Tim Cook…go take your meds and watch Price is Right
Seeing these systems just making shit up when they’re not sure on the answer is probably the closest they’ll ever come to human behaviour.
We’ve invented the virtual politician.
To be 100 percent sure is a hallucination. Probably he tried to say that he is less than 80 percent sure.
I’m 100% sure they can’t because what they call AI isn’t intelligence.
Intelligence is whatever does the job and gets it done well.
AI is whatever makes the dollar sign number get bigger
It’s intelligent in that regard…
Even people hallucinate. Under your definition intelligence doesn’t exist
This is some real “what else besides witches floats in water” ass-logic
Very small rocks!
Wow whoosh. The point is that “AI” isn’t actually “intelligent” like a human and thus can’t “hallucinate” like an intelligent human.
All of this anthropomorphic terminology is just misleading marketing bullshit.
Who said anything about human intelligence? AIs have a different kind of intelligence, an artificial kind. I’m tired of pretending they don’t
Ever heard of the Turing test? Ever since AIs could pass it it became not a thing. Before that, playing Go was the mark of AI.
Any time an AI achieves a new thing people move goalposts. So I ask you: what does AI need to achieve to have intelligence?
Have you ever heard of the Turing test?
https://en.m.wikipedia.org/wiki/Turing_test
Here you go since you’ve heard of it but don’t understand it.
Current AIs pass it, since most people can’t reasonably tell between AI and human-written stuff every time
It’s dead simple to see if you’re talking to an LLM. The latest models don’t pass the Turing test, not even close. Asking them simple shit causes them to crap themselves really quickly.
Ask ChatGPT how many r’s there are in “veryberry”. When it gets it wrong, tell it you’re disappointed and expect a correct answer. If you do that repeatedly, you can get it to claim there’s more r’s in the word than it has letters.
The same thing actually passing a turing test would require. You’ve obviously read the words “Turing test” somewhere and thought you understood what it meant, but no robot we’ve ever produced as a species has passed the turing test. It EXPLICITLY requires that intelligence equal to (or indistinguishable from) HUMAN intelligence is shown. Without a liar reading responses, no AI we’ll produce for decades will pass the turing test.
No large language model has intelligence. They’re just complicated call and response mechanisms that guess what answer we want based on a weighted response system (we tell it directly or tell another machine how to help it “weigh” words in a response). Obviously with anything that requires massive amounts of input or nuance, like language, it’ll only be right about what it was guided on, which is limited to areas it is trained in.
We don’t have any novel interactions with AI. They are regurgitation engines, bringing forward sentences that aren’t theirs piecemeal. Given ten messages, I’m confident no major LLM would pass a Turing test.
The chat bots will pass the Turing test in a few years, maybe 5. Would that be intelligence then?
The Turing test is flawed, because while it is supposed to test for intelligence it really just tests for a convincing fake. Depending on how you set it up I wouldn’t be surprised if a modern LLM could pass it, at least some of the time. That doesn’t mean they are intelligent, they aren’t, but I don’t think the Turing test is good justification.
For me the only justification you need is that they predict one word (or even letter!) at a time. ChatGPT doesn’t plan a whole sentence out in advance, it works token by token… The input to each prediction is just everything so far, up to the last word. When it starts writing “As…” it has no concept of the fact that it’s going to write “…an AI A language model” until it gets through those words.
Frankly, given that fact it’s amazing that LLMs can be as powerful as they are. They don’t check anything, think about their answer, or even consider how to phrase a sentence. Everything they do comes from predicting the next token… An incredible piece of technology, despite it’s obvious flaws.
The Turing test is flawed, because while it is supposed to test for intelligence it really just tests for a convincing fake.
This is just conjecture, but I assume this is because the question of consciousness is not really falsifiable, so you just kind of have to draw an arbitrary line somewhere.
Like, maybe tech gets so good that we really can’t tell the difference, and only god knows it isn’t really alive. But then, how would we know not to give the machine legal rights?
For the record, ChatGPT does not pass the turing test.
