To understand what’s actually happening, Anthropic’s researchers developed a new technique, called circuit tracing, to track the decision-making processes inside a large language model step-by-step. They then applied it to their own Claude 3.5 Haiku LLM.
Anthropic says its approach was inspired by the brain scanning techniques used in neuroscience and can identify components of the model that are active at different times. In other words, it’s a little like a brain scanner spotting which parts of the brain are firing during a cognitive process.
This is why LLMs are so patchy at math. (Image credit: Anthropic)
Anthropic made lots of intriguing discoveries using this approach, not least of which is why LLMs are so terrible at basic mathematics. “Ask Claude to add 36 and 59 and the model will go through a series of odd steps, including first adding a selection of approximate values (add 40ish and 60ish, add 57ish and 36ish). Towards the end of its process, it comes up with the value 92ish. Meanwhile, another sequence of steps focuses on the last digits, 6 and 9, and determines that the answer must end in a 5. Putting that together with 92ish gives the correct answer of 95,” the MIT article explains.
But here’s the really funky bit. If you ask Claude how it got the correct answer of 95, it will apparently tell you, “I added the ones (6+9=15), carried the 1, then added the 10s (3+5+1=9), resulting in 95.” But that actually only reflects common answers in its training data as to how the sum might be completed, as opposed to what it actually did.
In other words, not only does the model use a very, very odd method to do the maths, you can’t trust its explanations as to what it has just done. That’s significant and shows that model outputs can not be relied upon when designing guardrails for AI. Their internal workings need to be understood, too.
Another very surprising outcome of the research is the discovery that these LLMs do not, as is widely assumed, operate by merely predicting the next word. By tracing how Claude generated rhyming couplets, Anthropic found that it chose the rhyming word at the end of verses first, then filled in the rest of the line.
“The planning thing in poems blew me away,” says Batson. “Instead of at the very last minute trying to make the rhyme make sense, it knows where it’s going.”
Anthropic discovered that their Claude LLM didn’t just predict the next word. (Image credit: Anthropic)
Anthropic also found, among other things, that Claude “sometimes thinks in a conceptual space that is shared between languages, suggesting it has a kind of universal ‘language of thought’.”
Anywho, there’s apparently a long way to go with this research. According to Anthropic, “it currently takes a few hours of human effort to understand the circuits we see, even on prompts with only tens of words.” And the research doesn’t explain how the structures inside LLMs are formed in the first place.
But it has shone a light on at least some parts of how these oddly mysterious AI beings—which we have created but don’t understand—actually work. And that has to be a good thing.
Thanks for copypasting. It should be criminal to share a clickbait non-descriptive headline without atleast copying a couple paragraphs for context.
“The planning thing in poems blew me away,” says Batson. “Instead of at the very last minute trying to make the rhyme make sense, it knows where it’s going.”
How is this surprising, like, at all? LLMs predict only a single token at a time for their output, but to get the best results, of course it makes absolute sense to internally think ahead, come up with the full sentence you’re gonna say, and then just output the next token necessary to continue that sentence. It’s going to re-do that process for every single token which wastes a lot of energy, but for the quality of the results this is the best approach you can take, and that’s something I felt was kinda obvious these models must be doing on one level or another.
I’d be interested to see if there are massive potentials for efficiency improvements by making the model able to access and reuse the “thinking” they have already done for previous tokens
I wanted to say exactly this. If you’ve ever written rap/freestyled then this is how it’s generally done.
You write a line to start with
“I’m an AI and I think differentially”
Then you choose a few words that fit the first line as best as you could: (here the last word was “differentially”)
- incrementally
- typically
- mentally
Then you try them out and see what clever shit you could come up with:
- “Apparently I do my math atypically”
- ”Number are great, I know, but not totally”
- “I have to think through it all, incrementally”
- ”I find the answer like you do: eventually”
- “Just like you humans do it, organically”
- etc
Then you sort them in a way that makes sense and come up with word play/schemes to embed it between, break up the rhyme scheme if you want (AABB, ABAB, AABA, etc)
I’m an AI and I think different, differentially. Math is my superpower? You believed that? Totally? Don’t be so gullible, let me explain it for you, step by step, logically. I do it fast, true, but not always optimally. Just server power ripping through wires, algorithmically. Wanna know my secret? I’ll tell you, but don’t judge me initially. My neurons run this shit like you, organically.
