We not only have to stop ignoring the problem, we need to be absolutely clear about what the problem is.
LLMs don’t hallucinate wrong answers. They hallucinate all answers. Some of those answers will happen to be right.
If this sounds like nitpicking or quibbling over verbiage, it’s not. This is really, really important to understand. LLMs exist within a hallucinatory false reality. They do not have any comprehension of the truth or untruth of what they are saying, and this means that when they say things that are true, they do not understand why those things are true.
That is the part that’s crucial to understand. A really simple test of this problem is to ask ChatGPT to back up an answer with sources. It fundamentally cannot do it, because it has no ability to actually comprehend and correlate factual information in that way. This means, for example, that AI is incapable of assessing the potential veracity of the information it gives you. A human can say “That’s a little outside of my area of expertise,” but an LLM cannot. It can only be coded with hard blocks in response to certain keywords to cut it from answering and insert a stock response.
This distinction, that AI is always hallucinating, is important because of stuff like this:
But notice how Reid said there was a balance? That’s because a lot of AI researchers don’t actually think hallucinations can be solved. A study out of the National University of Singapore suggested that hallucinations are an inevitable outcome of all large language models. **Just as no person is 100 percent right all the time, neither are these computers. **
That is some fucking toxic shit right there. Treating the fallibility of LLMs as analogous to the fallibility of humans is a huge, huge false equivalence. Humans can be wrong, but we’re wrong in ways that allow us the capacity to grow and learn. Even when we are wrong about things, we can often learn from how we are wrong. There’s a structure to how humans learn and process information that allows us to interrogate our failures and adjust for them.
When an LLM is wrong, we just have to force it to keep rolling the dice until it’s right. It cannot explain its reasoning. It cannot provide proof of work. I work in a field where I often have to direct the efforts of people who know more about specific subjects than I do, and part of how you do that is you get people to explain their reasoning, and you go back and forth testing propositions and arguments with them. You say “I want this, what are the specific challenges involved in doing it?” They tell you it’s really hard, you ask them why. They break things down for you, and together you find solutions. With an LLM, if you ask it why something works the way it does, it will commit to the bit and proceed to hallucinate false facts and false premises to support its false answer, because it’s not operating in the same reality you are, nor does it have any conception of reality in the first place.
usually, what I see is that the REPL they are using is never introspective enough. The ai cant on its own revert to a prevous state or give notes to itself because the response being fast and in linear time matters for a chatbot. ChatGPT can make really cool stuff when you ask it to break it’s thoght process into steps. Ones it usually fails spectacularly at. It was like pulling teeth to get it to actually do the steps and not just give the bad answer anyway.
This right here is also the reason why AI fanboys get angry when they are told that LLMs are not intelligent or even thinking at all. They don’t understand that in order for rational intelligence to exist, the LLMs should be able to have an internal, referential inner world of symbols, to contrast external input (training data) against and that is also capable of changing and molding to reality and truth criteria. No, tokens are not what I’m talking about. I’m talking about an internally consistent and persistent representation of the world. An identity, which is currently antithetical with the information model used to train LLMs. Let me try to illustrate.
I don’t remember the details or technical terms but essentially, animal intelligence needs to experience a lot of things first hand in order to create an individualized model of the world which is used to direct behavior (language is just one form of behavior after all). This is very slow and labor intensive, but it means that animals are extremely good, when they get good, at adapting said skills to a messy reality. LLMs are transactional, they rely entirely on the correlation of patterns of input to itself. As a result they don’t need years of experience, like humans for example, to develop skilled intelligent responses. They can do it in hours of sensing training input instead. But at the same time, they can never be certain of their results, and when faced with reality, they crumble because it’s harder for it to adapt intelligently and effectively to the mess of reality.
LLMs are a solipsism experiment. A child is locked in a dark cave with nothing but a dim light and millions of pages of text, assume immortality and no need for food or water. As there is nothing else to do but look at the text they eventually develop the ability to understand how the symbols marked on the text relate to each other, how they are usually and typically assembled one next to the other. One day, a slit on a wall opens and the person receives a piece of paper with a prompt, a pencil and a blank page. Out of boredom, the person looks at the prompt, it recognizes the symbols and the pattern, and starts assembling the symbols on the blank page with the pencil. They are just trying to continue from the prompt what they think would typically follow or should follow afterwards. The slit in the wall opens again, and the person intuitively pushes the paper it just wrote into the slit.
For the people outside the cave, leaving prompts and receiving the novel piece of paper, it would look like an intelligent linguistic construction, it is grammatically correct, the sentences are correctly punctuated and structured. The words even make sense and it says intelligent things in accordance to the training text left inside and the prompt given. But once in a while it seems to hallucinate weird passages. They miss the point that, it is not hallucinating, it just has no sense of reality. Their reality is just the text. When the cave is opened and the person trapped inside is left into the light of the world, it would still be profoundly ignorant about it. When given the word sun, written on a piece of paper, they would have no idea that the word refers to the bright burning ball of gas above them. It would know the word, it would know how it is usually used to assemble text next to other words. But it won’t know what it is.
LLMs are just like that, they just aren’t actually intelligent as the person in this mental experiment. Because there’s no way, currently, for these LLMs to actually sense and correlate the real world, or several sources of sensors into a mentalese internal model. This is currently the crux and the biggest problem on the field of AI as I understand it.
How do hallucinations preclude an internal representation? Couldn’t hallucinations arise from a consistent internal representation that is not fully aligned with reality?
I think you are misunderstanding the role of tokens in LLMs and conflating them with internal representation. Tokens are used to generate a state, similar to external stimuli. The internal representation, assuming there is one, is the manner in which the tokens are processed. You could say the same thing about human minds, that the representation is not located anywhere like a piece of data; it is the manner in which we process stimuli.
