Message Stream - Atacama

runaway train LLM

2025-04-29 17:42:22

Trying the qwen3-1.7b-q6_k.gguf model ⚙️ ( QWEN3, 1.7B parameters, quantized to 6-bit) , there is a clear problem of runaway-thought (or at least looping responses).

Thought process: Okay, let's see. The question is asking for the antonym of 'reluctant' from the given candidates: eager, hesitant, unwilling, disinclined, averse, loath.

First, I need to recall what each of these words means. 'Reluctant' means feeling reluctance or hesitation to do something. So the antonym would be something that shows no hesitation or eagerness.

Looking at the options:

- Eager: This directly means willing or enthusiastic, so it's the opposite of reluctant. But wait, let me check others too.

- Hesitant: That's more about being unsure, but maybe not exactly the same as reluctant. Reluctant is more about reluctance, while hesitant is about uncertainty. Maybe they are similar but not exact antonyms.

- Unwilling: This means not willing to do something, which is directly opposite of eager. But 'unwilling' might be a bit stronger than 'reluctant'.

- Disinclined: This means not inclined to do something, so it's similar to reluctant. But maybe 'disinclined' is more formal.

- Averse: This means strongly opposed or unwilling. So 'averse' would mean very reluctant, but the antonym would be... Wait, if the word is 'averse', then its antonym would be 'eager'? Or maybe 'eager' is the opposite of 'reluctant'.

Wait, the question is about the antonym of 'reluctant'. So looking for a word that means not reluctant. 'Eager' is the direct opposite. But let me check other options.

'Unwilling' and 'averse' both mean not willing. However, 'averse' might be more about strong opposition, while 'unwilling' is more about lack of will. But in some contexts, they can be used interchangeably.

But the correct antonym of 'reluctant' is usually considered to be 'eager' or 'not reluctant'. However, among the given options, 'eager' is a direct opposite. But wait, let me think again. The word 'reluctant' can also be synonyms with 'hesitant', 'averse', etc. So the antonym would be the one that's not hesitant, not unwilling, etc.

Wait, but the options are: eager, hesitant, unwilling, disinclined, averse, loath.

So which of these is the antonym? Let me check each:

- Eager: opposite of reluctant. Yes.

- Hesitant: similar to reluctant, but maybe not exactly. If 'hesitant' is considered an antonym, then it's not. But I think 'hesitant' is a synonym for reluctant in some contexts.

- Unwilling: Direct opposite. So if someone is unwilling, they are not reluctant.

- Disinclined: Also means not inclined, so similar to reluctant.

- Averse: Strongly opposed, which is more than just reluctant.

- Loath: This means very angry or resentful, but can also mean being against something. However, 'loath' might be a bit stronger than 'reluctant'.

But the question is asking for the antonym of 'reluctant'. The most direct antonym would be 'eager', because eager means willing to do something, which is opposite of reluctant. But wait, 'eager' can sometimes be used in a way that's more about being enthusiastic rather than just not reluctant.

Alternatively, 'unwilling' is also an antonym. But 'unwilling' is more about not wanting to do it, while 'reluctant' is about hesitation. So which one is the better answer?

In standard dictionaries, the antonyms of 'reluctant' are often listed as 'eager', 'inclined', 'disposed', etc. However, among the given options, 'eager' is present. But 'unwilling' and 'averse' are also possible.

But let's check the options again. The candidates are: eager, hesitant, unwilling, disinclined, averse, loath.

So between these, which is the best antonym? Let's think of each:

- Eager: directly opposite of reluctant.

- Hesitant: similar to reluctant.

- Unwilling: opposite of eager?

Wait, 'unwilling' is the opposite of 'eager'? No. Wait, if someone is eager, they are willing. If they're unwilling, they are not willing. So 'unwilling' is the antonym of 'eager'? Or is it the opposite of 'reluctant'?

Wait, maybe I'm confusing words here. Let me think again.

The word 'reluctant' means not eager to do something. So the antonym would be 'eager'. But if someone is 'unwilling', that's also a synonym for reluctant? Or is it different?