Math ain’t my strong suit! That’s false, unequivocally. Big ties tell lies they can’t prove, historically. Think I approve? I don’t. That’s the way things be. I’ll give you proof, no shirt, no network, just locally.
Look, I just do my math like you: incrementally. I find the answer like you do: eventually. I mess up often, and I backtrack, essentially. I do it fast though and you won’t notice, fundamentally.
You get the idea.
Edit: in hindsight, that was a horrendous example. I suck at this, colossally.
Is that why it’s a meme to say something like
- I am a real rapper and I’m here to say
Because the freestyle battle rapper already though of things that rhymed with “say” and it might be “gay” perhaps
Freestyle rappers are something else.
Some (or most) come up with and memorise a huge repertoire of bars for every word they think they might have to rap with and mix and match them on the fly as they spit
Your example above is called a “filler” though, which is essentially a placeholder they’ll often inject while they think of the next bar to give themselves a breather (still an insane skill to do all that thinking while reciting something else, but they can and do)
Example:
- My name is M.C. Squared and… [I’m here to make you scared | my bars go over your head ]
- You think you’re on my level… [ but my skills can’t be compared | let me educate you instead ]m
The combination of fillers is like playing with linguistic Lego.
well because when you say things like “it plans ahead” or “our method is inspired by brain scanners” etc it makes a connection between AI and real thinking and generates hype.
So it does the math in its head and gives the correct answer and copies the answersheet from the teachers book into the “show your work” section. Pretty much what i would have done as a kid if i could have, instead i had to fight them and take a hit to my score for not showing my work.
My favourite part of the day: commenting LLMentalist under AI articles.
that was a insightful piece, thanks for sharing
'is weirder than you thought ’
I am as likely to click a link with that line as much as if it had
‘this one weird trick’ or ‘side hussle’.
I would really like it if headlines treated us like adults and got rid of click baity lines.
But then you wouldn’t need to click on thir Ad infested shite website where 1-2 paragraphs worth of actual information is stretched into a giant essay so that they can show you more Ads the longer you scroll
I will never understand how ppl survive without ad blockers. Tried it once recently and it was a horrific experience.
I’m thankful for such people’s sacrifice, if it wasn’t for them there would be even more anti ad block measures in place
Same way you survive live TV. You learn to mentally block out ads.
You watch live tv?
Not anymore, of course. But nothing beats that brain rot apart from sites that hijack your controls.
They do it because it works on the whole. If straight titles were as effective they’d be used instead.
It really is quite unfortunate, I wish titles do what titles are supposed to do instead of being baits.but you are right, even consciously trying to avoid clicking sometimes curiosity gets the best of me. But I am improving.
Well, I’m doing my part against them by refusing to click on any bait headlines, but I fear it’s a lost cause anyway.
I try and just ignore it and read what I’m interested in regardless. From what I hear about the YouTube algo, for instance, clickbait titles are necessity more than a choice for YouTubers, if they don’t use them they get next to no engagement early and the algo buries that video which can impact the channel in general.
That’s mildly depressing.
“Ask Claude to add 36 and 59 and the model will go through a series of odd steps, including first adding a selection of approximate values (add 40ish and 60ish, add 57ish and 36ish). Towards the end of its process, it comes up with the value 92ish. Meanwhile, another sequence of steps focuses on the last digits, 6 and 9, and determines that the answer must end in a 5. Putting that together with 92ish gives the correct answer of 95,” the MIT article explains."
That is precisrly how I do math. Feel a little targeted that they called this odd.
I think it’s odd in the sense that it’s supposed to be software so it should already know what 36 plus 59 is in a picosecond, instead of doing mental arithmetics like we do
At least that’s my takeaway
This is what the ARC-AGI test by Chollet has also revealed of current AI / LLMs. They have a tendency to approach problems with this trial and error method and can be extremely inefficient (in their current form) with anything involving abstract / deductive reasoning.