ChatGPT is not designed to fool us into thinking it’s a human. It produces language with a specific tone & direct references to the fact it is a language model. I am confident that an LLM trained specifically to speak naturally could do it. It still wouldn’t be intelligent, in my view.
The Turing Test says that any person could have any conversation with a machine and there’s no chance you could tell it’s a machine. It does not say that one person could have one conversation with a machine and not be able to tell.
Current text generation models out themselves all the damn time. It can’t actually understand the underlying concepts of words. It just predicts what bit of text would be most convincing to a human based on previous text.
Playing Go was never the mark of AI, it was the mark of improving game-playing machines. It doesn’t represent “intelligence”, only an ability to predict what should happen next based on a set of training data.
It’s worth noting that after Lee Se Dol lost to Alphago, researchers found a fairly trivial Go strategy that could reliably beat the machine. It was simply such an easy strategy to counter that none of the games in the training data had included anyone attempting that strategy, so the algorithm didn’t account for how to counter it. Because the computer doesn’t know Go theory, it only knows how to predict what to do next based on the training data.
Detecting the machine correctly once is not enough. You need to guess correctly most of the time to statistically prove it’s not by chance. It’s possible for some people to do this, but I’ve seen a lot of comments on websites accusing HUMAN answers of being written by AIs.
If the current chat bots improve to reliably not be detected, would that be intelligence then?
KataGo just fixed that bug by putting those positions into the training data. The reason it wasn’t in the training data is because the training data at first was just self-play games. When games that are losses for the AI from humans are included, the bug is fixed.
When games that are losses for the AI from humans are included, the bug is fixed.
You’re not grasping the fundamental problem here.
This is like saying a calculator understands math because when you plug in the right functions, you get the right answers.
The AI grasps the strategic aspects of the game really well. To the point that if you don’t let it “read” deeply into the game tree, but only “guess” moves (that is, only use the policy network) it still plays at a high level (below professional, but strong amateur)
People can mean different things. Intelligence can mean a calculator doing a sum, and it can mean the way humans talk to each other. AI can do some intelligent things without people agreeing that it’s intelligent in the latter sense.
Ever heard of the Turing test? Ever since AIs could pass it it became not a thing.
In place of the Turing test we have a new test that informs us whether an individual can properly identify a stochastic parrot
“Hallucination” is an anthropomorphized term for what’s happening. The actual cause is much simpler, there’s no semantic distinction between true and false statements. Both are equally plausible as far as a language model is concerned, as long as it’s semantically structured like an answer to the question being asked.
That’s also pretty true for people, unfortunately. People are deeply incapable of differentiating fact from fiction.
Like how many, five?
No that’s not it at all. People know that they don’t know some things. LLMs do not.
Exactly, the LLM isn’t “thinking,” it’s just matching inputs to outputs with some randomness thrown in. If your data is high quality, a lot of the time the answers will be appropriate given the inputs. If your data is poor, it’ll output surprising things more often.
It’s a really cool technology in how much we get for how little effort we put in, but it’s not “thinking” in any sense of the word. If you want it to “think,” you’ll need to put in a lot more effort.
Your brain is also “just” matching inputs to outputs using complex statistics, a huge number of interconnects and clever digital-analog mixed ionic circuitry.
LLMs aren’t even hallucinating thou. It’s a euphamistic term to make it’s limitations sound human like
No, really, if you understood how the language models work, you would understand it’s not really intelligence. We just tend to humanize it because that’s what our brains do.
There’s a lot of great articles that summarize how we got to this stage and it’s pretty interesting. I’ll try to update this post with a link later.
I think LLMs are useful (and fun) and have a place, but intelligence they are not.
I’m still waiting for the definition of intelligence that won’t have the same moving of goalposts the Turing Test did
I’m happy with the Oxford definition: “the ability to acquire and apply knowledge and skills”.
LLMs don’t have knowledge as they don’t actually understand anything. They are algorithmic response generators that apply scores to tokens, and spit out the highest scoring token considering all previous tokens.
