Hacker Newsnew | past | comments | ask | show | jobs | submit | zamalek's commentslogin

LLMs are notoriously terrible at multiplying large numbers: https://claude.ai/share/538f7dca-1c4e-4b51-b887-8eaaf7e6c7d3

> Let me calculate that. 729,278,429 × 2,969,842,939 = 2,165,878,555,365,498,631

Real answer is: https://www.wolframalpha.com/input?i=729278429*2969842939

> 2 165 842 392 930 662 831

Your example seems short enough to not pose a problem.


Modern LLMs, just like everyone reading this, will instead reach for a calculator to perform such tasks. I can't do that in my head either, but a python script can so that's what any tool-using LLM will (and should) do.

This is special pleading.

Long multiplication is a trivial form of reasoning that is taught at elementary level. Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper. That was Opus with extended reasoning, it had all the opportunity to get it right, but didn't. There are people who can quickly multiply such numbers in their head (I am not one of them).

LLMs don't reason.


I tried this with Claude - it has to be explicitly instructed to not make an external tool call, and it can get the right answer if asked to show its work long-form.

i assert that by your evidentiary standards humans don't reason.

presumably one of us is wrong.

therefore, humans don't reason.


Mathematics is not the only kind of reasoning, so your conclusion is false. The human brain also has compartments for different types of activities. Why shouldn't an AI be able to use tools to augment its intelligence?

I used the mathematics example only because the GP did. There are many other examples of non-reasoning, including some papers (as recent as Feb).

There are many examples of current limitations, but do you see a reason to think they are fundamental limitations? (I'm not saying they aren't, I'm curious what the evidence is for that.)

It's because of how transformers work, especially the fact that the output layer is a bunch of weights which we quite literally do a weighted random choice from. My hunch is that diffusion models would have a higher chance of doing real reasoning - or something like a latent space for reasoning.

Thinking that LLMs are intelligent arises from an incomplete understanding of how they work or, alternatively, having shareholders to keep happy.


Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper

LOL, talk about special pleading. Whatever it takes to reshape the argument into one you can win, I guess...

LLMs don't reason.

Let's see you do that multiplication in your head. Then, when you fail, we'll conclude you don't reason. Sound fair?


I can do it with a scratch pad. And I can also tell you when the calculation exceeds what I can do in my head and when I need a scratch pad. I can also check a long multiplication answer in my head (casting 9s, last digit etc.) and tell if there’s a mistake.

The LLMs also have access to a scratch pad. And importantly don’t know when they need to use it (as in, they will sometimes get long multiplication right if you ask them to show their work but if you don’t ask them to they will almost certainly get it wrong).


> And importantly don’t know when they need to use it

patently false, but hey at least you’re able to see the parallel between you with a scratch pad and an LLM with a python terminal


Sure, lets test that:

https://chatgpt.com/s/t_69c420f3118081919cf525123e39598c

https://chatgpt.com/s/t_69c4215daeb481919fdaf22498fb0c4f

Do you have a different definition of false? I'm referring to their reasoning context as their scratch pad if that wasn't clear.


The context is the scratch pad. LLMs have perfect recall (ignoring "lost in the middle") across the entire context, unlike humans. LLMs "think on paper."

The conclusion that LLMs don't reason is not a consequence of them not being able to do arithmetic, so your argument isn't valid.

Also, see https://news.ycombinator.com/newsguidelines.html

"Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.

Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.

When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."

Don't be curmudgeonly. Thoughtful criticism is fine, but please don't be rigidly or generically negative."

etc.


Plenty of humans can't do arithmetic. Can they also not reason.

Reasoning isn't a binary switch. It's a multidimensional continuum. AI can clearly reason to some extent even if it also clearly doesn't reason in the same way that a human would.


> Plenty of humans can't do arithmetic. Can they also not reason.

I just pointed out that this isn't valid reasoning ... it's a fallacy of denial of the antecedent. No one is arguing that because LLMs can't do arithmetic, therefore they can't reason. After all, zamalek said that he can't quickly multiply large numbers in his head, but he isn't saying that therefore he can't reason.

