Live data from Hacker News

The "confident idiot" problem: Why AI needs hard rules, not vibe checks

steerlabs.substack.com

331–340 of 399 posts

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#331

Confident idiot: I’m exploring using LLM for diagram creation. I’ve found after about 3 prompts to edit an image with Gemini, it will respond randomly with an entirely new image. Another quirk is it will respond “here’s the image with those edits” with no edits made. It’s like a toaster that will catch on fire every eighth or ninth time. I am not sure how to mitigate this behavior. I think maybe an LLM as a judge ste…

Have you considered that perhaps such things simply are not within its capabilities?

I mean, one of its flagship features is to make precise edits to images. And it's really good at it... until it randomly isn't.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#332

Confident idiot: I’m exploring using LLM for diagram creation. I’ve found after about 3 prompts to edit an image with Gemini, it will respond randomly with an entirely new image. Another quirk is it will respond “here’s the image with those edits” with no edits made. It’s like a toaster that will catch on fire every eighth or ninth time. I am not sure how to mitigate this behavior. I think maybe an LLM as a judge ste…

Whats your thoughts on the diagram as code movement? I'd prefer to have an LLM utilize those as it can atleast drive some determinism through it rather than deal with the slippery layer that is prompt control for visual LLMs.

I think that's the right approach and what I've been experimenting with. Diagram as code and then style transfer from output diagram to desired look. That's where I've had the most success.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#333

Confident idiot: I’m exploring using LLM for diagram creation. I’ve found after about 3 prompts to edit an image with Gemini, it will respond randomly with an entirely new image. Another quirk is it will respond “here’s the image with those edits” with no edits made. It’s like a toaster that will catch on fire every eighth or ninth time. I am not sure how to mitigate this behavior. I think maybe an LLM as a judge ste…

I had a similar result trying to create 16 similarly styled images. After half a dozen it just started kicking out the same image over and over again no matter what the prompt said. Even the “thinking” looked right, but the image was just a repeat. I don’t know if this is some type of context limitation or what. I got around it by using a new prompt/context for each image. This required some rethinking about how to m…

Oh smart. This is good guidance. Yeah fascinating how longer running context causes these side effects, especially the repeated image with no changes bug.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#334

We are trying to fix probability with more probability. That is a losing game. Thanks for pointing out the elephant in the room with LLMs. The basic design is non-deterministic. Trying to extract "facts" or "truth" or "accuracy" is an exercise in futility.

I couldn't agree with you more. I really do find it puzzling so many on HN are convinced LLM's reason or think and continue to entertain this line of reasoning. At the same time also somehow knowing what precisely the brain/mind does and constantly using CS language to provide correspondences where there are none. The simplest example being that LLM's somehow function in a similar fashion to human brains. They catego…

> The simplest example being that LLM's somehow function in a similar fashion to human brains. They categorically do not. I do not have most all of human literary output in my head and yet I can coherently write this sentence.

The ratio of cognition to knowledge is much higher in humans that LLMs. That is for sure. It is improving in LLMs, particularly small distillations of large models.

A lot of where the discussion gets hung up on is just words. I just used "knowledge" to mean ability to recall and recite a wide range of fasts. And "cognition" to mean the ability to generalize, notice novel patterns and execute algorithms.

> They don't actually understand anything about what they output. It's just text.

In the case of number multiplication, a bunch of papers have shown that the correct algorithm for the first and last digits of the number are embedded into the model weights. I think that counts as "understanding"; most humans I have talked to do not have that understanding of numbers.

> It's just an algorithm.

> I am surprised so many in the HN community have so quickly taken to assuming as fact that LLM's think or reason. Even anthropomorphising LLM's to this end.

I don't think something being an algorithm means it can't reason, know or understand. I can come up with perfectly rigorous definitions of those words that wouldn't be objectionable to almost anyone from 2010, but would be passed by current LLMs.

I have found anthropomorphizing LLMs to be a reasonably practical way to leverage the human skill of empathy to predict LLM performance. Treating them solely as text predictors doesn't offer any similar prediction; it is simply too complex to fit into a human mind. Paying a lot of attention to benchmarks, papers, and personal experimentation can give you enough data to make predictions from data, but it is limited to current models, is a lot of work, and isn't much more accurate than anthropomorphization.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#335

We are trying to fix probability with more probability. That is a losing game. Thanks for pointing out the elephant in the room with LLMs. The basic design is non-deterministic. Trying to extract "facts" or "truth" or "accuracy" is an exercise in futility.

I couldn't agree with you more. I really do find it puzzling so many on HN are convinced LLM's reason or think and continue to entertain this line of reasoning. At the same time also somehow knowing what precisely the brain/mind does and constantly using CS language to provide correspondences where there are none. The simplest example being that LLM's somehow function in a similar fashion to human brains. They catego…

I have had conversations at work, with people who I have reason to believe are smart and critical, in which they made the claim that humans and AI basically learn in the same way. My response to them, as to anyone that makes this claim, is that the amount of data ingested by someone with severe sensory dysfunction of one sort or another is very small. Helen Keller is the obvious extreme example, but even a person who is simply blind is limited to the bandwidth of their hearing.