Not really. Reality is mostly a social construction. If there’s not an other to check and bring about meaning, there is no reality, and therefore no hallucinations. More precisely, everything is a hallucination. As we cannot cross reference reality with LLMs and it cannot correct itself to conform to our reality. It will always hallucinate and it will only coincide with our reality by chance.
I’m not conflating tokens with anything, I explicitly said they aren’t an internal representation. They’re state and nothing else. LLMs don’t have an internal representation of reality. And they probably can’t given their current way of working.
You seem pretty confident that LLMs cannot have an internal representation simply because you cannot imagine how that capability could emerge from their architecture. Yet we have the same fundamental problem with the human brain and have no problem asserting that humans are capable of internal representation. LLMs adhere to grammar rules, present information with a logical flow, express relationships between different concepts. Is this not evidence of, at the very least, an internal representation of grammar?
We take in external stimuli and peform billions of operations on them. This is internal representation. An LLM takes in external stimuli and performs billions of operations on them. But the latter is incapable of internal representation?
And I don’t buy the idea that hallucinations are evidence that there is no internal representation. We hallucinate. An internal representation does not need to be “correct” to exist.
Yet we have the same fundamental problem with the human brain
And LLMs aren’t human brains, they don’t even work remotely similarly. An LLM has more in common with an Excel spreadsheet than with a neuron. Read on the learning models and pattern recognition theories behind LLMs, they are explicitly designed to not function like humans. So we cannot assume that the same emergent properties exist on an LLM.
That’s not how science works. You are the one claiming it does, you have the burden of proof to prove they have the same properties. Thus far, assuming they don’t as they aren’t human is the sensible rational route.
Read again. I have made no such claim, I simply scrutinized your assertions that LLMs lack any internal representations, and challenged that assertion with alternative hypotheses. You are the one that made the claim. I am perfectly comfortable with the conclusion that we simply do not know what is going on in LLMs with respect to human-like capabilities of the mind.
I fucking hate how OpenAi and other such companies claim their models “understand” language or are “fluent” in French. These are human attributes. Unless they made a synthetic brain, they can take these claims and shove them up their square tight corporate behinds.
It’s not circular. LLMs cannot be fluent because fluency comes from an understanding of the language. An LLM is incapable of understanding so it is incapable of being fluent. It may be able to mimic it but that is a different thing. (In my opinion)
You might agree with the conclusion, and the conclusion might even be correct, but the poster effectively argued ‘only humans can be fluent, and it’s not a human so it isn’t fluent’ and that is absolutely circular logic.
Hard to say. You claim they are incapable of understanding, which is why they can’t be fluent. however, really, the whole argument boils down to whether they are capable of understanding. You just state that as if it’s established fact, and I believe that’s an open question at this point.
So whether it is circular depends on why you think they are incapable of understanding. If it’s like the other poster, and it’s because that’s a human(ish) only trait, and they aren’t human…then yes.
Very long layman take. Why is your guesstimation so incredibly crucial to understand, then next thing important to understand then really, really important to understand, over and over, when you are not an expert?
I’m not convinced about the “a human can say ‘that’s a little outside my area of expertise’, but an LLM cannot.” I’m sure there are a lot of examples in the training data set that contains qualification of answers and expression of uncertainty, so why would the model not be able to generate that output? I don’t see why it would require an “understanding” for that specifically. I would suspect that better human reinforcement would make such answers possible.
Because humans can do introspection and think and reflect about our own knowledge against the perceived expertise and knowledge of other humans. There’s nothing in LLMs models capable of doing this. An LLM cannot asses it own state, and even if it could, it has nothing to contrast it to. You cannot develop the concept of ignorance without an other to interact and compare with.
Some researchers compared the results of questions between chat gpt 3 and 4. One of the questions was about stacking items in a stable way. Chat gpt 3 just, in line with what you are saying about “without understanding”, listed the items saying to place them one on top of each other. No way it would have worked.
Chat gpt 4, however, said that you should put the book down first, put the eggs in a 3 x 3 grid on top of the book, trap them in a way with a laptop so they don’t roll around, and then put the bottle on top of the laptop standing up, and then balance the nail on the top of it…even noting you have to put the flat end of the nail down. This sounds a lot like understanding to me and not just rolling the dice hoping to be correct.
Yes, AI confidently gets stuff wrong. But let’s all note that there is a whole subreddit dedicated to people being confidently wrong. One doesn’t need to go any further than Lemmy to see people confidently claiming to know the truth about shit they should know is outside of their actual knowledge. We’re all guilty of this. Including refusing to learn when we are wrong. Additionally, the argument that they can’t learn doesn’t make sense because models have definitely become better.
Now I’m not saying ai is conscious, I really don’t know, but all of your shortcomings you’ve listed humans are guilty of too. So to use it as examples as to why it’s always just a hallucination, or that our thoughts are not, doesn’t seem to hold much water to me.
the argument that they can’t learn doesn’t make sense because models have definitely become better.
They have to be either trained with new data or their internal structure has to be improved. It’s an offline process, meaning they don’t learn through chat sessions we have with them (if you open a new session it will have forgotten what you told it in a previous session), and they can’t learn through any kind of self-directed research process like a human can.
all of your shortcomings you’ve listed humans are guilty of too.
LLMs are sophisticated word generators. They don’t think or understand in any way, full stop. This is really important to understand about them.
They have to be either trained with new data or their internal structure has to be improved. It’s an offline process, meaning they don’t learn through chat sessions we have with them (if you open a new session it will have forgotten what you told it in a previous session), and they can’t learn through any kind of self-directed research process like a human can.