If asked to answer 10*5, they can’t reason through the math. They can only recognize 10, * and 5 as tokens in the training data that is usually followed by the 50 token. Thus, 50 is the highest scoring token, and is the answer it will choose. Things get more interesting when you ask questions that aren’t in the training data. If it has nothing more direct to copy from, it will regurgitate a sequence of tokens that sounds as close as possible to something in the training data: thus a hallucination.
This can be intuitively understood if you’ve gone through difficult college classes. There’s two ways to prepare for exams. You either try to understand the material, or you try to memorize it.
The latter isn’t good for actually applying the information in the future, and it’s most akin to what an LLM does. It regurgitates, but it doesn’t learn. You show it a bunch of difficult engineering problems, and it won’t be able to solve different ones that use the same principle.
The human could be described in very similar terms. People think we’re magic or something, but we to are just a weighted neural network assembling outputs based strictly on training data built from reinforcement. We are just for the moment much much better with massive models. Of course that is reductive but many seem to forget that brains suffer similarly when outside of training data.
That’s a strong claim. Got an academic paper to back that up?
I’m slightly confused. Which part needs an academic paper? I’ve made three admittedly reductive claims.
- Human brains are neural networks.
- Its outputs are based on training data built from reinforcement.
- We have a much more massive model than current artificial networks.
First, I’m not trying to make some really clever statement. I’m just saying there is a perspective where describing the human brain can generally follow a similar description. Nevertheless, let’s look at the only three assertions I make here. Given that the term neural network is given its namesake from the neurons that make up brains, I assume you don’t take issue with this. The second point, I don’t know if linking to scholarly research is helpful. Is it not well established that animals learn and use reward circuitry like the role of dopamine in neuromodulation? We also have… education, where we are fed information so that we retain it and can recount it down the road.
I guess maybe it is worth exploring the third, even though, I really wasn’t intending to make a scholarly statement. Here is an article in Scientific American that gives the number of neural connections around 100 trillion. Now, how that equates directly to model parameters is absolutely unclear, but even if you take glial cells where the number can be as low as 40-130 billion according to The search for true numbers of neurons and glial cells in the human brain: A review of 150 years of cell counting, that number is in the same order of magnitude of current models’ parameters. So I guess, if your issue is that AI models are actually larger than the human brain’s, I guess maybe there is something cogent. But given that there is likely at least a 1000:1 ratio of neural connections to neurons, I just don’t think that is really fair at all.
That’s an obsolete description of what a mammal’s brain is.
Do you have a better one?
I think the definition is “whichever is more emotionally important to you.” So, in your case, they would be very, very intelligent.
It’s kind of funny how AI has the exact same problems some humans have.
I always thought AI wouldn’t have that kind of problems, because they would be carefully fed accurate information.
Instead they are taught from things like Facebook and the thing formerly known as Twitter.
What an idiotic timeline we are in. LOLIt’s not the exact same problems humans have. It’s completely different. Marketers and hucksters just use anthropomorphic terminology to hype their dysfunctional programs.
The problem with AI hallucinations is not that the AI was fed inaccurate information, it’s that it’s coming up with information that it wasn’t fed in the first place.
As you say, this is a problem that humans have. But I’m not terribly surprised these AIs have it because they’re being built in mimicry of how aspects of the human mind works. And in some cases it’s desirable behaviour, for example when you’re using an AI as a creative assistant. You want it to come up with new stuff in those situations.
It’s just something you need to keep in mind when coming up with applications.
Not in the case of the google search AI. It quotes directly from unreliable sources.
Exactly, which is why I’ve objected in the past to calling Google Overview’s mistakes “hallucinations.” The AI itself is performing correctly, it’s giving an accurate overview of the search result it’s being told to create an overview for. It’s just being fed incorrect information.
I thought the main issue was that AI don’t really know how to say I don’t know or second guess themselves, as it would take a lot more robust architecture with multiple feedback loops. Like a brain.
Anyway, LLM’s aren’t the only AI that do this. So them being trained on Facebook data certainly isn’t the whole issue.
Yeah it’s the old garbage in, garbage out problem, the AI algorithms don’t really understand what they are outputting.
I think at this point voice recognition and text generation AI would be more useful as something like a phone assistant. You could tell it complex things like “Mute my phone for the next 2 hours” or “Notify me if I receive an email from John Smith.” Those sort of things could be easily done by AI algorithms that A) Understand your voice and B) Are programmed to know all the features of the OS. Hopefully with a known dataset like a phone OS there shouldn’t be hallucination problems, the AI could just act as an OS concierge.
Seems Siri and Alexa could already do things like that without needing LLMs trained on Facebook shit.
The narrow purpose models seem to be the most successful, so this would support the idea that a general AI isn’t going to happen from LLMs alone. It’s interesting that hallucinations are seen as a problem yet are probably part of why LLMs can be creative (much like humans). We shouldn’t want to stop them, but just control when they happen and be aware of when the AI is off the tracks. A group of different models working together and checking each other might work (and probably has already been tried, it’s hard to keep up).
Instead they are taught from things like Facebook and the thing formerly known as Twitter.
Imagine they would teach in our schools to inform yourself about all the important things, and therefore you should read as many toilet walls as newspapers…
Right? In all science fiction, artificial intelligence starts out better than us, and the only question is whether it can capture some idiosyncratic element of “being human.” Instead, AI has started out dumber than us, and we’re all standing around saying “uh what is this good for?”
What weirds me out is that the things it has issues with when generating images/video are basically a list of things lucid dreamers check on to see if they’re awake or dreaming.
-
Hands. Are your hands… Hands? Do they make sense?
-
Written language. Does it look like normal written language?
(3. Turn the lights off/4. Pinch your nose and breath through it) - these two not so much
-
How did I get here? Where was I before this? Does the transition make sense?
-
Mirrors. Are they accurate?
-
Displays on digital devices. Do they look normal?
-
Clocks. Digital and analog… Do they look like they’re telling time? Even if they do, look away and check again.
(9. Physics, try to do something physically impossible, like poking your finger through your palm. 10. Do you recognize people/do they recognize you) - on two more that aren’t relevant.
But still… It’s kinda remarkable.
Also, Nvidia launched their earth 2 earth simulator recently. So, simulation theory confirmed, I guess.
Also, check your cell phone. Despite how ubiquitous they are in our daily lives, I don’t think I’ve seen a single cell phone in my dreams. Or any other phone, for that matter.
And now that I think about it, I’ve definitely had a dream of being in my living room where there’s a TV, but I don’t remember the TV actually being in the dream.
Weird.
-
There’s also the fact that they can’t tell reality apart from fiction in general, because they don’t understand anything in the first place.
LLMs have no way of differentiating a fantasy RPG elements from IRL things. So they can lose the plot on what is being discussed suddenly, and for seemingly no reason.
LLMs don’t just “learn” facts from their training data. They learn how to pretend to be thinking, they can mimic but not really comprehend. If there were facts in the training data, it can regurgitate them, but it doesn’t actually know which facts apply to which subjects, or when to not make some up.
They learn how to pretend
True, and they are so darn good at it, that it can be somewhat confusing at times.
But the current AIs are not the ones we read about in SciFi.I’d argue that referring to it as “AI” is a stretch since it’s all A and no I.
This is why I strictly refer to these things as LLMs. That’s what they are.
This is the best summary I could come up with:
Even Apple CEO Tim Cook isn’t sure the company can fully stop AI hallucinations.
In an interview with The Washington Post, Cook said he would “never claim” that its new Apple Intelligence system won’t generate false or misleading information with 100 percent confidence.
These features will let you generate email responses, create custom emoji, summarize text, and more.
Recent examples of how AI can get things wrong include last month’s incident with Google’s Gemini-powered AI overviews telling us to use glue to put cheese on pizza or a recent ChatGPT bug that caused it to spit out nonsensical answers.
The voice assistant will turn to ChatGPT when it receives a question better suited for the chatbot, but it will ask for your permission before doing so.
In the demo of the feature shown during WWDC, you can see a disclaimer at the bottom of the answer that reads, “Check important info for mistakes.”
The original article contains 334 words, the summary contains 153 words. Saved 54%. I’m a bot and I’m open source!