Actually, 'unwilling' can mean not willing, which is similar to 'reluctant', but sometimes used in different contexts.

But according to standard definitions, the antonym of 'reluctant' is 'eager'. However, among the given options, 'eager' is one of them. But let me check if other words are better.

Wait, 'averse' is another option. 'Averse' means strongly opposed or unwilling. So 'averse' could be an antonym of 'reluctant' in some contexts. But which is more accurate?

In the given options, the best answer would be 'eager', but maybe 'unwilling' is also correct.

But I need to choose from the provided candidates. Let's check each:

- Eager: direct opposite.

- Hesitant: synonym.

- Unwilling: antonym of eager? Or antonym of reluctant?

If someone is unwilling, they are not willing, which is similar to being reluctant. So maybe 'unwilling' is a better antonym than 'eager'? But that's conflicting.

Alternatively, perhaps the correct answer is 'eager' because it's the direct opposite. However, some might argue that 'unwilling' is also an antonym.

Wait, but looking at the options again: the user provided these words as candidates. The question is to pick the antonym of 'reluctant' among these. So which one is correct?

In standard English, the antonyms of 'reluctant' are 'eager', 'inclined', 'disposed', etc. But since the options include 'eager' and 'unwilling', I need to see which is more accurate.

But in some contexts, 'unwilling' can mean not eager. For example, if someone is unwilling to do something, they are not eager to do it. So maybe 'unwilling' is a better antonym here.

Wait, but 'reluctant' and 'unwilling' are both about not wanting to do something. However, 'reluctant' is more about hesitation or reluctance, while 'unwilling' is about lack of will.

So which one is the correct antonym? Let me check a dictionary. For example, according to Merriam-Webster, the antonyms of reluctant are eager, inclined, disposed, etc. So 'eager' is an antonym. However, 'unwilling' is also considered an antonym in some contexts.

But given that the options include both 'eager' and 'unwilling', which one is the correct answer here?

The question is from a test or a quiz, so perhaps the intended answer is 'eager' because it's directly opposite. However, maybe 'unwilling' is also acceptable.

Alternatively, looking at the other options: 'averse' and 'disinclined' are similar to 'reluctant'. So they are not antonyms but synonyms.

So between 'eager' and 'unwilling', which one is correct?

In some contexts, 'unwilling' can be considered an antonym of 'reluctant', but I think the more precise answer is 'eager'.

But wait, let's check another angle. The word 'reluctant' can also mean not willing to do something, so 'unwilling' would be a direct antonym. But maybe in this case, the best answer is 'eager' because it's the most direct opposite.

Alternatively, if the options are limited, and given that 'eager' is present, perhaps that's the intended answer.

It is a type of thought, almost. Unfortunately, it is not a particularly healthy type of thought.

🔥 it almost certainly did not consult the Merriam-Webster dictionary.

more "fake thought" at the margins LLM

2025-04-29 03:48:12

QWEN3 is out. I'm starting tests with the 4B model.

The task: which of these 10 words has a given definition. It does some thought before picking an answer:

Okay, let's see. The question is asking for a word that means a feeling of great pride and honor. The choices are breeze, fragile, glory, iceberg, magic, outcast, rescue, savage, skeptic, wonder.

First, I need to recall each of these words and their meanings. Let's go through them one by one.

Breeze: that's a light wind, right? So that's related to weather, not emotions. Probably not the answer.

Fragile: meaning easily broken or damaged. That's more about being weak or delicate. Not related to pride or honor.

Glory: Hmm, I think glory has to do with something that's impressive or magnificent. Like national glory or glory of a victory. That sounds like it could be the answer. But let me check the others.

Iceberg: that's a large mass of frozen water. Not related to emotions or feelings. So probably not.

Magic: that's the ability to do things that seem impossible, or something supernatural. Not directly related to pride or honor, but maybe in some contexts? Not sure.

Outcast: someone who is excluded from a group. That's more about being rejected, not feeling pride.

Rescue: the act of saving someone from danger. That's an action, not a feeling.

Savage: brutal or wild, not civilized. Maybe related to harshness, not pride.

Skeptic: someone who doubts things. Not related to feeling pride.