Most human training is done through the guidance of another, additionally, most of this is training is done through an automated process where some computer is just churning through data. And while you are correct that the context does not exist from one session to the next, you can in fact teach it something and it will maintain it during the session. It’s just like moving to a new session is like talking to completely different person, and you’re basically arguing “well, I explained this one thing to another human, and this human doesn’t know it. . .so how can you claim it’s thinking?” And just imagine the disaster that would happen if you would just allow it to be trained by anyone on the web. It would be spitting out memes, racism, and right wing propaganda within days. lol
They don’t think or understand in any way, full stop.
I just gave you an example where this appears to be untrue. There is something that looks like understanding going on. Maybe it’s not, I’m not claiming to know, but I have not seen a convincing argument as to why. Saying “full stop” instead of an actual argument as to why just indicates to me that you are really saying “stop thinking.” And I apologize but that’s not how I roll.
Most human training is done through the guidance of another
Let’s take a step back and not talk about training at all, but about spontaneous learning. A baby learns about the world around it by experiencing things with its senses. They learn a language, for example, simply by hearing it and making connections - getting corrected when they’re wrong, yes, but they are not trained in language until they’ve already learned to speak it. And once they are taught how to read, they can then explore the world through signs, books, the internet, etc. in a way that is often self-directed. More than that, humans are learning at every moment as they interact with the world around them and with the written word.
An LLM is a static model created through exposure to lots and lots of text. It is trained and then used. To add to the model requires an offline training process, which produces a new version of the model that can then be interacted with.
you can in fact teach it something and it will maintain it during the session
It’s still not learning anything. LLMs have what’s known as a context window that is used to augment the model for a given session. It’s still just text that is used as part of the response process.
They don’t think or understand in any way, full stop.
I just gave you an example where this appears to be untrue. There is something that looks like understanding going on.
You seem to have ignored the preceding sentence: “LLMs are sophisticated word generators.” This is the crux of the matter. They simply do not think, much less understand. They are simply taking the text of your prompts (and the text from the context window) and generating more text that is likely to be relevant. Sentences are generated word-by-word using complex math (heavy on linear algebra and probability) where the generation of each new word takes into account everything that came before it, including the previous words in the sentence it’s a part of. There is no thinking or understanding whatsoever.
This is why Voroxpete@sh.itjust.works said in the original post to this thread, “They hallucinate all answers. Some of those answers will happen to be right.” LLMs have no way of knowing if any of the text they generate is accurate for the simple fact that they don’t know anything at all. They have no capacity for knowledge, understanding, thought, or reasoning. Their models are simply complex networks of words that are able to generate more words, usually in a way that is useful to us. But often, as the hallucination problem shows, in ways that are completely useless and even harmful.
An LLM is a static model created through exposure to lots and lots of text. It is trained and then used. To add to the model requires an offline training process, which produces a new version of the model that can then be interacted with.
But this is a deliberate decision, not an inherent limitation. The model could get feedback from the outside world, in fact this is how it’s trained (well, data is fed back into the model to update it). Of course we are limiting it to words, rather than a whole slew of inputs that a human gets. But keep in mind we have things like music and image generation AI as well. So it’s not like it can’t be also be trained on these things. Again, deliberate decision rather than inherent limitation.
We both even agree it’s true that it can learn from interacting with the world, you just insist that because it isn’t persisting, that doesn’t actually count. But it does persist, just not the the new inputs from users. And this is done deliberately to protect the models from what would inevitably happen. That being said, it’s also been fed arguably more input than a human would get in their whole life, just condescended into a much smaller period of time. So if it’s “total input” then the AI is going to win, hands down.
You seem to have ignored the preceding sentence: “LLMs are sophisticated word generators.”
I’m not ignoring this. I understand that it’s the whole argument, it gets repeated around here enough. Just saying it doesn’t make it true, however. It may be true, again I’m not sure, but simply stating and saying “full stop” doesn’t amount to a convincing argument.
They simply do not think, much less understand.
It’s not as open and shut as you wish it to be. If anyone is ignoring anything here, it’s you ignoring the fact that it went from basically just, as you said, randomly stacking objects it was told to stack stably, to actually doing so in a way that could work and describing why you would do it that way. Additionally there is another case where they asked chat gpt4 to draw a unicorn using an obscure programming language. And you know what? It did it. It was rudimentary, but it was clearly a unicorn. This is something that wasn’t trained on images at all. They even messed with the code, turning the unicorn around, removing the horn, fed it back in, and then asked it to replace the horn, and it put it back on correctly. It seemed to understand not only what an unicorn looked like, but what was the horn and where it should go when it was removed.
So to say it just can “generate more words” is something you can accuse us of as well, or possibly even just overly reductive of what it’s capable of even now.
But often, as the hallucination problem shows, in ways that are completely useless and even harmful.
There are all kinds of problems with human memory, where we imagine things all of the time. You’ve ever taken acid? If so, you would see how unreliable our brains are at always interpreting reality. And you want to really trip? Eye witness testimony is basically garbage. I exaggerate a bit, but there are so many flaws with it with people remembering things that didn’t happen, and it’s so easy to create false memories, that it’s not as convincing as it should be. Hell, it can even be harmful by convicting an innocent person.
Every short coming you’ve used to claim AI isn’t real thinking is something shared with us. It might just be inherent to intelligence to be wrong sometimes.
It’s exciting either way. Maybe it’s equivalent to a certain lobe of the brain, and we’re judging it for not being integrated with all the other parts.
I think where you are going wrong here is assuming that our internal perception is not also a hallucination by your definition. It absolutely is. But our minds are embodied, thus we are able check these hallucinations against some outside stimulus. Your gripe that current LLMs are unable to do that is really a criticism of the current implementations of AI, which are trained on some data, frozen, then restricted from further learning by design. Imagine if your mind was removed from all stimulus and then tested. That is what current LLMs are, and I doubt we could expect a human mind to behave much better in such a scenario. Just look at what happens to people cut off from social stimulus; their mental capacities degrade rapidly and that is just one type of stimulus.