That’s what it comes by not really understanding what you’re doing. Most of the AI models I work with are the state of the art just because they happen to work.
In my case, when I solve a PDE using finite difference schemes, there are precise mathematical conditions that guarantee you if the method is going to be stable or not. When I do the same using AI, I can’t tell if my method is going to work or not unless I run it. Moreover, I’ve had it sometimes fail and sometimes succeed.
It’s just the way it is for now.
I mean companies world wide just jumped in the AI bandwagon like a lot of people did with the NFT one. Mostly because AI actually has solid use cases and can make a big difference in broad situations.
Just since people are just slapping AI in everything it’s gonna end up being another fad to raise stock prices, like firing people last year.
Let’s just hope when all of the hype blows over and the general public thinks of AI as the marketing buzzword that never works quite right we’ll keep AI in the things it’s actually useful for
AI interest has come and gone. Some decades ago, people would slap the AI label to expert systems. If we go further back, one would call AI to solving problems in blocks world. It’s eventually going to fade away, just like all the previous waves did.
That’s like saying you can’t be 100% sure you never have fake news at the top of search query results. It’s just a fact.
If Apple can stop AI hallucination, any other AI company can also stop AI hallucination. Which is something they could have already done instead of making AI seem a joke on purpose. AI hallucinations are a sort of phenomena that nobody has control over. Why would Tim Cook have unique control over it?
I’m sure Tesla can do it within the decade! /s
You mean xAI?
Unless Apple became the first to figure out how, then they suddenly have a huge leg up on the rest. Which is kinda how Apple has been making their bread for most of their successes in my lifetime
Yeah. When Apple says it’s coming into a market, they mean they have already perfected it.
(Or let other companies polish up a feature/concept for a few years, slap a coat of Space Gray on it, and release it as a revolutionary “new” feature for apple)
Like the revolutionary space gray USB-C port?
eh. I don’t think Apple’s gonna be a pioneer in AI. If anybody can do it, it would be openai figuring it out first. Happy to be proven wrong tho.
Oh I’m not suggesting the will or are able to, I’m coming from a strategic standpoint
He’s only being honest for the sake of the shareholders.
Saying anything else would be lying
They can’t. AI has hallucinations. Google has shown that AI can’t even rely on external sources, either.
At least LLMs will. The only real fix we’ve seen was running it through additional specialized LLMs to try to massage out errors, but that just increases costs and scale for marginally low results.
I only trust moguls and political figures that are 100% sure of everything. I really like the confidence and it makes me feel like they deserve big paychecks and special rights because they must be so smart to have have no room for the doubt like the rest of us spineless imps. This guy is displaying weakness and should be shamed!
I bet Tim Apple is going to fire his ass.
I’m not exaggerating when I say there’s only like a dozen true experts for generative AI on the planet and even they’re not completely sure what’s going on in that blackbox. And as far as I’m aware Tim Cook isn’t even one of them. How would he know?
These programs are averaging massive amounts of data into a massive averaging function. There’s no way that a human could ever understand what’s going on inside that kind function. Humans can’t hold millions of weights/etc in their head and comprehend what it means. Otherwise, if humans could do this, there would be no point in doing this kind of statistics with computers.
I would expect that Apple has hired some of those experts and they told him.
They could make Siri change its voice and Genmoji based on the degree of certainty of the response:
- Trust me: Arnold as Terminator 😎
- Eehhhh, could be bullshit: shrugging old man meme 🤷🏻♂️
- Just kiddin’ here: whacky Jerry Lewis 🤪
They could sell different voice packages. Revive the ringtone market.
The AI is confidently wrong, that’s the whole problem. If there was an easy way to know if it could be wrong we wouldn’t have this discussion
this paper tries to do that: arxiv.org/pdf/2404.04689
there are also several other techniques I think
While it can’t “know” its own confidence level, it can distinguish between general knowledge (12” in 1’) and specialized knowledge that requires supporting sources.
At one point, I had a chatGPT memory designed for it to automatically provide sources for specialized knowledge and it did a pretty good job.