And yet, nobody would argue that a blind person is any less intelligent that a sighted person. And so the amount of data a human ingests is not correlated with intelligence. Intelligence is something else.

When LLMs were first proposed as useful tools for examining data and proving answers to questions, I wondered to myself how they would solve the problem of there being no a-priori knowledge of truth in the models. How they would find a way of sifting their terabytes of training data so that the models learnt only true things.

Imagine my surprise that not only did they not attempt to do this, but most people did not appear to understand that this was a fundamental and unsolvable problem at the heart of every LLM that exists anywhere. That LLMs, without this knowledge, are just random answer generators. Many, many years ago I wrote a fun little Markov-chain generator I called "Talkback", that you could feed a short story to and then have a chat with. It enjoyed brief popularity at the University I attended, you could ask it questions and it would sort-of answer. Nobody, least of all myself, imagined that the essential unachievable idea - "feed in enough text and it'll become human" - would actually be a real idea in real people's heads.

This part of your answer though;

"My paper and pen version of the latest LLM .... My paper and pen version of the latest LLM"

Is just a variation of the Chinese Room argument, and I don't think it holds water by itself. It's not that it's just an algorithm, it's that learning anything usefully correct from the entire corpus of human literary output by itself is fundamentally impossible.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#336
post #40

Earlier quoted context omitted.

I find it amusing that once you try to take LLMs and do productive work with them either this problem trips you up constantly OR the LLM ends up becoming a shallow UI over an existing app (not necessarily better, just different).

The UI of the Internet (search) has recently gotten quite bad. In this light it is pretty obvious why Google is working heavily on these models. I fully expect local modes to eat up most other LLM applications—there’s no reason for your chat buddy or timer setter to reach out to the internet, but LLMs are pretty good at vibes based search, and that will always require looking at a bunch of websites, so it should slot…

The reason search got so bad, even pretending google themselves are some beneficial actors, is because it is a directly adversarial process. It is profitable to be higher in search results than you "naturally" would be, so of course people attack it.

Google's entire theory of founding was that you could do better than Yahoo hand picking websites with an algorithm, and pagerank was the demonstration, but IMO that was only possible with a dataset that was non-adversarial because you couldn't "attack" yahoo and friend's processes from the data itself.

The moment that changed, the moment pagerank was used in production, the game was up. As long as you try to use content to judge search ranking, content will be changed, modified, abused, cheated to increase your search rank.

The very moment it becomes profitable to do the same for LLM "search", it will happen. LLMs are rather vulnerable to "attack", and will run into the exact same adversarial environment that nullified the effectiveness of pagerank.

This is orthogonal also to if you believe Google let search be shittier to improve their ad empire. LLM "search" will have exactly this same problem if you believe it exists.

If you build a credit card fraud model on a dataset that contains no attacks, you will build a rather bad fraud model. The same is true of pagerank and algorithmic search.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#337

Earlier quoted context omitted.

I suspect a lot of the em-dash usage also comes from transcriptions of verbal media. In the spoken word, people use the kinds of asides that elicit an em-dash a lot.

I would bet like a dollar that the supposed em-dash usage (which I'm not convinced is an accurate take in the first place) would have come from an enterprising dev somewhere being like "Well, we probably don't need multiple tokens for hyphens" and coercing every dash type thing to just one hyphen like token. But I'm also showing off my ignorance with how these machines turn text into tokens in practice.

I think all the em-dashes came from scraping Wordpress blogs. Wordpress editor does "typography", then thus introduced em-dashes survive HTML to Markdown process used to scrap them, and end up in datasets.

EDIT: Also PDFs authored in MS Word.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#338

> We are trying to fix probability with more probability. That is a losing game. > The next time the agent runs, that rule is injected into its context. It essentially allows me to “Patch” the model’s behavior without rewriting my prompt templates or redeploying code. Must be satire, right?

The first thing I do on Hacker News when there's an AI post is run to the comments for a good time. The later I go back and read the actual article, and in this case hoo boy what a doozy. An AI-written summary of a seemingly not vibe-coded python library written by a human being who apparently genuinely believes that you can fix LLM hallucinations with enough Regular Expressions.

It would be magnificent if this is satire. Wonderful.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#339

Earlier quoted context omitted.

I disagree. You're still human if you're deaf and mute. Our intellectual processing powers, or of animals for that matter, has nothing to do with language.

Being deaf and mute doesn't imply lack of language. But being unable to communicate absolutely strikes me as non-human.

Ok say you grew up alone in the woods, are you no longer human? The capability to learn language is no doubt unique, but language itself isn't the basis of intelligence.

Re: The "confident idiot" problem: Why AI needs hard rules, not vibe checks

#340

Earlier quoted context omitted.

No, you cannot. Our abstract language abilities (especially the written word part) are a very thin layer on top of hundreds of millions of years of evolution in an information dense environment.

Sure, but language is the only thing that meaningfully separates us from other great apes

Not it isn't most animals also have a language and humans do way more things differently, than just speak.
Post reply on HN