Another problem with your analysis is that you expect the AI to do something that humans cannot do: cite sources without an external reference. Go ahead right now and from memory cite some source for something you know. Do not Google search, just remember where you got that knowledge. Now who is the one that cannot cite sources? The way we cite sources generally requires access to the source at that moment. Current LLMs do not have that by design. Once again, this is a gripe with implementation of a very new technology.
The main problem I have with so many of these “AI isn’t really able to…” arguments is that no one is offering a rigorous definition of knowledge, understanding, introspection, etc in a way that can be measured and tested. Further, we just assume that humans are able to do all these things without any tests to see if we can. Don’t even get me started on the free will vs illusory free will debate that remains unsettled after centuries. But the crux of many of these arguments is the assumption that humans can do it and are somehow uniquely able to do it. We had these same debates about levels of intelligence in animals long ago, and we found that there really isn’t any intelligent capability that is uniquely human.
This seems to be a really long way of saying that you agree that current LLMs hallucinate all the time.
I’m not sure that the ability to change in response to new data would necessarily be enough. They cannot form hypotheses and, even if they could, they have no way to test them.
My thesis is that we are asserting the lack of human-like qualities in AIs that we cannot define or measure. Assertions should be made on data, not uneasy feelings arising when an LLM falls into the uncanny valley.
But we do know how they operate. I saw a post a while back where somebody asked the LLM how it was calculating (incorrectly) the date of Easter. It answered with the formula for the date of Easter. The only problem is that that was a lie. It doesn’t calculate. You or I can perform long multiplication if asked to, but the LLM can’t (ironically, since the hardware it runs on is far better at multiplication than we are).
We do not know how LLMs operate. Similar to our own minds, we understand some primitives, but we have no idea how certain phenomenon emerge from those primitives. Your assertion would be like saying we understand consciousness because we know the structure of a neuron.
They do not have any comprehension of the truth or untruth of what they are saying, and this means that when they say things that are true, they do not understand why those things are true.
Which can be beautifully exploited with sponsored content.
Alternative title for this year Google I/O: AI vomit. You can watch Verge’s TL;DW video on Google I/O. There is no panel that did not mention AI. Most of it is “user centric”.
AI can deliver and gather ad data. The bread and butter for Google.
As to how it relates to the quote. It is up to Google to make it as truthful as they want it to be. And given ads is their money driver.
Well stated and explained. I’m not an AI researcher but I develop with LLMs quite a lot right now.
Hallucination is a huge problem we face when we’re trying to use LLMs for non-fiction. It’s a little bit like having a friend who can lie straight-faced and convincingly. You cannot distinguish whether they are telling you the truth or they’re lying until you rely on the output.
I think one of the nearest solutions to this may be the addition of extra layers or observer engines that are very deterministic and trained on only extremely reputable sources, perhaps only peer reviewed trade journals, for example, or sources we deem trustworthy. Unfortunately this could only serve to improve our confidence in the facts, not remove hallucination entirely.
It’s even feasible that we could have multiple observers with different domains of expertise (i.e. training sources) and voting capability to fact check and subjectively rate the LLMs output trustworthiness.
But all this will accomplish short term is to perhaps roll the dice in our favor a bit more often.
The perceived results from the end users however may significantly improve. Consider some human examples: sometimes people disagree with their doctor so they go see another doctor and another until they get the answer they want. Sometimes two very experienced lawyers both look at the facts and disagree.
The system that prevents me from knowingly stating something as true, despite not knowing, without some ability to back up my claims is my reputation and my personal values and ethics. LLMs can only pretend to have those traits when we tell them to.
Consider some human examples: sometimes people disagree with their doctor so they go see another doctor and another until they get the answer they want. Sometimes two very experienced lawyers both look at the facts and disagree.
This actually illustrates my point really well. Because the reason those people disagree might be
Different awareness of the facts (lawyer A knows an important piece of information lawyer B doesn’t)
Different understanding of the facts (lawyer might have context lawyer B doesn’t)
Different interpretation of the facts (this is the hardest to quantify, as its a complex outcome of everything that makes us human, including personality traits such as our biases).
Whereas you can ask the same question to the same LLM equipped with the same data set and get two different answers because it’s just rolling dice at the end of the day.
If I sit those two lawyers down at a bar, with no case on the line, no motivation other than just friendly discussion, they could debate the subject and likely eventually come to a consensus, because they are sentient beings capable of reason. That’s what LLMs can only fake through smoke and mirrors.
Honestly I feel people are using them completely wrong.
Their real power is their ability to understand language and context.
Turning natural language input into commands that can be executed by a traditional software system is a huge deal.
Microsoft released an AI powered auto complete text box and it’s genius.
Currently you have to type an exact text match in an auto complete box. So if you type cats but the item is called pets you’ll get no results. Now the ai can find context based matches in the auto complete list.
This is their real power.
Also they’re amazing at generating non factual based things. Stories, poems etc.
So if you type cats but the item is called pets get no results. Now the ai can find context based matches in the auto complete list.
Google added context search to Gmail and it’s infuriating. I’m looking for an exact phrase that I even put in quotes but Gmail returns a long list of emails that are vaguely related to the search word.
It shouldn’t even automatically fallback. If I am looking for an exact phrase and it doesn’t exist, the result should be “nothing found”, so that I can search somewhere else for the information. A prompt, “Nothing found. Look for related information?” Would be useful.
But returning a list of related information when I need an exact result is worse than not having search at all.
Fuzzy matching is a search technique that uses a set of fuzzy rules to compare two strings. The fuzzy rules allow for some degree of similarity, which makes the search process more efficient.