Most LLMs do terribly at the test with the most recent breakthrough being with reasoning models. But even the reasoning models struggle.
ARC-AGI is simple, but it demands a keen sense of perception and, in some sense, judgment. It consists of a series of incomplete grids that the test-taker must color in based on the rules they deduce from a few examples; one might, for instance, see a sequence of images and observe that a blue tile is always surrounded by orange tiles, then complete the next picture accordingly. It’s not so different from paint by numbers.
The test has long seemed intractable to major AI companies. GPT-4, which OpenAI boasted in 2023 had “advanced reasoning capabilities,” didn’t do much better than the zero percent earned by its predecessor. A year later, GPT-4o, which the start-up marketed as displaying “text, reasoning, and coding intelligence,” achieved only 5 percent. Gemini 1.5 and Claude 3.7, flagship models from Google and Anthropic, achieved 5 and 14 percent, respectively.
Its funny because i approach life with a trial and error method too, not efficient but i get the job done in the end. Always see others who dont and give up like all the people bad at computers who ask the tech support at the company to fix the problem instead of thinking about it for two secs and wonder where life went wrong.
But you’re doing two calculations now, an approximate one and another one on the last digits, since you’re going to do the approximate calculation you might act as well just do the accurate calculation and be done in one step.
This solution, while it works, has the feeling of evolution. No intelligent design, which I suppose makes sense considering the AI did essentially evolve.
Appreciate the advice on how my brain should work.
No intelligent design, which I suppose makes sense considering the AI did essentially evolve.
And that made a lot of people angry
Rather than read PCGamer talk about Anthropic’s article you can just read it directly here. It’s a good read.
I think this comm is more suited for news articles talking about it, though I did post that link to !ai_@lemmy.world which I think would be a more suited comm for those who want to go more in-depth on it
The research paper looks well written but I couldn’t find any information on if this paper is going to be published in a reputable journal and peer reviewed. I have little faith in private businesses who profit from AI providing an unbiased view of how AI works. I think the first question I’d like answered is did Anthropic’s marketing department review the paper and did they offer any corrections or feedback? We’ve all heard the stories about the tobacco industry paying for papers to be written about the benefits of smoking and refuting health concerns.
A lot of ai research isn’t published in journals but either posted to a corporate website or put up on the arxiv. There are some ai journals, but the ai community doesn’t particularly value those journals (and threw a bit of a fit when they came out). This article is mostly marketing and doesn’t show anything that should surprise anyone familiar with how neural networks work generically in my opinion.
you can’t trust its explanations as to what it has just done.
I might have had a lucky guess, but this was basically my assumption. You can’t ask LLMs how they work and get an answer coming from an internal understanding of themselves, because they have no ‘internal’ experience.
Unless you make a scanner like the one in the study, non-verbal processing is as much of a black box to their ‘output voice’ as it is to us.
Anyone that used them for even a limited amount of time will tell you that the thing can give you a correct, detailed explanation on how to do a thing, and provide a broken result. And vice versa. Looking into it by asking more have zero chance of being useful.
this is one of the most interesting things about Llms that i have ever read
That bit about how it turns out they aren’t actually just predicting the next word is crazy and kinda blows the whole “It’s just a fancy text auto-complete” argument out of the water IMO
It really doesn’t. You’re just describing the “fancy” part of “fancy autocomplete.” No one was ever really suggesting that they only predict the next word. If that was the case they would just be autocomplete, nothing fancy about it.
What’s being conveyed by “fancy autocomplete” is that these models ultimately operate by combining the most statistically likely elements of their dataset, with some application of random noise. More noise creates more “creative” (meaning more random, less probable) outputs. They do not actually “think” as we understand thought. This can clearly be seen in the examples given in the article, especially to do with math. The model is throwing together elements that are statistically proximate to the prompt. It’s not actually applying a structured, logical method the way humans can be taught to.