> Reasoning isn't a binary switch. It's a multidimensional continuum.

Indeed, and a lot of humans are very bad at it, as is clear from the comments I'm responding to.

> AI can clearly reason to some extent

The claim was about LLMs, not AI. This is like if someone said that chihuahuas are little and someone responded by saying that dogs are tall to some extent.

LLMs do not reason ... they do syntactic pattern matching. The appearance of reasoning is because of all the reasoning by humans that is implicit in the training data.

I've had this argument too many times ... it never goes anywhere. So I won't respond again ... over and out.


Indeed, and a lot of humans are very bad at it, as is clear from the comments I'm responding to.

This is your idea of "conversing curiously" and "editing out swipes," I suppose.

I've had this argument too many times ... it never goes anywhere. So I won't respond again ... over and out.

A real reasoning entity might pause for self-examination here. Maybe run its chain of thought for a few more iterations, or spend some tokens calling research tools. Just to probe the apparent mismatch between its own priors and those of "a lot of humans," most of whom are not, in fact, morons.


Comments should get more thoughtful and substantive

Yes, they should, but instead we're stuck with the stochastic-parrot crowd, who log onto HN and try their best to emulate a stochastic parrot.


LLMs don't use tools. Systems that contain LLMs are programmed to use tools under certain circumstances.

you’re just abstracting it away into this new “systems” definition

when someone says LLMs today they obviously mean software that does more than just text, if you want to be extra pedantic you can even say LLMs by themselves can’t even geenrate text since they are just model files if you don’t add them to a “system” that makes use of that model files, doh


This doesn’t address the author’s point about novelty at all. You don’t need 100% accuracy to have the capability to solve novel problems.

It does address the GP comment about math.

I thought it might do better if I asked it to do long-form multiplication specifically rather than trying to vomit out an answer without any intermediate tokens. But surprisingly, I found it doesn't do much better.

Other comments indicate that asking it to do long multiplication does work, but the varying results makes sense: LLMs are probabilistic, you probably rolled an unlikely result.

Specifically, you need to use a reasoning model. Applying more test time compute is analogous to Kahneman's System 2 thinking, while directly taking the first output of an LLM is analogous to System 1.

This is true for solving difficult novel problems as well, with the addition of tools that an agent can use to research the problem autonomously.


This hasn't been true for a while now.

I asked Gemini 3 Thinking to compute the multiplication "by hand." It showed its work and checked its answer by casting out nines and then by asking Python.

Sonnet 4.6 with Extended Thinking on also computed it correctly with the same prompt.


Pijul isn't a CRDT is it? It's theory of patches (i.e. DARCS++) alongside native conflicts.

Its author says it implements a CRDT in its theory documentation.

I generally use flatpak for things that are important to keep extremely updated, e.g. my browser for vulnerability reasons.

I can completely understand how you were driven away. If you ever want to give it a go again:

> there's "Flakes" which I never quite understood

Nix never clicked for me until I started using flakes. There's a lot of internal drama surrounding them that honestly childish; that's why they are marked as experimental and not the official recommendation. You are going to have a worse time with Nix if you go with the official recommendation, flakes are significantly more intuitive. The Determinate Systems installer enables them by default, and whatever documentation they have is on the happier path (except for FlakeHub, I haven't figured that one out yet).

On the most fundamental level, flakes allow you to take /etc/nixos/nixos.nix (or whatever, it has been forever) out of /etc and into a git repository. Old-style nix may be able to do that, but I discovered flakes before trying. I did previously attempt to use git on /etc/nix, but git was falling to pieces with bizarre ownership problems.

What this means is that I could install and completely configure a machine, once booted into a nix iso, by running: nixos-install --flake https://github.com/.../repo.git. I manage all of my system config out of /home/$user/$clone

As for /home there is home-manager and, again, you are not steered towards it (the tutorial pushes you towards nix profiles/nix-env instead). Home-manager will do for your home directory what the system config does for your system, and has many program modules. You can even declare home-level systemd units and whatnot.

> manually edited /etc files.

You can use environment.etc for these files[1]. systemd.tmpfiles can be used for things outside of etc. Home-manager has the equivalent for .config, .local, .cache. [2].

