…Duh. 🤓
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.
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.
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.
'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.
They do it because it works on the whole. If straight titles were as effective they’d be used instead.
The one weird trick that makes clickbait work
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.
That’s mildly depressing.
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.
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.
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.
Thanks for copypasting here. I wonder if the “prediction” is not as expected only in that case, when making rhymes. I also notice that its way of counting feels interestingly not too different from how I count when I need to come up fast with an approximate sum.
Isn’t that the “new math” everyone was talking about?
Thanks
🙏
Is that a weird method of doing math?
I mean, if you give me something borderline nontrivial like, say 72 times 13, I will definitely do some similar stuff. “Well it’s more than 700 for sure, but it looks like less than a thousand. Three times seven is 21, so two hundred and ten, so it’s probably in the 900s. Two times 13 is 26, so if you add that to the 910 it’s probably 936, but I should check that in a calculator.”
Do you guys not do that? Is that a me thing?
I do much the same in my head.
Know what’s crazy? We sling bags of mulch, dirt and rocks onto customer vehicles every day. No one, neither coworkers nor customers, will do simple multiplication. Only the most advanced workers do it. No lie.
Customer wants 30 bags of mulch. I look at the given space:
“Let’s do 6 stacks of 5.”
Everyone proceeds to sling shit around in random piles and count as we go. And then someone loses track and has to shift shit around to check the count.
Yeah, one of my family members is a bricklayer and he can work out a bill of materials in his head based on the dimensions in an architectural plan: given these dimensions and this thickness of mortar joint, I’ll need this many bricks, this many bags of mortar, this many bags of sand, this many hours of labor, etc. It’s just addition and multiplication, but his colleagues regard him as a freak. And when he first started doing it, if you’d ask him to break down his reasoning, he’d find that difficult.
This is pretty normal, in my opinion. Every time people complain about common core arithmetic there are dozens of us who come out of the woodwork to argue that the concepts being taught are important for deeper understanding of math, beyond just rote memorization of pencil and paper algorithms.
The problem with common core math isn’t that rounding is inherently bad, it’s that you don’t start with that as a framework.
Rote memorization should be minimized in school curriculum
Memory can improve with training, and it’s useful in a large number of contexts. My major beef with rote memorization in schools is that it’s usually made to be excruciatingly boring. I’d say that’s the bigger problem.
Well, I guess I do a bit of the same:) I do (70+2)(10+3) -> 700+210+20+6
Nah I do similar stuff. I think very few people actually trace their own lines of thought, so they probably don’t realize this is how it often works.
Huh. I visualize a whiteboard in my head. Then I…do the math.
I’m also fairly certain I’m autistic, so… ¯\_(ツ)_/¯
72 * 10 + 70 * 3 + 2 * 3
That’s what I do in my head if I need an exact result. If I’m approximateing I’ll probably just do something like 70 * 15 which is much easier to compute (70 * 10 + 70 * 5 = 700 + 350 = 1050).
(72 * 10) + (2 * 3) = x
There, fixed, because otherwise order of operation gets fucky.
No it doesn’t, multiplication and division always take precedence over addition and subtraction. You’d need parentheses to clarify what is in the divisor since that can be ambiguous with line notation.
OK, I’ve been willing to just let the examples roll even though most people are just describing how they’d do the calculation, not a process of gradual approximation, which was supposed to be the point of the way the LLM does it…
…but this one got me.
Seriously, you think 70x5 is easier to compute than 70x3? Not only is that a harder one to get to for me in the notoriously unfriendly 7 times table, but it’s also further away from the correct answer and past the intuitive upper limit of 1000.
The 7 times table is unfriendly?
I love 7 timeses. If numbers were sentient, I think I could be friends with 7.
I’ve always hated it and eight. I can only remember the ones that are familiar at a glance from the reverse table and to this day I sometimes just sum up and down from those “anchor” references. They’re so weird and slippery.
Huh.
Going back to the “being friends” thing, I think you and I could be friends due to applying qualities to numbers; but I think it might be challenging because I find 7 and 8 to be two of the best. They’re quirky, but interesting.
Thank you for the insight.
Times 5 and times 10 tables are really easy for me. So yeah, in my mind it’s an easier comuptation.
That being said having a result of a little over a 1000 gives me an estimate for the magnitude of a number – it’s around a thousand. It might be more or less but it’s not far from there.
For me personally, anything times 5 can be reached by halving the number, then multiplying that number by 10.