That allows for mis typing etc. it doesn’t allow context based searching at all. Cat doesn’t fuzz with pet. There is no similarity.
They’re really, really bad at context. The main failure case isn’t making things up, it’s having text or image in part of the result not work right with text or image in another part because they can’t even manage context across their own replies.
See images with three hands, where bow strings mysteriously vanish etc.
New models are like really good at context, the amount of input that can be given to them has exploded (fairly) recently… So you can give whole datasets or books as context and ask questions about them.
Searching with synonym matching is almost.decades old at this point. I worked on it as an undergrad in the early 2000s.and it wasn’t new then, just complicated. Google’s version improved over other search algorithms for a long time.and then trashed it by letting AI take over.
Google’s algorithm has pretty much always used AI techniques.
It doesn’t have to be a synonym. That’s just an example.
Typing diabetes and getting medical services as a result wouldn’t be possible with that technique unless you had a database of every disease to search against for all queries.
The point is AI means you don’t have to have a giant lookup of linked items as it’s trained into it already.
Yes, synonym searching doesn’t strictly mean the thesaurus. There are a lot of different ways to connect related terms and some variation in how they are handled from one system to the next. Letting machine learning into the mix is a very new step in a process that Library and Information Sci has been working on for decades.
Yes and no. I’ve had to insert a LOT of meaning to get a story worth any substance, and I’ve had to do a lot of editing to get good images. It’s really good at giving me a figure that’s 90% done, but that last 10% touching up still often takes me a day or so of work.
That’s why I only use Perplexity. ChatGPT can’t give me sources unless I pay, so I can’t trust information it gives me and it also hallucinated a lot when coding, it was faster to search in the official documentation rather than correcting and debugging code “generated” by ChatGPT.
I use Perplexity + SearXNG, so I can search a lot faster, cite sources and it also makes summaries of your search, so it saves me time while writing introductions and so.
It sometimes hallucinates too and cites weird sources, but it’s faster for me to correct and search for better sources given the context and more ideas. In summary, when/if you’re correcting the prompts and searching apart from Perplexity, you already have something useful.
BTW, I try not to use it a lot, but it’s way better for my workflow.
Exactly. The big problem with LLMs is that they're so good at mimicking understanding that people forget that they don't actually have understanding of anything beyond language itself.
The thing they excel at, and should be used for, is exactly what you say - a natural language interface between humans and software.
Like in your example, an LLM doesn't know what a cat is, but it knows what words describe a cat based on training data - and for a search engine, that's all you need.
Hallucinations, like depression, is a multifaceted issue. Training data is only a piece of it. Quantized models, overfitted training models rely on memory at the cost of obviously correct training data. Poorly structured Inferences can confuse a model.
It will never be solved. Even the greatest hypothetical super intelligence is limited by what it can observe and process. Omniscience doesn’t exist in the physical world. Humans hallucinate too - all the time. It’s just that our approximations are usually correct, and then we don’t call it a hallucination anymore. But realistically, the signals coming from our feet take longer to process than those from our eyes, so our brain has to predict information to create the experience. It’s also why we don’t notice our blinks, or why we don’t see the blind spot our eyes have.
AI representing a more primitive version of our brains will hallucinate far more, especially because it cannot verify anything in the real world and is limited by the data it has been given, which it has to treat as ultimate truth. The mistake was trying to turn AI into a source of truth.
Hallucinations shouldn’t be treated like a bug. They are a feature - just not one the big tech companies wanted.
When humans hallucinate on purpose (and not due to illness), we get imagination and dreams; fuel for fiction, but not for reality.
Could not have said it better. The whole reason contemporary programs haven’t been able to adapt to the ambiguity of real world situations is because they require rigidly defined parameters to function. LLMs and AI make assumptions and act on shaky info - That’s the whole point. If people waited for complete understanding of every circumstance and topic, we’d constantly be trapped in indecision. Without the ability to test their assumptions in the real world, LLMs will be like children.
I’m not an expert in AI, I will admit. But I’m not a layman either. We’re all anonymous on here anyways. Why not leave a comment explaining what you disagree with?
I want to just understand why people get so passionate about explaining how things work, especially in this field where even the experts themselves just don’t understand how it works? It’s just an interesting phenomenon to me
The not understanding hlw it works thing isn’t universal in ai from my understanding. And people understand how a lot of it works even then. There may be a few mysterious but its not sacrificing chickens to Jupiter either.
Seems like there are a lot of half baked ideas online about AI that seem to come from assumptions based on some sci-fi ideal or something. People are shocked that an artificial intelligence gets things wrong when they themselves have probably made a handful of incorrect assumptions today. This Tom Scott talk is a great explanation of how truth can never be programmed into anything. And will never really be obtainable to humanity in the foreseeable future.
Yeah! That’s probably a good portion of it, but exasterbated by the general hate for ai, which is understandable due to the conglomorates abusive training data
What exactly are your bona fides that you get to play the part of the exasperated “expert” here? And, more importantly, why should I give a fuck?
I constantly hear this shit from other self-appointed experts in this field as if no one is allowed to discuss, criticize, or form opinions on the implications of this technology besides those few who ‘truly understand’.
Hallucinations in AI are fairly well understood as far as I’m aware. Explained in high level on the Wikipedia page for it.And I’m honestly not making any objective assessment of the technology itself. I’m making a deduction based on the laws of nature and biological facts about real life neural networks. (I do say AI is driven by the data it’s given, but that’s something even a layman might know)
How to mitigate hallucinations is definitely something the experts are actively discussing and have limited success in doing so (and I certainly don’t have an answer there either), but a true fix should be impossible.
