I think AI is neat.
As someone who has loves Asimov and read nearly all of his work.
I absolutely bloody hate calling LLM’s AI, without a doubt they are neat. But they are absolutely nothing in the ballpark of AI, and that’s okay! They weren’t trying to make a synethic brain, it’s just the culture narrative I am most annoyed at.
I look at all these kids glued to their phones and I ask 'Where’s the Frankenstein Complex now that we really need it?"
Ok, but so do most humans? So few people actually have true understanding in topics. They parrot the parroting that they have been told throughout their lives. This only gets worse as you move into more technical topics. Ask someone why it is cold in winter and you will be lucky if they say it is because the days are shorter than in summer. That is the most rudimentary “correct” way to answer that question and it is still an incorrect parroting of something they have been told.
Ask yourself, what do you actually understand? How many topics could you be asked “why?” on repeatedly and actually be able to answer more than 4 or 5 times. I know I have a few. I also know what I am not able to do that with.
I don’t think actual parroting is the problem. The problem is they don’t understand a word outside of how it is organized. They can’t be told to do simple logic because they don’t have a simple understanding of each word in their vocabulary. They can only reorganize things to varying degrees.
https://en.m.wikipedia.org/wiki/Chinese_room
I think they’re wrong, as it happens, but that’s the argument.
I guess, I just am looking at from an end user vantage point. I’m not saying the model cant understand the words its using. I just don’t think it currently understands that specific words refer to real life objects and there are laws of physics that apply to those specific objects and how they interact with each other.
Like saying there is a guy that exists and is a historical figure means that information is independently verified by physical objects that exist in the world.
In some ways, you are correct. It is coming though. The psychological/neurological word you are searching for is “conceptualization”. The AI models lack the ability to abstract the text they know into the abstract ideas of the objects, at least in the same way humans do. Technically the ability to say “show me a chair” and it returns images of a chair, then following up with “show me things related to the last thing you showed me” and it shows couches, butts, tables, etc. is a conceptual abstraction of a sort. The issue comes when you ask “why are those things related to the first thing?” It is coming, but it will be a little while before it is able to describe the abstraction it just did, but it is capable of the first stage at least.
Some systems clearly do that though or are you just talking about llms?
Just llms
It’s like saying bro, this mouse can’t even type text if I don’t use an on screen keyboard
It doesn’t need to understand the words to perform logic because the logic was already performed by humans who encoded their knowledge into words. It’s not reasoning, but the reasoning was already done by humans. It’s not perfect of course since it’s still based on probability, but the fact that it can pull the correct sequence of words to exhibit logic is incredibly powerful. The main hard part of working with LLMs is that they break randomly, so harnessing their power will be a matter of programming in multiple levels of safe guards.
I feel that knowing what you don’t know is the key here.
An LLM doesn’t know what it doesn’t know, and that’s where what it spouts can be dangerous.
Of course there’s a lot of actual people that applies to as well. And sadly they’re often in positions of power.
There are more than a couple research agents in development
We need something that can real time fact check without error that would fuck twitter up lol
This is only one type of intelligence and LLMs are already better at humans at regurgitating facts. But I think people really underestimate how smart the average human is. We are incredible problem solvers, and AI can’t even match us in something as simple as driving a car.
Lol @ driving a car being simple. That is one of the more complex sensory somatic tasks that humans do. You have to calculate the rate of all vehicles in front of you, assess for collision probabilities, monitor for non-vehicle obstructions (like people, animals, etc.), adjust the accelerator to maintain your own velocity while terrain changes, be alert to any functional changes in your vehicle and be ready to adapt to them, maintain a running inventory of laws which apply to you at the given time and be sure to follow them. Hell, that is not even an exhaustive list for a sunny day under the best conditions. Driving is fucking complicated. We have all just formed strong and deeply connected pathways in our somatosensory and motor cortexes to automate most of the tasks. You might say it is a very well-trained neural network with hundreds to thousands of hours spent refining and perfecting the responses.