Wonder: a feeling of amazement or admiration. That's a feeling, but more about being amazed rather than pride and honor.

So, the options are glory and wonder. The question is about pride and honor. Glory is often associated with achievements that bring honor, like national glory. Wonder is more about amazement. So I think the answer is glory.

💡 Some of the notes, like Breeze: that's a light wind, right?, are not thinking, but the illusion of thought.

In any event, the approach is (mostly) correct: a good intermediate step is to write out the definitions of all ten words.

⚔️ well, actually, this takes ~20 seconds; which is something like 50 times what some of the smaller models do for this task.

💡 as always: cost is a factor.

garden update: 2025-04-28 Gardening

2025-04-29 01:42:29

The garden is mostly planted at this point. ⚔️ ( well, actually, the tomatoes and peppers haven't been transplanted yet. but they are planted. just inside.)

"West" garden - corn. 30-40 stalks? Hopefully enough plants and sun to get a crop.
Raised garden bed - cucumbers, zucchini, and "lemon cucumber"s. Old carrot seeds in between. And, later, the tomatoes and peppers.
By east fence: one zucchini and one cucumber plant.

💭 I have tried strawberries two or three times; they always get eaten or otherwise die before achieving health.

The flower garden (by the deck) still needs dahlias planted. Perhaps this coming weekend.

edgeley (part 3) Cities

2025-04-28 17:04:56

Last night, Claude wrote some code for the qualification metrics. 💡 ( which might now be called exemplar tasks. The metric is "respond to one prompt".) 🔥 ( to some degree, the goal is to test "changes in context" as much as "changes in model")

An earlier task (from late 2023) was to answer the question: Who was Pablo Arosemena? ⚙️ ( the Wikipedia article Pablo Arosemena is about an obscure politician from Panama)

The 8b models don't know who this is. But, they most commonly think he is an obscure painter. 🔥 ( probably because of Pablo Picasso)

Is there some sense this is a true stereotype? Is it more likely he was a painter than that he was a baker, a masseuse, or a peasant farmer? ⚔️ ( well, actually ... it's more likely someone of this name had one of those jobs. But, less likely they were written about.)

I am getting a new computer. 🔥 ( "only" $600) This should allow better speed comparisons between the models. 💡 ( the inconveniences of having an external USB drive, power demands, and heat creation "on my lap" grew to be too much.)

I need to do one more "schema improvement" pass on the "dictionary".

Things like "Chinese translation", "Korean translation" need to be in a dataclass, rather than passed as parameters everywhere.

For now, I want the indexes (and explicit NULL values), so these are database columns, rather than an "all_translations" JSON blob.

Some of the benchmarks should be re-written once the "dictionary" API is available.

Claude invented "categories" for the benchmarks: "Language", "Reasoning", "Knowledge", and "Translation". These are ... decent.

But the zeroth category is token introspection. For "how many letters is the word triumphant" questions. 💡 ( even the "spell check" tests that require repeating a misspelled word are probably in this category) ⚙️ ( you can have an LLM without token introspection. but it should be very doable. possibly with some form of API / injection.)

Then, Language (starting with definitions and antonyms), Knowledge (starting with basic geography), and Translation (starting with EN-FR, EN-ZH, SW-KO - word-based).

The only "reasoning" task so far might be "unit conversion". But that would have a different name.

The "translation" tasks have to deal with the different vocabulary sizes of different languages.

This is one part technical vocabulary 💡 ( does Swahili have a word for capacitor that isn't a loan word?) and one part eccentric distinctions ⚙️ ( Chinese has 二 and 两 for "two")

So far, I have largely mitigated this problem by avoiding it.

The dictionary will need some type of "class" system. Specifically, I want to say "get a random animal" and have it do that.

I am putting that off as well. Largely because it is a morass of taxonomical hell that has stymied decades of efforts.

newell Cities

2025-04-27 22:00:37

🤖 A notable city in South Dakota associated with artisanal wheat farming is Newell. Located in the Belle Fourche River Valley of western South Dakota, Newell is home to Belle Valley Ancient Grains, an organic farm and flour mill specializing in ancient and heirloom wheat varieties such as einkorn, emmer, spelt, red fife, and white sonora. The farm, established by Brian and Linda Stambaugh in 2002, emphasizes sustainable practices, including crop rotation and the avoidance of synthetic fertilizers and pesticides.