I can’t exactly say why I’m passionate about it. In part I want people to be informed about what AI is and is not, because knowledge about the technology allows us to make more informed decision about the place AI takes in our society. But I’m also passionate about human psychology and creativity, and what we can learn about ourselves from the quirks we see in these technologies.
Not really, no, because these aren’t biological, and the scientists that work with it is more interested in understanding why it works at all.
It is very interesting how the brain works, and our sensory processing is predictive in nature, but no, it’s not relevant to machine learning which works completely different
Technically if we were to develop a real general artificial general intelligence, it would be limited to the amount of knowledge that it has, but so is any given human. And it’s advantage would still be scale of operations compared to a human, since it can realistically operate on all known theoretical and practical information, where as for a human that’s simply not possible.
Though presumably, it would also be influenced by AI posting that we already have now, to some degree, the question is how it responds to that, and how well it can determine the difference between that and real human posting.
the reason why hallucinations are such a big problem currently is simply due to the fact that it’s literally a predictive text model, it doesn’t know anything. That simply wouldn’t be true for a general artificial intelligence. Not that it couldn’t hallucinate, but it wouldn’t hallucinate to the same degree, and possibly with greater motives in mind.
A lot of the reason human biology tends to obfuscate certain things is simply due to the way it’s evolved, as well as it’s potential advantages in our life. The reason we can’t see our blindspots is due to the fact that it would be much more difficult to process things otherwise. It’s the same reason our eyesight is flipped as well. It’s the same reason pain is interpreted the way that it is.
a big mistake you are making here is stating that it must be fed information that it knows to be true, this is not inherently true. You can train a model on all of the wrong things to do, as long it has the capability to understand this, it shouldn’t be a problem.
For predictive models? This is probably the case, but you can also poison the well so to speak, when it comes to those even.
Yes, a theoretical future AI that would be able to self-correct would eventually become more powerful than humans, especially if you could give it ways to run magnitudes more self-correcting mechanisms at the same time. But it would still be making ever so small assumptions when there is a gap in the information it has.
It could be humble enough to admit it doesn’t know, but it can still be mistaken and think it has the right answer when it doesn’t. It would feel neigh omniscient, but it would never truly be.
A roundtrip around the globe on glass fibre takes hundreds of milliseconds, so even if it has the truth on some matter, there’s no guarantee that didn’t change in the milliseconds it needed to become aware that the truth has changed. True omniscience simply cannot exists since information (and in turn the truth encoded by that information) also propagates at the speed of light.
a big mistake you are making here is stating that it must be fed information that it knows to be true, this is not inherently true. You can train a model on all of the wrong things to do, as long it has the capability to understand this, it shouldn’t be a problem.
The dataset that encodes all wrong things would be infinite in size, and constantly change. It can theoretically exist, but realistically it will never happen. And if it would be incomplete it has to make assumptions at some point based on the incomplete data it has, which would open it up to being wrong, which we would call a hallucination.
It could be humble enough to admit it doesn’t know, but it can still be mistaken and think it has the right answer when it doesn’t. It would feel neigh omniscient, but it would never truly be.
yeah and so are humans, so i mean, shit happens. Even then it’d likely be more accurate than a human just based off of the very fact that it knows more subjects than any given human. And all humans alive, because it’s knowledge is based off of the written works of the entirety of humanity, theoretically.
A roundtrip around the globe on glass fibre takes hundreds of milliseconds, so even if it has the truth on some matter, there’s no guarantee that didn’t change in the milliseconds it needed to become aware that the truth has changed. True omniscience simply cannot exists since information (and in turn the truth encoded by that information) also propagates at the speed of light.
well yeah, if we’re defining the ultimate truth as something that propagates through the universe at the highest known speed possible. That would be how that works, since it’s likely a device of it’s own accord, and or responsive to humans, it likely wouldn’t matter, as it would just wait a few seconds anyway.
The dataset that encodes all wrong things would be infinite in size, and constantly change. It can theoretically exist, but realistically it will never happen. And if it would be incomplete it has to make assumptions at some point based on the incomplete data it has, which would open it up to being wrong, which we would call a hallucination.
at that scale yes, but at this scale, with our current LLM technology, which was what i was talking about specifically, it wouldn’t matter. But even at that scale i don’t think it would classify as a hallucination, because a hallucination is a very specific type of being wrong. It’s literally pulling something out a thin air, and a theoretical general intelligence AI wouldn’t be pulling shit out of thin air, at best it would elaborate on what it knows already, which might be everything, or nothing, depending on the topic. But it shouldn’t just make something up out of thin air. It could very well be wrong about something, but that’s not likely to be a hallucination.
Yes, it would be much better at mitigating it and beat all humans at truth accuracy in general. And truths which can be easily individually proven and/or remain unchanged forever can basically be 100% all the time. But not all truths are that straight forward though.
What I mentioned can’t really be unlinked from the issue, if you want to solve it completely. Have you ever found out later on that something you told someone else as fact turned out not to be so? Essentially, you ‘hallucinated’ a truth that never existed, but you were just that confident it was correct to share and spread it. It’s how we get myths, popular belief, and folklore.
For those other truths, we simply ascertain the truth to be that which has reached a likelihood we consider it to be certain. But ideas and concepts we have in our minds constantly float around on that scale. And since we cannot really avoid talking to other people (or intelligent agents) to ascertain certain truths, misinterpretations and lies can sneak in to cause us to treat as truth that which is not. To avoid that would mean the having to be pretty much everywhere to personally interpret the information straight from the source. But then things like how fast it can process those things comes in to play. Without making guesses about what’s going to happen, you basically can’t function in reality.