The issue that AI has right now is that we are only running 1 to 3 sub-AIs to optimize and calculate results. Once that number goes up, they will be capable of a lot more. For instance: one AI for finding similarities, one for categorizing them, one for mapping them into a use case hierarchy to determine when certain use cases apply, one to analyze structure, one to apply human kineodynamics to the structure and a final one to analyze for effectiveness of the kineodynamic use cases when done by a human. This would be a structure that could be presented an object and told that humans use it and the AI brain could be able to piece together possible uses for the tool and describe them back to the presenter with instructions on how to do so.
AI can beat me in driving a car, and I have a degree.
Jokes on them. I don’t even calculate when I need to parrot. I am beyond such lowly needs.
Few people truly understand what understanding means at all, i got teacher in college that seriously thinked that you should not understand content of lessons but simply remember it to the letter
I am so glad I had one that was the opposite. I discussed practical applications of the subject material after class with him and at the end of the semester he gave me a B+ even though I only got a C by score because I actually grasped the material better than anyone else in the class, even if I was not able to evaluate it as well on the tests.
I’m glad for you) out teacher liked to offer discussion only to shoot us down when we tried to understand something, i was like duh that’s what teachers are for, to help us understand, if teachers don’t do that, then it’s the same as watching YouTube lectures
I once ran an LLM locally using Kobold AI. Said thing has an option to show the alternative tokens for each token it puts out, and what their probably for being chosen was. Seeing this shattered the illusion that these things are really intelligent for me. There’s at least one more thing we need to figure out before we can build an AI that is actually intelligent.
It’s cool what statistics can do, though.
That’s actually pretty neat. I tried Kobold AI a few months ago but the novelty wore off quickly. You made me curious, I’m going to check out that option once I get home. Is it just a toggleable opyiont option or do you have to mess with some hidden settings?
Just as I was about to give up, it somehow worked: https://imgchest.com/p/9p4ne9m9m4n I didn’t really do anything different this time around, so no idea why it didn’t work at first.
It’s been about a year since I saw the probabilities. I took another look at it just now, and while I can find the toggle in the settings, I can’t find the context menu where the probabilities are shown.
So… LLMS are… teenagers?
The way I’ve come to understand it is that LLMs are intelligent in the same way your subconscious is intelligent.
It works off of kneejerk “this feels right” logic, that’s why images look like dreams, realistic until you examine further.
We all have a kneejerk responses to situations and questions, but the difference is we filter that through our conscious mind, to apply long-term thinking and our own choices into the mix.
LLMs just keep getting better at the “this feels right” stage, which is why completely novel or niche situations can still trip it up; because it hasn’t developed enough “reflexes” for that problem yet.
LLMs are intelligent in the same way books are intelligent. What makes LLMs really cool is that instead of searching at the book or page granularity, it searches at the word granularity. It’s not thinking, but all the thinking was done for it already by humans who encoded their intelligence into words. It’s still incredibly powerful, at it’s best it could make it so no task ever needs to be performed by a human twice which would have immense efficiency gains for anything information based.
ah, yes, prejudice
They also reason which is really wierd
Is that Summer from Rick and Morty?
AI: “Keep Summer safe”
They’re predicting the next word without any concept of right or wrong, there is no intelligence there. And it shows the second they start hallucinating.
I have a silly little model I made for creating Vogoon poetry. One of the models is fed from Shakespeare. The system works by predicting the next letter rather than the next word (and whitespace is just another letter as far as it’s concerned). Here’s one from the Shakespeare generation:
KING RICHARD II:
Exetery in thine eyes spoke of aid.
Burkey, good my lord, good morrow now: my mother’s said
This is silly nonsense, of course, and for its purpose, that’s fine. That being said, as far as I can tell, “Exetery” is not an English word. Not even one of those made-up English words that Shakespeare created all the time. It’s certainly not in the training dataset. However, it does sound like it might be something Shakespeare pulled out of his ass and expected his audience to understand through context, and that’s interesting.