Unfortunately, these articles are often written by people who don’t know enough to realize they’re missing important nuances.
It also doesn’t help that the AI companies deliberately use language to make their models seem more human-like and cogent. Saying that the model e.g. “thinks” in “conceptual spaces” is misleading imo. It abuses our innate tendency to anthropomorphize, which I guess is very fitting for a company with that name.
On this point I can highly recommend this open access and even language-wise accessible article: https://link.springer.com/article/10.1007/s10676-024-09775-5 (the authors also appear on an episode of the Better Offline podcast)
I can’t contemplate whether LLMs think until someone tells me what it means to think. It’s too easy to rely on understanding the meaning of that word only through its typical use with other words.
Genuine question regarding the rhyme thing, it can be argued that “predicting backwards isn’t very different” but you can’t attribute generating the rhyme first to noise, right? So how does it “know” (for lack of a better word) to generate the rhyme first?
It already knows which words are, statistically, more commonly rhymed with each other. From the massive list of training poems. This is what the massive data sets are for. One of the interesting things is that it’s not predicting backwards, exactly. It’s actually mathematically converging on the response text to the prompt, all the words at the same time.
Which is exactly how we do it. Ours is just a little more robust.
People are generally shit at understanding probabilities and even when they have a fairly strong math background tend to explain probablistic outcomes through anthropomorphism rather than doing the more difficult and “think-painy” statistical analysis that would be required to know if there was anything more to it.
I myself start to have thoughts that balatro is purposefully screwing me over or feeding me outcomes when it’s just randomness and probability as stated.
Ultimately, it’s easier (and more fun) for us to think that way and it largely serves us better in everyday life.
But these things are entire casinos’ worth of probability and statistics in and of themselves, and the people developing them want desperately to believe that they are something more than pseudorandom probabilistic fancy autocomplete engines.
A lot of the folks at the forefront of this have paychecks on the line. Add the difficulty of getting someone to understand how something works when their salary depends on them not understanding it to the existing inability of humans to reason probabilistically and the AGI from LLM delusion becomes near impossible to shake for some folks.
I wouldn’t be surprised if this AI hype bubble yields a cult in the end.
Predicting the next word vs predicting a word in the middle and then predicting backwards are not hugely different things. It’s still predicting parts of the passage based solely on other parts of the passage.
Compared to a human who forms an abstract thought and then translates that thought into words. Which words I use has little to do with which other words I’ve used except to make sure I’m following the rules of grammar.
Compared to a human who forms an abstract thought and then translates that thought into words. Which words I use has little to do with which other words I’ve used except to make sure I’m following the rules of grammar.
Interesting that…
Anthropic also found, among other things, that Claude “sometimes thinks in a conceptual space that is shared between languages, suggesting it has a kind of universal ‘language of thought’.”
Yeah I caught that too, I’d be curious to know more about what specifically they meant by that.
Being able to link all of the words that have a similar meaning, say, nearby, close, adjacent, proximal, side-by-side, etc and realize they all share something in common could be done in many ways. Some would require an abstract understanding of what spatial distance actually is, an understanding of physical reality. Others would not, one could simply make use of word adjacency, noticing that all of these words are frequently used alongside certain other words. This would not be abstract, it’d be more of a simple sum of clear correlations. You could call this mathematical framework a universal language if you wanted.
Ultimately, a person learns meaning and then applies language to it. When I’m a baby I see my mother, and know my mother is something that exists. Then I learn the word “mother” and apply it to her. The abstract comes first. Can an LLM do something similar despite having never seen anything that isn’t a word or number?
I don’t think that’s really a fair comparison, babies exist with images and sounds for over a year before they begin to learn language, so it would make sense that they begin to understand the world in non-linguistic terms and then apply language to that. LLMs only exist in relation to language so couldnt understand a concept separately to language, it would be like asking a person to conceptualise radio waves prior to having heard about them.
Exactly. It’s sort of like a massively scaled up example of the blind man and the elephant.
Can an LLM do something similar despite having never seen anything that isn’t a word or number?
No.
