kersplomp

I didn’t mean my opportunity cost from the toilet.

I did, that’s literally what I’ve been talking about this whole time. I’m asking you to stop wasting yours and my time.

Neither did chatgpt, the post is fake. See the other comments, people have verified

I see, thanks for clarifying. If you’re arguing that PRNG is not random, then you’re likely confusing non-technical readers. Additionally, it is an implementation detail whether it’s pseudorandom or actually random since /dev/random takes in actual random signals like network packets.

If it used a seeded PRNG it’s repeatable, but repeatability does not imply predictability which is what a non-technical reader might assume. Remember, most people on here are non-technical.

re: the kv cache thing, I don’t think that’s correct but I don’t have the energy to prove it sorry. shared kv cache sounds like a security nightmare but ymmv

Honest question, you used the word opportunity cost, but do you not understand what that is? With an opportunity cost, you compare two possible actions that you can take. It doesn’t make sense to compare your action (supporting those that propagate misinformation) to the actions of an entire economy (investing in ai infrastructure), because the actor is not the same in both.

With an opportunity cost, you can only compare the actions that you can take, and you alone. I listed a few comparable actions in my previous comment: research climate change, research the climate effects of ai, post about factory farms, read up on solar, invite others to donate money to clean water causes. These can all be done on the toilet. Any of these is a better use of your time than joining a mob of people against a cause you don’t even fully understand.

Almost all clients do some random sampling after softmax using temperature. I’m confused why someone who knows about kv caching would not know about temperature. Also shared kv cache while plausible is not standard in open source as of a year or so ago, so i’m curious what you are basing this off of. Did I miss a research paper?

This doesn’t seem real, have any of you actually tried this?

And you as a commenter have an opportunity cost too. You could be spending your time raising money for PETA or Oxfam, or researching climate change, or commenting about the effects of factory farms, but instead you’re just griping about the current popular thing to gripe about.

I donate 100% of my salary (I’m retired but still work) to fighting income inequality, climate change, animal abuse, and transphobia. It’s so frustrating to see people waste their time hating on things they don’t even understand just to fit in. Maybe this doesn’t describe everyone commenting, but if it does maybe they should get off their high horse. Sometimes the best thing a person can do for the world is shut up and give space for an actual expert to talk.

Please people, you need to stop spreading this misinformation.

AI is not killing the environment. If you track the sources of these claims, you will find that they first were spread by McKinsey and Bloomberg, who have a vested interest in publicly traded oil companies and other polluting corporations. They love to spread this misinformation because it distracts from the REAL environmental harms being caused by fossil fuels, meat farming, and concrete production. See drawdown.org for the specific numbers.

We need to stay focused on climate change and not get distracted. Our efforts should be focused on stopping new coal plants and factory farms. Datacenters, and especially the one in indianapolis which wouldn’t even have used water for cooling, have minimal environmental impact compared to trump’s coal, oil, and farming policies which will kill tens of millions in the long run and have already sparked wars.

We can fix climate change. We were so close to replacing fossil fuels that oil companies got scared; we can’t afford to give up at the last mile.

This reads like an ad for claude

Ah if only I had self control

No I mean it’s literally AI generated. I am super confident. I work on these models on a daily basis so I know it when I see it. Sorry that doesn’t fit into your preconceived worldview.

You write with too much confidence, I think. I work on these models professionally so I don’t know what to tell you besides your surface level understanding of LLMs seems wrong. 🤷‍♀️

Models do use emdashes disproportionately often, along with shorter sentences than the average human, because they are trained with RLHF where humans rate the model outputs based on their preferences. This is different than copying human behavior as you suggested they do. With RLHF (and later more efficient methods), humans tended to rate models that used emdashes and that had shorter, matter of fact sentences higher because it projected more confidence. In fact, models that were confidently wrong were more likely to be rated higher than models that were unconfidently correct.

Sources cited:

• https://surgehq.ai/blog/lmarena-is-a-plague-on-ai
• https://ai.stackexchange.com/questions/49107/why-do-ai-language-models-overuse-em-dashes-compared-to-human-writers
• https://en.wikipedia.org/wiki/Reinforcement_learning_from_human_feedback

Respectfully, if you do the stereotypical armchair expert thing and double down, I will block you. I don’t have the patience to deal with people incorrectly mansplaining my own field to me. I just want to spread knowledge.

AI detectors don’t work on the latest generation, they’re all trained on older open source models. You can usually tell AI generated stuff by the em dashes, the tone, and the contrast-driven sentence structure.

E.g. the most AI generated sentence is “That’s not being rude—that’s self-care .”

This one looks like Grok. It has a tendency to generate short sentences and pithy one liners like “Let that sink in.”

It’s a nice story and all but it’s clearly AI generated and I’m not a fan of that

This doesn’t surprise me. I know a gal who reported retaliation at Google and was immediately put on a secret HR blacklist that prevented her from getting above a certain perf rating. She was later “laid off” without cause.

She recorded her chat with HR, and this is a direct quote from the HR person: “Oh, that’s not retaliation. That’s a very specific legal construct. It sounds like what your manager is doing is more like retribution.”

It’s probably not true if that makes you feel better. futurism is usually fast and loose with the articles 🤷‍♀️

you seem to be under the wrong impression that “random dice roll” == “random dice roll from a uniform distribution”.

Almost all dice are uniformly random. Unless you’re using weighted dice? Which, seeing how defensive you get when wrong, might actually make sense 🤔

Knowledge isn’t a competition. Nobody cares about any of this. We will all die and be forgotten. You’re on an internet forum with a bunch of people who have no idea who you are and who could care less about your knowledge of statistics.

Respectfully blocked. I have better things to do with my time.

Apologies for the late reply, but it turns out I can’t let that sit. Sorry for the rant, but I work in RL and saying “it’s just dice rolls” is insulting to my entire line of work. :(

A probability distribution is not the same as random dice roll. Dice rolls are uniformly and independently random, whereas the probability distributions for LLMs are conditional on the context and the model’s learned parameters. Additionally, all modern LLMs use top K and p sampling–which filters the probability distribution to only high confidence words–so the probability of it choosing to say random garbage is exactly zero.

The issues with LLMs have nothing to do with their sampling from random distributions. That’s just a minor part of their training, and some LLMs don’t even do random sampling since they use tree search. The issues with LLMs are the result of people trying to teach it intelligence using behavior cloning on a corpus of human words and images. Words can’t encode wisdom, only knowledge. Wisdom can only be gained through lived experience.

How well do you think you would perform if you were born into a cave, forced to read a thousand dictionaries in order with no context, and then your only interaction with the outside world was a single question from a single human, and then you died? If you ask me, the LLMs are doing suprisingly well given their “lived experiences”.

Is decomp.dev supposed to be down?

PSA: Don’t buy a gaming laptop. They are trash. The plastic case will melt, the wifi card will come loose, the battery will die within minutes. A steam deck is truly your best option.

And never ever buy alienware. Screw them in particular.