Wow, sounds amazing, big probs to you! Are you planning on releasing the model? Would be interested tbh :D
Nothing special about it, really. I only followed this TensorFlow tutorial:
https://www.tensorflow.org/text/tutorials/text_generation
The Shakespeare dataset is on there. I also have another mode that uses entries from the Joyce Kilmer Memorial Bad Poetry Contest, and also some of the works of William Topaz McGonagall (who is basically the Tommy Wiseau of 19th century English poetry). The code is the same between them, however.
Nice, thx
…yeah dude. Hence artificial intelligence.
There aren’t any cherries in artificial cherry flavoring either 🤷♀️ and nobody is claiming there is
They are a bit like you’d take just the creative writing center of a human brain. So they are like one part of a human mind without sentience or understanding or long term memory. Just the creative part, even though they are mediocre at being creative atm. But it’s shocking because we kind of expected that to be the last part of human minds to be able to be replicated.
Put enough of these “parts” of a human mind together and you might get a proper sentient mind sooner than later.
Exactly. Im not saying its not impressive or even not useful, but one should understand the limitation. For example you can’t reason with an llm in a sense that you could convince it of your reasoning. It will only respond how most people in the used dataset would have responded (obiously simplified)
You repeat your point but there already was agreement that this is how ai is now.
I fear you may have glanced over the second part where he states that once we simulated other parts of the brain things start to look different very quickly.
There do seem to be 2 kind of opinions on ai.
-
those that look at ai in the present compared to a present day human. This seems to be the majority of people overall
-
those that look at ai like a statistic, where it was in the past, what improved it and project within reason how it will start to look soon enough. This is the majority of people that work in the ai industry.
For me a present day is simply practice for what is yet to come. Because if we dont nuke ourselves back to the stone age. Something, currently undefinable, is coming.
I didn’t, I just focused on how it is today. I think it can become very big and threatening but also helpful, but that’s just pure speculation at this point :)
What i fear is AI being used with malicious intent. Corporations that use it for collecting data for example. Or governments just putting everyone in jail that they are told by an ai
I’d expect governments to use it to craft public relation strategies. An extension of what they do now by hiring the smartest sociopaths on the planet. Not sure if this would work but I think so. Basically you train an AI on previous messaging and results from polls or voting. And then you train it to suggest strategies to maximize for support for X. A kind of dumbification of the masses. Of course it’s only going to get shittier from there on out.
-
…or you might not.
It’s fun to think about but we don’t understand the brain enough to extrapolate AIs in their current form to sentience. Even your mention of “parts” of the mind are not clearly defined.
There are so many potential hidden variables. Sometimes I think people need reminding that the brain is the most complex thing in the universe, we don’t full understand it yet and neural networks are just loosely based on the structure of neurons, not an exact replica.
True it’s speculation. But before GPT3 I never imagined AI achieving creativity. No idea how you would do it and I would have said it’s a hard problem or like magic, and poof now it’s a reality. A huge leap in quality driven just by quantity of data and computing. Which was shocking that it’s “so simple” at least in this case.
So that should tell us something. We don’t understand the brain but maybe there isn’t much to understand. The biocomputing hardware is relatively clear how it works and it’s all made out of the same stuff. So it stands to reason that the other parts or function of a brain might also be replicated in similar ways.
Or maybe not. Or we might need a completely different way to organize and train other functions of a mind. Or it might take a much larger increase in speed and memory.
You say maybe there’s not much to understand about the brain but I entirely disagree, it’s the most complex object in the known universe and we haven’t discovered all of it’s secrets yet.
Generating pictures from a vast database of training material is nowhere near comparable.
Ok, again I’m just speculating so I’m not trying to argue. But it’s possible that there are no “mysteries of the brain”, that it’s just irreducible complexity. That it’s just due to the functionality of the synapses and the organization of the number of connections and weights in the brain? Then the brain is like a computer you put a program in. The magic happens with how it’s organized.
And yeah we don’t know how that exactly works for the human brain, but maybe it’s fundamentally unknowable. Maybe there is never going to be a language to describe human consciousness because it’s entirely born out of the complexity of a shit ton of simple things and there is no “rhyme or reason” if you try to understand it. Maybe the closest we get are the models psychology creates.
Then there is fundamentally no difference between painting based on a “vast database of training material” in a human mind and a computer AI. Currently AI generated images is a bit limited in creativity and it’s mediocre but it’s there.
Then it would logically follow that all the other functions of a human brain are similarly “possible” if we train it right and add enough computing power and memory. Without ever knowing the secrets of the human brain. I’d expect the truth somewhere in the middle of those two perspectives.
Another argument in favor of this would be that the human brain evolved through evolution, through random change that was filtered (at least if you do not believe in intelligent design). That means there is no clever organizational structure or something underlying the brain. Just change, test, filter, reproduce. The worst, most complex spaghetti code in the universe. Code written by a moron that can’t be understood. But that means it should also be reproducible by similar means.
Possible, yes. It’s also entirely possible there’s interactions we are yet to discover.
I wouldn’t claim it’s unknowable. Just that there’s little evidence so far to suggest any form of sentience could arise from current machine learning models.
That hypothesis is not verifiable at present as we don’t know the ins and outs of how consciousness arises.
Then it would logically follow that all the other functions of a human brain are similarly “possible” if we train it right and add enough computing power and memory. Without ever knowing the secrets of the human brain. I’d expect the truth somewhere in the middle of those two perspectives.
Lots of things are possible, we use the scientific method to test them not speculative logical arguments.
Functions of the brain
These would need to be defined.
But that means it should also be reproducible by similar means.
Can’t be sure of this… For example, what if quantum interactions are involved in brain activity? How does the grey matter in the brain affect the functioning of neurons? How do the heart/gut affect things? Do cells which aren’t neurons provide any input? Does some aspect of consciousness arise from the very material the brain is made of?
As far as I know all the above are open questions and I’m sure there are many more. But the point is we can’t suggest there is actually rudimentary consciousness in neural networks until we have pinned it down in living things first.
Knowing that LLMs are just “parroting” is one of the first steps to implementing them in safe, effective ways where they can actually provide value.
The next step is to understand much more and not get stuck on the most popular semantic trap
Then you can begin your journey man
There are so, so many llm chains that do way more than parrot. It’s just the last popular catchphrase.
Very tiring to keep explaining that because just shallow research can make you understand more than it’s a parrot comment. We are all parrots. It’s extremely irrelevant to the ai safety and usefulness debates
Most llm implementations use frameworks to just develop different understandings, and it’s shit, but it’s just not true that they only parrot known things they have internal worlds especially when looking at agent networks
I think a better way to view it is that it’s a search engine that works on the word level of granularity. When library indexing systems were invented they allowed us to look up knowledge at the book level. Search engines allowed look ups at the document level. LLMs allow lookups at the word level, meaning all previously transcribed human knowledge can be synthesized into a response. That’s huge, and where it becomes extra huge is that it can also pull on programming knowledge allowing it to meta program and perform complex tasks accurately. You can also hook them up with external APIs so they can do more tasks. What we have is basically a program that can write itself based on the entire corpus of human knowledge, and that will have a tremendous impact.
LLMs definitely provide value its just debatable whether they’re real AI or not. I believe they’re going to be shoved in a round hole regardless.
People learn the same way, we do things that bring us satisfaction and get us approval.
We use words to describe our thoughts and understanding. LLMs order words by following algorithms that predict what the user wants to hear. It doesn’t understand the meaning or implications of the words it’s returning.
It can tell you the definition of an apple, or how many people eat apples, or whatever apple data it was trained on, but it has no thoughts of it’s own about apples.