I think you’re giving a glorified encyclopedia too much credit. The difference between us and “AI” is that we can approach knowledge from a problem solving position. We do approximate the laws of physics, but we don’t blindly take our beliefs and run with it. We put we come up with a theory that then gets rigorously criticized, then come up with ways to test that theory, then be critical of the test results and eventually we come to consensus that based on our understandings that thing is true. We’ve built entire frameworks to reduce our “hallucinations”. The reason we even know we have blind spots is because we’re so critical of our own “hallucinations” that we end up deliberately looking for our blind spots.
But the “AI” doesn’t do that. It can’t do that. The “AI” can’t solve problems, it can’t be critical of itself or what information its giving out. All our current “AI” can do is word vomit itself into a reasonable answer. Sometimes the word vomit is factually correct, sometimes it’s just nonsense.
You are right that theoretically hallucinations cannot be solved, but in practicality we ourselves have come up with solutions to minimize it. We could probably do something similar with “AI” but not when the AI is just a LLM that fumbles into sentences.
I’m not sure where you think I’m giving it too much credit, because as far as I read it we already totally agree lol. You’re right, methods exist to diminish the effect of hallucinations. That’s what the scientific method is. Current AI has no physical body and can’t run experiments to verify objective reality. It can’t fact check itself other than be told by the humans training it what is correct (and humans are fallible), and even then if it has gaps in what it knows it will fill it up with something probable - but which is likely going to be bullshit.
All my point was, is that to truly fix it would be to basically create an omniscient being, which cannot exist in our physical world. It will always have to make some assumptions - just like we do.
The fundamental difference is that the AI doesn’t know anything. It isn’t capable of understanding, it doesn’t learn in the same sense that humans learn. A LLM is a (complex!) digital machine that guesses the next most likely word based on essentially statistics, nothing more, nothing less.
It doesn’t know what it’s saying, nor does it understand the subject matter, or what a human is, or what a hallucination is or why it has them. They are fundamentally incapable of even perceiving the problem, because they do not perceive anything aside from text in and text out.
Many people’s entire thought process is an internal monologue. You think that voice is magic? It takes input and generates a conceptual internal dialogue based on what it’s previously experienced (training data for long term, context for short term). What do you mean when you say you understand something? What is the mechanism that your brain undergoes that’s defined as understanding?
Because for me it’s an internal conversation that asserts an assumption based on previous data and then attacks it with the next most probable counter argument systematically until what I consider a “good idea” emerges that is reasonably vetted. Then I test it in the real world by enacting the scientific process. The results are added to my long term memory (training data).
It doesn’t need to verify reality, it needs to be internally consistent and it’s not.
For example I was setting up logging pipeline and one of the filters didn’t work. There was seemingly nothing wrong with configuration itself and after some more tests with dummy data I was able to get it working, but it still didn’t work with the actual input data. So I have the working dummy example and the actual configuration to chatGPT and asked why the actual configuration doesn’t work. After some prompts going over what I had already tried it ended up giving me the exact same configuration I had presented as the problem. Humans wouldn’t (or at least shouldn’t) make that error because it would be internally inconsistent, the problem statement can’t be the solution.
But the AI doesn’t have internal consistency because it doesn’t really think. It’s not making sure what it’s saying is logical based on the information it knows, it’s not trying to make assumptions to solve a problem, it can’t even deduce that something true is actuality true. All it can do is predict what we would perceive as the answer.
Indeed. It doesn’t even trend towards consistency.
It’s much like the pattern-matching layer of human consciousness. Its function isn’t to filter for truth, its function is to match knowns and potentials to patterns in its environment.
AI has no notion of critical thinking. It is purely positive “thinking”, in a technical sense - it is positing based on what it “knows”, but there is no genuine concept of self, nor even of critical thinking, nor even a non-conceptual logic or consistency filter.
The simple solution is not to rely upon AI. It’s like a misinformed relative after a jar of moonshine, they might be right some of the time, or they might be totally full of shit.
I honestly don’t know why people are obsessed with relying on AI, is it that difficult to look up the answer from a reliable source?
because some jobs have to produce a bunch of bullshit text that no one will read quickly, or else parse a bunch of bullshit text for a single phrase in the midst of it all and put it in a report.
Sites like that can be blacklisted with web browser plugins. Vastly improved my DuckDuckGo experience for a while, but it’ll be a Whack-A-Mole game from both sides, and yet again my searches are littered with SEO garbage at best, and AI-generated SEO garbage full with made up stuff at worst.
is it that difficult to look up the answer from a reliable source?
With the current state of search engines and their content (almost completely unrelated garbage and shitty blogs make in like 3 minutes with 1/4 of the content poorly copy-pasted out of context from stackoverflow and most of the rest being pop-ups and ads), YES
It’s not hallucination, it’s confabulation. Very similar in its nuances to stroke patients.
Just like the pretrained model trying to nuke people in wargames wasn’t malicious so much as like how anyone sitting in front of a big red button labeled ‘Nuke’ might be without a functioning prefrontal cortex to inhibit that exploratory thought.
Human brains are a delicate balance between fairly specialized subsystems.
Right now, ‘AI’ companies are mostly trying to do it all in one at once. Yes, the current models are typically a “mixture of experts,” but it’s still all in one functional layer.
Hallucinations/confabulations are currently fairly solvable for LLMs. You just run the same query a bunch of times and see how consistent the answer is. If it’s making it up because it doesn’t know, they’ll be stochastic. If it knows the correct answer, it will be consistent. If it only partly knows, it will be somewhere in between (but in a way that can be fine tuned to be detected by a classifier).
This adds a second layer across each of those variations. If you want to check whether something is safe, you’d also need to verify that answer isn’t a confabulation, so that’s more passes.
It gets to be a lot quite quickly.
As the tech scales (what’s being done with servers today will happen around 80% as well on smartphones in about two years), those extra passes aren’t going to need to be as massive.