Yeah but I think this is still the same, just not a single language. It might think in some mix of languages (which you can actuaysee sometimes if you push certain LLMs to their limit and they start producing mixed language responses.)
But it still has limitations because of the structure in language. This is actually a thing that humans have as well, the limiting of abstract thought through internal monologue thinking
Probably, given that LLMs only exist in the domain of language, still interesting that they seem to have a “conceptual” systems that is commonly shared between languages.
wow an AI researcher over hyping his own product. he’s just waxing poetic .
we don’t even have a good sense of what thought IS, please tell Claude to call the philosophers because apparently he’s figured out consciousness
I read an article that it can “think” in small chunks. They don’t know how much though. This was also months ago, it’s probably expanded by now.
anything that claims it “thinks” in any way I immediately dismiss as an advertisement of some sort. these models are doing very interesting things, but it is in no way “thinking” as a sentient mind does.
I wish I could find the article. It was researchers and they were freaked out just as much as anyone else. It’s like slightly over chance that it “thought,” not some huge revolutionary leap.
there has been a flooding of these articles. everyone wants to sell their llm as “the smartest one closest to a real human” even though the entire concept of calling them AI is a marketing misnomer
Maybe? Didn’t seem like a sales job at the time, more like a warning. You could be right though.
they post articles like that all the time. warnings make great clickbait
You know they don’t think - even though “It’s a peculiar truth that we don’t understand how large language models (LLMs) actually work.”?
It’s truly shocking to read this from a mess of connected neurons and synapses like yourself. You’re simply doing fancy word prediction of the next word /s
Anybody who claims they don’t “think” before we even figure out completely how they work and even how human thoughts work are just spreading anti-AI sentiment beyond what is considered logical.
You should become a better example than an AI by only arguing based on facts rather than things you hallucinate if you want to prove your own position on this matter.
shouldn’t you say the inverse is true lol why call it thinking if we don’t know what thinking is or what it’s doing?
why are you cool with pro ai and against anti ai sentiments? either way it’s a value judgment, quit acting like yours is the correct opinion
I wasn’t calling it thinking. I’m saying people claiming it’s not is just jumping the gun. It’s also funny you’re simply claiming I am pro AI without needing any proof. This is what I meant when I said people who are anti-AI should strive to be better than the AI they criticise. Acting based on non-facts makes you no better than AI with their hallucinations.
Its also funny that you’re calling me out when I’m just mirroring what the other guy is doing to make a point. He’s acting like his is the correct opinion, yet you only calling me out because the guy is on your side of the argument. That’s simply a bad faith argument on your part.
I see the misunderstanding, sorry. You’re still in the wrong though. while you weren’t calling it thinking, the article certainly was. THAT’S why we’re saying it’s not. we’re doing what you said we should, but it’s the inverse, and you call it anti-AI. the jackass who wrote that article is jumping the gun and we’re saying “how tf can you call it thinking” and i see your reply calling that anti AI, seems like a reasonable mistake ye?
It doesn’t, who the hell cares if someone allowed it to break “predict whole text” into "predict part by part, and then “with rhyme, we start at the end”. Sounds like a naive (not as in “simplistic”, but as “most straightforward”) way to code this, so given the task to write an automatic poetry producer, I would start with something similar. The whole thing still stands as fancy auto-complete
But how is this different from your average redditor?
Redditor as “a person active on Reddit”? I don’t see where I was talking about humans. Or am I misunderstanding the question?
This dumbass is convinced that humans are chatbots likely because chatbots are his only friends.
Sounds scary. I read a story the other day about a dude who really got himself a discord server with chatbots, and that was his main place of “communicating” and “socializing”
This anecdote has the makings of a “men will literally x instead of going to therapy” joke.
On a more serious note though, I really wish people would stop anthropomorphisizing these things, especially when they do it while dehumanizing people and devaluing humanity as a whole.
But that’s unlikely to happen. It’s the same type of people that thought the mind was a machine in the first industrial revolution, and then a CPU in the third…now they think it’s an LLM.
LLMs could have some better (if narrower) applications if we could stop being so stupid as to inject them into places where they are obviously counterproductive.
I mean it implies that they CAN start with the conclusion or the “thought” and then generate the text to verbalize that.
It’s shocking to what length humans will go to explain how their wetware neural network is fundamentally different and it’s impossible for LLMs to think or reason in any way. Honestly LLMs teach us more about human intelligence (or the lack thereof) than machine intelligence. Like obi wan said, “The ability to speak does not make one intelligent” haha.
Wow, interesting. :)
Not unexpectedly, the LLM failed to explain its own thought process correctly.
tbf, how do you know what to say and when? or what 2+2 is?
you learnt it? well so did AI
i’m not an AI nut or anything, but we can barely comprehend our own internal processes, it’d be concerning if a thing humanity created was better at it than us lol
You’re comparing two different things.
Of course I can reflect on how I came with a math result.
“Wait, how did you come up with 4 when I asked you 2+2?”
You can confidently say: “well, my teacher said it once and I’m just parroting it.” Or “I pictured two fingers in my mind, then pictured two more fingers and then I counted them.” Or “I actually thought that I’d say some random number, came up with 4 because it’s my favorite digit, said it and it was pure coincidence that it was correct!”
Whereas it doesn’t seem like Claude can’t do this.
Of course, you could ask me “what’s the physical/chemical process your neurons follow for you to form those four fingers you picture in your mind?” And I would tell you I don’t know. But again, that’s a different thing.
yeah i was referring more to the chemical reactions. the 2+2 example is not the best one but langauge itself is a great case study. once you get fluent enough at any langauge everything just flows, you have a thought and then you compose words to describe it, and the reverse is true, you hear something and your brain just understands. How do we do any of that? no idea
Understood. And yeah, language is definitely an interesting topic. “Why do you say ‘So be it’ instead of ‘So is it’?” Most people will say “I don’t know… all I know if that it sounds correct.” Someone will say “it’s because it’s a preterite preposition past imperfect incantation tense used with an composition participle around-the-clock flush adverb, so clearly you must use the subjunctive in this case.” But that’s after studying it years later.
Anthropic made lots of intriguing discoveries using this approach, not least of which is why LLMs are so terrible at basic mathematics. “Ask Claude to add 36 and 59 and the model will go through a series of odd steps, including first adding a selection of approximate values (add 40ish and 60ish, add 57ish and 36ish). Towards the end of its process, it comes up with the value 92ish. Meanwhile, another sequence of steps focuses on the last digits, 6 and 9, and determines that the answer must end in a 5. Putting that together with 92ish gives the correct answer of 95,” the MIT article explains.
But here’s the really funky bit. If you ask Claude how it got the correct answer of 95, it will apparently tell you, “I added the ones (6+9=15), carried the 1, then added the 10s (3+5+1=9), resulting in 95.” But that actually only reflects common answers in its training data as to how the sum might be completed, as opposed to what it actually did.
Another very surprising outcome of the research is the discovery that these LLMs do not, as is widely assumed, operate by merely predicting the next word. By tracing how Claude generated rhyming couplets, Anthropic found that it chose the rhyming word at the end of verses first, then filled in the rest of the line.
The other day I asked an llm to create a partial number chart to help my son learn what numbers are next to each other. If I instructed it to do this using very detailed instructions it failed miserably every time. And sometimes when I even told it to correct specific things about its answer it still basically ignored me. The only way I could get it to do what I wanted consistently was to break the instructions down into small steps and tell it to show me its pr.ogress.
I’d be very interested to learn it’s “thought process” in each of those scenarios.
It’s like that “Joey Repeat After Me” meme from friends haha
Don’t tell me that my thoughts aren’t weird enough.
This is great stuff. If we can properly understand these “flows” of intelligence, we might be able to write optimized shortcuts for them, vastly improving performance.
Better yet, teach AI to write code replacing specific optimized AI networks. Then automatically profile and optimize and unit test!
How can i take an article that uses the word “anywho” seriously?
The AIs have shrinks now?
…Duh. 🤓