That’s the point that OOP was making. People confuse ordering words with understanding. It has no understanding about anything. It’s a large language model - it’s not capable of independent thought.
I think that the question of what “understanding” is will become important soon, if not already. Most people don’t really understand as much as you might think we do, an apple for example has properties like flavor, texture, appearance, weight and firmness it also is related to other things like trees and is in categories like food or fruit. A model can store the relationship of apple to other things and the properties of apples, the model could probably be given “personal preferences” like a preferred flavor profile and texture profile and use this to estimate if apples would be preferred by the preferences and give reasonings for it.
Unique thought is hard to define and there is probably a way to have a computer do something similar enough to be indistinguishable, probably not through simple LLMs. Maybe using a LLM as a way to convert internal “ideas” to external words and external words to internal “ideas” to be processed logically probably using massive amounts of reference materials, simulation, computer algebra, music theory, internal hypervisors or some combination of other models.
Keep seething, OpenAI’s LLMs will never achieve AGI that will replace people
That was never the goal… You might as well say that a bowling ball will never be effectively used to play golf.
I agree, but it’s so annoying when you work as IT and your non-IT boss thinks AI is the solution to every problem.
At my previous work I had to explain to my boss at least once a month why we can’t have AI diagnosing patients (at a dental clinic) or reading scans or proposing dental plans… It was maddening.
I find that these LLMs are great tools for a professional. So no, you still need the professional but it is handy if an ai would say, please check these places. A tool, not a replacemenrt.
That was never the goal…
Most CEOs seem to not have got the memo…
They did, I think. It’s most of the general public that don’t know. CEOs just take advantage of this to sell shit.
Next you’ll tell me that the enemies that I face in video games arent real AI either!
I said AGI deliberately…
Buddy, nobody ever said it would
Keep seething
Keep projecting
Unfortunately the majority of people are idiots who just do this in real life, parroting populous ideology without understanding anything more than the proper catchphrase du jour. And there are many employed professionals who are paid to read a script, or output mundane marketing content, or any “content”. And for that, LLMs are great.
It’s the elevator operator of technology as applied to creative writers. Instead of “hey intern, write the next article about 25 things these idiots need to buy and make sure 90% of them are from our sponsors” it goes to AI. The writer was never going to purchase a few different types of each product category, blindly test them and write a real article. They are just shilling crap they are paid to shill making it look “organic” because many humans are too stupid to not know it’s a giant paid for ad.
I think LLMs are neat, and Teslas are neat, and HHO generators are neat, and aliens are neat…
…but none of them live up to all of the claims made about them.
HHO generators
…What are these? Something to do with hydrogen? Despite it not making sense for you to write it that way if you meant H2O, I really enjoy the silly idea of a water generator (as in, making water, not running off water).
HHO generators are a car mod that some backyard scientists got into, but didn’t actually work. They involve cracking hydrogen from water, and making explosive gasses some claimed could make your car run faster. There’s lots of YouTube videos of people playing around with them. Kinda dangerous seeming… Still neat.
Thanks! I hadn’t heard of this before, hydrogen fueled cars, sure, but not this. 😄
So super informed OP, tell me how they work. technically, not CEO press release speak. explain the theory.
I’m not OP, and frankly I don’t really disagree with the characterization of ChatGPT as “fancy autocomplete”. But…
I’m still in the process of reading this cover-to-cover, but Chapter 12.2 of Deep Learning: Foundations and Concepts by Bishop and Bishop explains how natural language transformers work, and then has a short section about LLMs. All of this is in the context of a detailed explanation of the fundamentals of deep learning. The book cites the original papers from which it is derived, most of which are on ArXiv. There’s a nice copy on Library Genesis. It requires some multi-variable probability and statistics, and an assload of linear algebra, reviews of which are included.
So obviously when the CEO explains their product they’re going to say anything to make the public accept it. Therefore, their word should not be trusted. However, I think that when AI researchers talk simply about their work, they’re trying to shield people from the mathematical details. Fact of the matter is that behind even a basic AI is a shitload of complicated math.