This is a problem that will eventually go away, just not for a single pass at a single layer, which is 99% of the instances where people are complaining this is an issue.
You just run the same query a bunch of times and see how consistent the answer is.
A lot of people are developing what I’d call superstitions on some way to overcome LLm limitations. I remember someone swearing they fixed the problem by appending “Ensure the response does not contain hallucinations” to every prompt.
In my experience, what you describe is not a reliable method. Sometimes it’s really attached to the same sort of mistakes for the same query. I’ve seen it double down, when instructed a facet of the answer was incorrect and to revise, several times I’d get “sorry for the incorrect information”, followed by exact same mistake. On the flip side, to the extent it “works”, it works on valid responses too, meaning an extra pass to ward off “hallucinations” you end up gaslighting the model and it changes the previously correct answer as if it were a hallucination.
For the SelfCheckGPT paper, which was basically this method, it was very sample dependent, continuing to see improvement up to 20 samples (their limit), but especially up to around 6 iterations…
I’ve seen it double down, when instructed a facet of the answer was incorrect and to revise, several times I’d get “sorry for the incorrect information”, followed by exact same mistake.
You can’t continue with it in context or it ruins the entire methodology. You are reintroducing those tokens when you show it back to the model, and the models are terrible at self-correcting when instructed that it is incorrect, so the step is quite meritless anyways.
You need to run parallel queries and identify shared vs non-shared data points.
It really depends on the specific use case in terms of the full pipeline, but it works really well. Even with just around 5 samples and intermediate summarization steps it pretty much shuts down completely errant hallucinations. The only class of hallucinations it doesn’t do great with are the ones resulting from biases in the relationship between the query and the training data, but there’s other solutions for things like that.
And yes, it definitely does mean inadvertently eliminating false negatives, which is why a balance has to be struck in terms of design choices.
without reading the article, this is the best summary I could come up with:
Mainstream government tied media keeps hallucinatin up facts. Republican, democrat, doesn’t matter; they hallucinate up facts. Time to stop ignoring human’s hallucination problem. At least with ai, they don’t have some subversive agenda beneath the surface when they do it. Time to help ai take over the world bbl
I’m a bit annoyed at all the people being pedantic about the term hallucinate.
Programmers use preexisting concepts as allegory for computer concepts all the time.
Your file isn’t really a file, your desktop isn’t a desk, your recycling bin isn’t a recycling bin.
[Insert the entirety of Object Oriented Programming here]
Neural networks aren’t really neurons, genetic algorithms isn’t really genetics, and the LLM isn’t really hallucinating.
But it easily conveys what the bug is. It only personifies the LLM because the English language almost always personifies the subject. The moment you apply a verb on an object you imply it performed an action, unless you limit yourself to esoteric words/acronyms or you use several words to overexplain everytime.
What I don’t like about it is that it makes it sound more benign than it is. Which also points to who decided to use that term - AI promoters/proponents.
Edit: it’s like all of the bills/acts in congress where they name them something like “The Protect Children Online Act” and you ask, “well, what does it do?” And they say something like, “it lets local police read all of your messages so they can look for any dangers to children.”
In terms of LLM hallucination, it feels like the name very aptly describes the behavior and severity. It doesn’t downplay what’s happening because it’s generally accepted that having a source of information hallucinate is bad.
I feel like the alternatives would downplay the problem. A “glitch” is generic and common, “lying” is just inaccurate since that implies intent to deceive, and just being “wrong” doesn’t get across how elaborately wrong an LLM can be.
Hallucination fits pretty well and is also pretty evocative. I doubt that AI promoters want to effectively call their product schizophrenic, which is what most people think when hearing hallucination.
Ultmately all the sciences are full of analogous names to make conversations easier, it’s not always marketing. No different than when physicists say particles have “spin” or “color” or that spacetime is a “fabric” or [insert entirety of String theory]…
After thinking about it more, I think the main issue I have with it is that it sort of anthropomorphises the AI, which is more of an issue in applications where you’re trying to convince the consumer that the product is actually intelligent. (Edit: in the human sense of intelligence rather than what we’ve seen associated with technology in the past.)
You may be right that people could have a negative view of the word “hallucination”. I don’t personally think of schizophrenia, but I don’t know what the majority think of when they hear the word.
You could invent a new word, but that doesn’t help people understand the problem.
You are looking for an existing word that describes providing unintentionally incorrect thoughts but is totally unrelated to humans. I suspect that word doesn’t exist. Every thinking word gets anthropomorphized.
The term “hallucination” has been used for years in AI/ML academia. I reading about AI hallucinations ten years ago when I was in college. The term was originally coined by researchers and mathematicians, not the snake oil salesman pushing AI today.
It’s easily the worst problem of Lemmy. Sometimes one guy has an issue with something and suddenly the whole thread is about that thing, as if everyone thought about it. No, you didn’t think about it, you just read another person’s comment and made another one instead of replying to it.
I never heard anyone complain about the term “hallucination” for AIs, but suddenly in this one thread there are 100 clonic comments instead of a single upvoted ones.
I get it, you don’t like “hallucinate”, just upvote the existing comment about it and move on. If you have anything to add, reply to that comment.
I don’t know why this specific thing is so common on Lemmy though, I don’t think it happened in reddit.
The AI isn’t alive, it’s not hallucinating… We will likely never have true AI until we figure out the Hard Problem of Consciousness, till we know what makes a human alive, we can’t make a machine alive.
The Italians actually have a name for that kind of gibberish talking that sounds real. I did some VO work on a project being directed by an Italian guy and he explained what he wanted me to do by explaining the term to me first. I’m afraid it’s been way too long since he told me for me to remember it though.
Another example would be the La Linea cartoons, where the main character speaks a gibberish which seems to approximate Italian to my ears.
Add comment