Example: 66 x 5 = Y
-
(66/2) x (5x2) = Y
-
cancel out the division by creating equal multiplication in the other number
-
66/2 = 33
-
5x2 = 10
-
-
33 x 10 = Y
-
33 x 10 = 330
-
Y = 330
-
See, for me, it’s not that 7*5 is easier to compute than 7*3, it’s that 5*7 is easier to compute than 7*3.
I saw your other comment about 8’s, too, and I’ve always found those to be a pain, so I reverse them, if not outright convert them to arithmetic problems. 8x4 is some unknown value, but X*8 is always X*10-2X, although do have most of the multiplication tables memorized for lower values.
8*7 is an unknown number that only the wisest sages can compute, however.
I think what’s wild about it is that it really is surprisingly similar to how we actually think. It’s very different from how a computer (calculator) would calculate it.
So it’s not a strange method for humans but that’s what makes it so fascinating, no?
Yes, agreed. And calculators are essentially tabulators, and operate almost just like a skilled person using an abacus.
We shouldn’t really be surprised because we designed these machines and programs based on our own human experiences and prior solutions to problems. It’s still neat though.
That’s what’s fascinating about how it does language in general.
The article is interesting in both the ways in which things are similar and the ways they’re different. The rough approximation thing isn’t that weird, but obviously any human would have self-awareness of how they did it and not accidentally lie about the method, especially when both methods yield the same result. It’s a weirdly effective, if accidental example of human-like reasoning versus human-like intelligence.
And, incidentally, of why AGI and/or ASI are probably much further away than the shills keep claiming.
I mean neural networks are modeled after biological neurons/brains after all. Kind of makes sense…
I wouldn’t even attempt that in my head.
I can’t keep track of things and then recall them later for the final result.Pen and paper maths I’m pretty decent at, but ask me to calculate anything in my head and it’s anyone’s guess if I remembered to carry the 1 or not. Ever since learning about aphantasia I’m wondering if the lack of being able to visually store values has something to do with it.
I can visually store values and I still struggle. :(
Ever since learning about aphantasia I’m wondering if the lack of being able to visually store values has something to do with it.
Here’s some anecdotal evidence. Until I was 12 or 13, I could do absurdly complex arithmetical calculations in my head. My memory of it was of visualizing intermediate calculations as if they were on a screen in my head. I’d close my eyes to minimize distracting external stimuli. I’d get pocket money because my dad would get his friends to bet on whether I could correctly multiply two 7-digit phone numbers, and when I won, which I always did, he’d give the money to me. He had an old-school electromechanical calculator he’d use to check the results.
Neither of my parents and none of my many siblings had this ability.
I was able to use a similar visualization technique to memorize long passages of music and text. That stayed with me post-puberty, though again at a lesser extent. I’ve also been able to learn languages more quickly than most.
Once puberty kicked in, my ability to visualize declined significantly, though to compensate, I learned some mental arithmetics tricks that I still use now. I was able to get an MS in mathematics without much effort, since that relied on higher-level reasoning and not all that much on powerful memory or visualization. I didn’t pursue a Ph.D. due to lack of money but I think I could have gotten one (though I despise academic politics).
So I think your comment about aphantasia is at least directionally correct, at least as applied to people. But there’s little reason to assume LLMs would do things the same way a human mind does, though both might operate under some similar information-theoretic constraints that would cause convergent evolution.
How I’d do it is basically
72 * (10+3)
(72 * 10) + (72 * 3)
(720) + (3*(70+2))
(720) + (210+6)
(720) + (216)
936
Basically I break the numbers apart into easier chunks and then add them together.
This is what I do, except I would add 700 and 236 at the end.
Well except I would probably add 700 and 116 or something, because my working memory fucking sucks and my brain drops digits very easily when there’s more than 1
But you wouldn’t multiply, say, 74*14 to get the answer.
Not, but I’d do 75*10 + 75*4, then subtract the extra.
The LLM method of doing it with multiple numbers without proper interpolation though makes it extra weird
I might. Then I can subtract 74 to get 74*14, and subtract 28 to get 72*13.
I don’t generally do that to ‘weird’ numbers, I usually get closer to multiples of 5, 9, 10, or 11.
But a computer stores information differently. Perhaps it moves closer to numbers with simpler binary addresses.
I would do 720 + 3 * 70 + 3 * 2
This reminds me of learning a shortcut in math class but also knowing that the lesson didn’t cover that particular method. So, I use the shortcut to get the answer on a multiple choice question, but I use method from the lesson when asked to show my work. (e.g. Pascal’s Pyramid vs Binomial Expansion).
It might not seem like a shortcut for us, but something about this LLM’s training makes it easier to use heuristics. That’s actually a pretty big deal for a machine to choose fuzzy logic over algorithms when it knows that the teacher wants it to use the algorithm.
You’re antropomorphising quite a bit there. It is not trying to be deceptive, it’s building two mostly unrelated pieces of text and deciding the fuzzy logic is getting it the most likely valid response once and that the description of the algorithm is the most likely response to the other. As far as I can tell there’s neither a reward for lying about the process nor any awareness of what the process was anywhere in this.
Still interesting (but unsurprising) that it’s not getting there by doing actual maths, though.
Maybe you’re right. Maybe it’s Markov chains all the way down.
The only way I can think to test this would be to “poison” the training data with faulty arithmetic to see if it is just recalling precedent or actually implementing an algorithm.
Well, that’s supposed to be the point of the paper in the first place. They seem to be tracing paths through the neural net and seeing what lights up when they do things step by step. Someone posted a link to the source article somewhere in this thread.
Best they can tell, as per the article, they say the math answer and the answer to how it got to the answer are being generated independently.
“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
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.
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.
My favourite part of the day: commenting LLMentalist under AI articles.
that was a insightful piece, thanks for sharing
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.
“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 use a calculator. Which an AI should also be and not need to do weird shit to do math.
Yes, you shove it off onto another to do for you instead of doing it yourself and the ai doesnt.
Function calling is a thing chatbots can do now
Fascist. If someone does maths differently than your preference, it’s not “weird shit”. I’m facile with mental math despite what’s perhaps a non-standard approach, and it’s quite functional to be able to perform simple to moderate levels of mathematics mentally without relying on a calculator.
I am talking about the AI. It’s already a computer. It shouldn’t need to do anything other than calculate the equations. It doesn’t have a brain, it doesn’t think like a human, so it shouldn’t need any special tools or ways to help it do math. It is a calculator, after all.
Fascist
Wat
Thought police mate. You don’t tell people the way they think is weird shit just because they think differently than you. Break free from that path.
The reply was literally “*I* use a calculator” followed by “AI should use one too”. Are you suggesting that you’re an LLM or how did you cut a piece of cloth for yourself out of that?
Calling someone a fascist for that is obviously a bit OTT but you’ve ignored the “do weird shit” part of the response so it wasn’t literally what you said. Taking the full response into account you can easily interpret it as “I don’t bother with mental maths but use a calculator instead, anyone who isn’t like me is weird as shit”
That is a bit thought police-y
Wtf hahahahaha
Kek
OK but the llm is evidently shit at math so its “non-standard” approach should still be adjusted
A regular AI should use a calculator subroutine, not try to discover basic math every time it’s asked something.
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
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.
Don’t tell me that my thoughts aren’t weird enough.
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.
If the llm already knows the full sentence it’s going to output from the first word it “guesses” I wonder if you could short circuit it and say just give the full sentence instead of doing a cycle for each word of the sentence, could maybe cut down on llm energy costs.
I don’t think it knows the full sentence, it just doesn’t search for the words in the order they will be in the sentence. It finds the end-words first to make the poem rhyme, than looks for the rest of the words. I do it this way as well just like many other people trying to create any kind of rhyming text.
interestingly, too, this is a technique when you’re improvising songs, it’s called Target Rhyming.
The most effective way is to do A / B^1 / C / B^2 rhymes. You pick the B^2 rhyme, let’s say, “ibruprofen” and you get all of A and B^1 to think of a rhyme
Oh its Christmas time
And I was up on my roof when
I heard a jolly old voice
Ask me for ibuprofenAnd the audience thinks you’re fucking incredible for complex rhymes.
How can i take an article that uses the word “anywho” seriously?
Someone put 69 to research and then to article. Nice trolling.
The AIs have shrinks now?
You can become one too! Get your certification here https://mt.cert.ccc.de/
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.
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 [email protected] which I think would be a more suited comm for those who want to go more in-depth on it
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.
This is not surprising. LLMs are not designed to have any introspection capabilities.
Introspection could probably be tacked onto existing architectures in a few different ways, but as far as I know nobody’s done it yet. It will be interesting to see how that might change LLM behavior.
Then take that concept further, and let it keep introspecting and inspecting how it comes to the conclusions it does and eventually…
I’m surprised that they are surprised by this as well. What did they expect, and why? How much of this is written to imply LLMs - their business - are more advanced/capable than they actually are?