At least from personal experience, people tend to get really aggressive when I try to explain math concepts to them. So they’re probably assuming based on their experience that you would be better served by some clumsy heuristic explanation.
IMO it is super important for tech-inclined people interested in making the world a better place to learn the fundamentals and limitations of machine learning (what we typically call “AI”) and bring their benefits to the common people. Clearly, these technologies are a boon for the wealthy and powerful, and like always, have been used to fuck over everyone else.
IMO, as it is, AI as a technology has inherent patterns that induce centralization of power, particularly with respect to the requirement of massive datasets, particularly for LLMs, and the requirement to understand mathematical fundamentals that only the wealthy can afford to go to school long enough to learn. However, I still think that we can leverage AI technologies for the common good, particularly by developing open-source alternatives, encouraging the use of open and ethically sourced datasets, and distributing the computing load so that people who can’t afford a fancy TPU can still use AI somehow.
I wrote all this because I think that people dismiss AI because it is “needlessly” complex and therefore bullshit. In my view, it is necessarily complex because of the transformative potential it has. If and only if you can spare the time, then I encourage you to learn about machine learning, particularly deep learning and LLMs.
That’s my point. OP doesn’t know the maths, has probably never implemented any sort of ML, and is smugly confident that people pointing out the flaws in a system generating one token at a time are just parroting some line.
These tools are excellent at manipulating text (factoring in the biases they have, I wouldn’t recommended trying to use one in a multinational corporation in internal communications for example, as they’ll clobber non euro derived culture) where the user controls both input and output.
Help me summarise my report, draft an abstract for my paper, remove jargon from my email, rewrite my email in the form of a numbered question list, analyse my tone here, write 5 similar versions of this action scene I drafted to help me refine it. All excellent.
Teach me something I don’t know (e.g. summarise article, answer question etc?) disaster!
They can summarize articles fairly well
No, they can summarise articles very convincingly! Big difference.
They have no model of what’s important, or truth. Most of the time they probably do ok but unless you go read the article you’ll never know if they left out something critical, hallucinated details, or inverted the truth or falsity of something.
That’s the problem, they’re not an intern they don’t have a human mind. They recognise patterns in articles and patterns in summaries, they non deterministically adjust the patterns in the article towards the patterns in summaries of articles. Do you see the problem? They produce stuff that looks very much like an article summary but do not summarise, there is no intent, no guarantee of truth, in fact no concern for truth at all except what incidentally falls out of the statistical probability wells.
That’s a good way of explaining it. I suppose you’re using a stricter definition of summary than I was.
I think it’s really important to keep in mind the separation between doing a task and producing something which looks like the output of a task when talking about these things. The reason being that their output is tremendously convincing regardless of its accuracy, and given that writing text is something we only see human minds do it’s so easy to ascribe intent behind the emission of the model that we have no reason to believe is there.
Amazingly it turns out that often merely producing something which looks like the output of a task apparently accidentally accomplishes the task on the way. I have no idea why merely predicting the next plausible word can mean that the model emits something similar to what I would write down if I tried to summarise an article! That’s fascinating! but because it isn’t actually setting out to do that there’s no guarantee it did that and if I don’t check the output will be indistinguishable to me because that’s what the models are built to do above all else.
So I think that’s why we to keep them in closed loops with person -> model -> person, and explaining why and intuiting if a particularly application is potentially dangerous or not is hard if we don’t maintain a clear separation between the different processes driving human vs llm text output.
Yeah this is basically how I think of them too. It’s fascinating that it works.
You are so extremely outdated in your understanding, For one that attacks others for not implementing their own llm
They are so far beyond the point you are discussing atm. Look at autogen and memgpt approaches, the way agent networks can solve and develop way beyond that point we were years ago.
It really does not matter if you implement your own llm
Then stay out of the loop for half a year
It turned out that it’s quite useless to debate the parrot catchphrase, because all intelligence is parroting
It’s just not useful to pretend they only “guess” what a summary of an article is
They don’t. It’s not how they work and you should know that if you made one
Fact of the matter is that behind even a basic AI is a shitload of complicated math.
Depending on how simple something can be to be considered an AI, the math is surprisingly simple compared to what an average person might expect. The theory behind it took a good amount of effort to develop, but to make something like a basic image categorizer (eg. optical character recognition) you really just need some matrix multiplication and calculating derivatives-- non-math-major college math type stuff.
Come on… It’s not impressive to just not be aware of where the bar is for most people. No, it’s not complex math but you are debating people that read headlines only and then go fully into imagination of what it says
you really just need some matrix multiplication and calculating derivatives-- non-math-major college math type stuff.
Well sure you don’t need a math degree for that, but most people really need to put some time into those topics. I.e., that kind of math is complex enough to constitute a barrier to entry into the field, particularly people with no free time to self-study or money for school.
Said differently: matrix math and basic calculus is hard, just not for you and I.
Point taken
Doesn’t matter had sex
you… you fucked the LLM?
Yusssss
https://imdb.com/title/tt0470752 Ex Machina (2014)
A young programmer is selected to participate in a ground-breaking experiment in synthetic intelligence by evaluating the human qualities of a highly advanced humanoid A.I.
Dude you gave me a heart attack, I was like NO WAY that came out in 2004. It didn’t, it was 2014, which is still like probably twice as old as I would’ve guessed but not as bad.
And yes it is a fantastic movie, go watch it if you haven’t seen it.
Hahaha I’ll fix it
Unfortunately they made it seem like a horror movie in the trialers when it’s more a dramatic thriller. Interesting watch and even holds up well today TEN YEARS LATER.
The reason it’s dangerous is because there are a significant number of jobs and people out there that do exactly that. Which can be replaced…
People making content should immediately pivot to become the approvers, not the generators.
The world needs waaaay more editors, too…
This post isn’t true, LLMs do have an understanding of things.
SELF-RAG: Improving the Factual Accuracy of Large Language Models through Self-Reflection
Chess-GPT’s Internal World Model
POKÉLLMON: A Human-Parity Agent for Pokémon Battle with Large Language Models
Whilst everything you linked is great research which demonstrates the vast capabilities of LLMs, none of it demonstrates understanding as most humans know it.
This argument always boils down to one’s definition of the word “understanding”. For me that word implies a degree of consciousness, for others, apparently not.
To quote GPT-4:
LLMs do not truly understand the meaning, context, or implications of the language they generate or process. They are more like sophisticated parrots that mimic human language, rather than intelligent agents that comprehend and communicate with humans. LLMs are impressive and useful tools, but they are not substitutes for human understanding.
You are moving goal posts
“understanding” can be given only when you reach like old age as a human and if you meditated in a cave
That’s my definition for it
No one is moving goalposts, there is just a deeper meaning behind the word “understanding” than perhaps you recognise.
The concept of understanding is poorly defined which is where the confusion arises, but it is definitely not a direct synonym for pattern matching.
When people say that the model “understands”, it means just that, not that it is human, and not that it does so exactly humans do. Judging its capabilities by how close it’s mimicking humans is pointless, just like judging a boat by how well it can do the breast stroke. The value lies in its performance and output, not in imitating human cognition.
Understanding is a human concept so attributing it to an algorithm is strange.
It can be done by taking a very shallow definition of the word but then we’re just entering a debate about semantics.
Understanding is a human concept so attributing it to an algorithm is strange.
Yes sorry probably shouldn’t have used the word “human”. It’s a concept that we apply to living things that experience the world.
Animals certainly understand things but it’s a sliding scale where we use human understanding as the benchmark.
My point stands though, to attribute it to an algorithm is strange.
I’m starting to wonder about you though.