[1]: https://search.nixos.org/options?channel=unstable&query=envi... [2]: https://home-manager-options.extranix.com/?query=xdg.configF...


Yep, i am doing the same. I have a central remote flake repo where all my machines, services, etc are defined and they all run tweaked autoupdaters to periodically do full updates. I push commits and wait and forget. It feels like maintaining your distro everywhere, no matter where you ssh in. And soon, i will migrate that repo off a central platform (github) into radicle or something and turn some of my machines into seeders. Then, with offsite data backups, my house could burn down and github go dark, i could still recover, maybe in the future even bootstrap from my smartphone. A big step towards digital sovereignity.

Great comment -- thank you!

> services.desktopManager.gnome.extraGSettingsOverrides =

You can set dconf settings more declaratively: https://tangled.org/jonathan.dickinson.id/nix/blob/7c895ada8...


Give the kid a device that is age-controlled. No need for all devices to support it.

the law is there because parents are fucking clueless unprincipled whining crybabies, who need a lot of support, and sometimes that includes a bit of pushing ...

or who knows what problem is this supposed to fix. orphans buying phones? kids buying secret phones behind their parents back?


> [Article] What works in GNOME might not work in KDE. What works in both might not work in Sway.

If you subtract GNOME from the set then things become a lot more sane. "Compositor-specific extensions" are really "everyone besides GNOME extensions." The system tray extension isn't KDE-specific. Sure, window positions might not be available at all (because they don't make sense for a TWM), or a user might not have a system tray bar (or you might be on GNOME). However, if they did have a system tray it would be the StatusNotifierItem protocol. Ideally, these should be handled like other platform features like accelerometers etc.. That may not be possible, either way a lot of them can safely noop.

> [Article] For Avalonia, this means "Wayland support" isn't one implementation, it's potentially dozens. We're not just writing a Wayland backend; we're writing a GNOME-Wayland backend, a KDE-Wayland backend, a Sway-Wayland backend,

If you're making per-WM backends then you've fundamentally misunderstood how extensions are supposed to work. Other Wayland client libraries do not have a independent backends for KDE, Sway, and GNOME. Maybe quirks would be needed because you're attempting to support an existing UI library - but those should be few and far between.

IIRC Avalonia supports Vulkan as a rendering backend? Wayland protocols are the same line of thinking as Vulkan extensions.

wlroot and smithay are good examples of what extensions are used in the real world.


My experience was the same while helping to adapt a Steam Deck game for wider Linux support. The issue wasn't Waylandisms, most of those have already by figured out. It was GNOME. Their preferred resolution to issues seems to be dropping support rather than bug fixes, and they go out of their way to adopt implementations that are against the momentum of the wider community. I can get why they make some of their decisions, but things like killing the tray indicator or server side decorations are insane. To be an outlier in name of a greater or grander goal is one thing, then there is whatever GNOME is doing.

Server Side Decorations are a wayland extension though. Its not in the baseline spec.

Fwiw on Vulkan or GL it also often makes sense to implement entirely different code paths for different *sets* of supported extensions, instead of handling all possible extension combinations in a single code path (which leads to messy code and a combinatorial explosion of test cases, since each unique extension might theoretically be supported or unsupported independently from all other extensions - but you need to test each combination).

The whole idea of granular and independent extensions is pretty stupid across GL, Vulkan and Wayland. It makes more sense to have a handful "tiers" or "profiles" which guarantee a specific set of features (eg how all other 3D APIs do it) - e.g. Wayland should have a "desktop profile" instead of dozens of optional extensions needed for desktop scenarios, other profiles could be "mobile" and "kiosk".


Despite how much they would have you believe it, human rights are not a political issue. Politics are used to expand practiced rights (or abused to reduce them), just like politics are involved with providing you access to water.

This is just structured logging. Slog in go, tracing in rust, serilog in C#, etc. You should be combining this with the likes of otel or datadog, which will keep those fields structured and filterable.

Micro-orms (mapping to parameters and from columns) are generally fine last i read. It is the general aversion to writing a SELECT that is suspect.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: