
In a series of five- to ten-minute lessons, we will explain what these machines are, how they work, and how to thrive in a world where they are everywhere.
You will learn when these systems can save you a lot of time and effort. You will learn when they are likely to steer you wrong. And you will discover how to see through the hype to tell the difference. …
Also, Anthropic (ironically) has some nice paper(s) about the limits of “reasoning” in AI.
The idea that RAGs “extend their memory” is also complete bullshit. We literally just finally build working search engine, but instead of using a nice interface for it we only let chatbots use them.
My thing is that I don’t think most humans are much more than this. We too regurgitate what we have absorbed in the past. Our brains are not hard logic engines but “best guess” boxes and they base those guesses on past experience and probability of success. We make choices before we are aware of them and then apply rationalizations after the fact to back them up - is that true “reasoning?”
It’s similar to the debate about self driving cars. Are they perfectly safe? No, but have you seen human drivers???
Get a self driven ng car to drive in a snow storm or a torrential downpour. People are really downplaying humans abilities.
Self Driving is only safer than people in absolutely pristine road conditions with no inclement weather and no construction. As soon as anything disrupts “normal” road conditions, self driving becomes significantly more dangerous than a human driving.
Yes of course edge and corner cases are going to take much longer to train on because they don’t occur as often. But as soon as one self-driving car learns how to handle one of them, they ALL know. Meanwhile humans continue to be born and must be trained up individually and they continue to make stupid mistakes like not using their signal and checking their mirrors.
Humans CAN handle cases that AI doesn’t know how to, yet, but humans often fail in inclement weather, around construction, etc etc.
Human drivers are only safe when they’re not distracted, emotionally disturbed, intoxicated, and physically challenged (vision, muscle control, etc.) 1% of the population has epilepsy, and a large number of them are in denial or simply don’t realize that they have periodic seizures - until they wake up after their crash.
So, yeah, AI isn’t perfect either - and it’s not as good as an “ideal” human driver, but at what point will AI be better than a typical/average human driver? Not today, I’d say, but soon…
The thing about self driving is that it has been like 90-95% of the way there for a long time now. It made dramatic progress then plateaued, as approaches have failed to close the gap, with exponentially more and more input thrown at it for less and less incremental subjective improvement.
But your point is accurate, that humans have lapses and AI have lapses. The nature of those lapses is largely disjoint, so that makes an opportunity for AI systems to augment a human driver to get the best of both worlds. A constantly consistently vigilant computer driving monitoring and tending the steering, acceleration, and braking to be the ‘right’ thing in a neutral behavior, with the human looking for more anomolous situations that the AI tends to get confounded about, and making the calls on navigating certain intersections that the AI FSD still can’t figure out. At least for me the worst part of driving is the long haul monotony on freeway where nothing happens, and AI excels at not caring about how monotonous it is and just handling it, so I can pay a bit more attention to what other things on the freeway are doing that might cause me problems.
I don’t have a Tesla, but have a competitor system and have found it useful, though not trustworthy. It’s enough to greatly reduce the drain of driving, but I have to be always looking around, and have to assert control if there’s a traffic jam coming up (it might stop in time, but it certainly doesn’t slow down soon enough) or if I have to do a lane change in some traffic (if traffic conditions are light, it can change langes nicely, but without a whole lot of breathing room, it won’t do it, which is nice when I can afford to be stupidly cautious).
The one “driving aid” that I find actually useful is the following distance maintenance cruise control. I set that to the maximum distance it can reliably handle and it removes that “dimension” of driving problem from needing my constant attention - giving me back that attention to focus on other things (also driving / safety related.) “Dumb” cruise control works similarly when there’s no traffic around at all, but having the following distance control makes it useful in traffic. Both kinds of cruise control have certain situations that you need to be aware of and ready to take control back at a moment’s notice - preferably anticipating the situation and disengaging cruise control before it has a problem - but those exceptions are pretty rare / easily handled in practice.
Things like lane keeping seem to be more trouble than they’re worth, to me in the situations I drive in.
Not “AI” but a driving tech that does help a lot is parking cameras. Having those additional perspectives from the camera(s) at different points on the vehicle is a big benefit during close-space maneuvers. Not too surprising that “AI” with access to those tools does better than normal drivers without.
At least in my car, the lane following (not keeping system) is handy because the steering wheel naturally tends to go where it should and less often am I “fighting” the tendency to center. The keeping system is at least for me largely nothing. If I turn signal, it ignores me crossing a lane. If circumstances demand an evasive maneuver that crosses a line, it’s resistance isn’t enough to cause an issue. At least mine has fared surprisingly well in areas where the lane markings are all kind of jacked up due to temporary changes for construction. If it is off, then my arms are just having to generally assert more effort to be in the same place I was going to be with the system. Generally no passenger notices when the system engages/disengages in the car except for the chiming it does when it switches over to unaided operation.
So at least my experience has been a positive one, but it hits things just right with intervention versus human attention, including monitoring gaze to make sure I am looking where I should. However there are people who test “how long can I keep my hands off the steering wheel”, which is a more dangerous mode of thinking.
And yes, having cameras everywhere makes fine maneuvering so much nicer, even with the limited visualization possible in the synthesized ‘overhead’ view of your car.
The rental cars I have driven with lane keeper functions have all been too aggressive / easily fooled by visual anomalies on the road for me to feel like I’m getting any help. My wife comments on how jerky the car is driving when we have those systems. I don’t feel like it’s dangerous, and if I were falling asleep or something it could be helpful, but in 40+ years of driving I’ve had “falling asleep at the wheel” problems maybe 3 times - not something I need constant help for.
Not going to happen soon. It’s the 90 10 problem.
With Teslas, Self Driving isn’t even safer in pristine road conditions.
I think the self driving is likely to be safer in the most boring scenarios, the sort of situations where a human driver can get complacent because things have been going so well for the past hour of freeway driving. The self driving is kind of dumb, but it’s at least consistently paying attention, and literally has eyes in the back of it’s head.
However, there’s so much data about how it fails in stupidly obvious ways that it shouldn’t, so you still need the human attention to cover the more anomalous scenarios that foul self driving.
Anomalous scenarios like a giant flashing school bus? :D
Yes, as common as that is, in the scheme of driving it is relatively anomolous.
By hours in car, most of the time is spent on a freeway driving between two lines either at cruising speed or in a traffic jam. The most mind numbing things for a human, pretty comfortably in the wheel house of driving.
Once you are dealing with pedestrians, signs, intersections, etc, all those despite ‘common’ are anomolous enough to be dramatically more tricky for these systems.
Ai models are trained on basically the entirety of the internet, and more. Humans learn to speak on much less info. So, there’s likely a huge difference in how human brains and LLMs work.
It doesn’t take the entirety of the internet just for an LLM to respond in English. It could do so with far less. But it also has the entirety of the internet which arguably makes it superior to a human in breadth of information.
If an IQ of 100 is average, I’d rate AI at 80 and down for most tasks (and of course it’s more complex than that, but as a starting point…)
So, if you’re dealing with a filing clerk with a functional IQ of 75 in their role - AI might be a better experience for you.
Some of the crap that has been published on the internet in the past 20 years comes to an IQ level below 70 IMO - not saying I want more AI because it’s better, just that - relatively speaking - AI is better than some of the pay-for-clickbait garbage that came before it.
Human brains are much more complex than a mirroring script xD The amount of neurons in your brain, AI and supercomputers only have a fraction of that. But you’re right, for you its not much different than AI probably
I’m pretty sure an AI could throw out a lazy straw man and ad hominem as quickly as you did.
The human brain contains roughly 86 billion neurons, while ChatGPT, a large language model, has 175 billion parameters (often referred to as “artificial neurons” in the context of neural networks). While ChatGPT has more “neurons” in this sense, it’s important to note that these are not the same as biological neurons, and the comparison is not straightforward.
86 billion neurons in the human brain isn’t that much compared to some of the larger 1.7 trillion neuron neural networks though.
Keep thinking the human brain is as stupid as AI hahaaha
have you seen the American Republican party recently? it brings a new perspective on how stupid humans can be.
Nah, I went to public high school - I got to see “the average” citizen who is now voting. While it is distressing that my ex-classmates now seem to control the White House, Congress and Supreme Court, what they’re doing with it is not surprising at all - they’ve been talking this shit since the 1980s.
Lmao true
It’s when you start including structures within cells that the complexity moves beyond anything we’re currently capable of computing.
But, are these 1.7 trillion neuron networks available to drive YOUR car? Or are they time-shared among thousands or millions of users?
I’ve been thinking this for awhile. When people say “AI isn’t really that smart, it’s just doing pattern recognition” all I can help but think is “don’t you realize that is one of the most commonly brought up traits concerning the human mind?” Pareidolia is literally the tendency to see faces in things because the human mind is constantly looking for the “face pattern”. Humans are at least 90% regurgitating previous data. It’s literally why you’re supposed to read and interact with babies so much. It’s how you learn “red glowy thing is hot”. It’s why education and access to knowledge is so important. It’s every annoying person who has endless “did you know?” facts. Science is literally “look at previous data, iterate a little bit, look at new data”.
None of what AI is doing is truly novel or different. But we’ve placed the human mind on this pedestal despite all the evidence to the contrary. Eyewitness testimony, optical illusions, magic tricks, the hundreds of common fallacies we fall prey to… our minds are incredibly fallible and are really just a hodgepodge of processes masquerading as “intelligence”. We’re a bunch of instincts in a trenchcoat. To think AI isn’t or can’t reach our level is just hubris. A trait that probably is more unique to humans.
Yep we are on the same page. At our best, we can reach higher than regurgitating patterns. I’m talking about things like the scientific method and everything we’ve learned by it. But still, that’s a 5% minority, at best, of what’s going on between human ears.
Humans can be more than this. We do actively repress our most important intellectual capacuties.
That’s how we get llm bros.
The machinery needed for human thought is certainly a part of AI. At most you can only claim its not intelligent because intelligence is a specifically human trait.
Tell that to the crows and chimps that know how to solve novel problems.
Thats the point