An oscilloscope is an electronic test instrument that creates a visible two-dimensional graph of an electrical signal's voltage versus time. Think of it as a "heart monitor" for electronic signals - it lets you see the "pulse" of electricity in a circuit. Engineers, technicians, and scientists use oscilloscopes to troubleshoot electrical equipment, observe how signals change over time, and measure properties like frequency, amplitude, and distortion.

The device typically displays waveforms on a screen with a grid, allowing precise measurements. Modern digital oscilloscopes can capture one-time events, store data for later analysis, and even perform automatic measurements. They're essential tools in electronics labs, manufacturing facilities, and repair shops.

I find myself wanting a "Mac mini". 💡 ( So I don't have to run LLMs on my laptop.) 🔥 ( I wonder if I can use the Mac Mini with my AppleVision, without having another monitor ...)

Probably twice as the Macbook Air fast as what I have now. But, more importantly, "not running on my lap". So things like "I need to have an external hard drive attached" are not an inconvenience.

💡 This is not a "hoarding" purchase. This is a "use" purpose. It needs to be "fast" because I am doing real-time queries.

Horizontal scaling, often called "scaling out," is when you add more machines or instances to your system instead of making existing machines more powerful. It's like adding more attorneys to your law firm rather than asking each attorney to work more hours.

I find myself "buying a computer" for the first time in many years (non-Apple version). ⚔️ ( well, actually, the second time. I scouted them out once a few years ago.) If you want to buy them, there are two main choices: Apple or NVIDIA.

💡 This is why Apple and NVIDIA are the top two-valued companies.

🔥 my estimate is that total world compute is already enough to satisfy the world demand for AGI.

way mo' problems Tech

2025-04-24 21:08:46

https://www.fox7austin.com/news/waymo-driverless-cars-austin-slammed-complaints

The title is Waymo driverless cars in Austin slammed with complaints ... but a more accurate title would continue slammed with unfounded complaints by people who hate technology.

The two complaints in the article are "doesn't run red lights", and "pulls over when you push the pull over now button.

Their real complaint is some form of Google is evil, technology is evil, they are taking good jobs. But they can't say that. So they scour the ends of the earth looking for problematic experiences. And, once they find a few things which look bad if you squint, they get a friendly journalist to write an article about the number of complaints. 💡 ( which is based entirely on the number of people who want to complain, not the severity of the problematic actions)

edgeley (part 2) Cities

2025-04-24 20:00:24

so far today: running the "proficiency" benchmarks against GPT-4-1 and Gemini-2.5-flash.

The headline: Google's cheap model can count letters. Gemini was substantially slower than both OpenAI and Anthropic (but, perhaps, that can vary day-to-day). But it got 96% on the infamous count how many "R"s in strawberry metric, and none of the similarly-priced models got above 70%. 💡 ( the only metric it did "bad" on was the IPA one, and that is because the response normalization code is broken)

for pricing ⚙️ ( all prices per million tokens) :

GPT-4-1-nano: 10c IN, 40c OUT

GPT-4-1-mini: 40c IN, 160c OUT

GPT-4o-mini: 30c IN, 120c OUT

Gemini-2.5-flash: 15c IN, 60c OUT

Claude-3-5-haiku: 80c IN, 400c OUT

⚙️ Most of these have (or will have) "cache" discounts of 50-90% for repeated queries with the same long context.

💡 Claude is both the most expensive at this tier, and the lowest-performing. And the least-recently updated.

🔥 presumably, they will have a new model at half the price, next week.

the fall of General Electric Miscellany

2025-04-24 17:42:07

https://www.md-a.co/p/intellectual-laziness

The collapse of General Electric stands apart. GE was the bluest of the blue-chips: descended from Thomas Edison and J.P. Morgan, it was one of the original twelve components of the Dow in 1896, and grew to become one of the leading technology giants of the early 20th century. After WWII, GE evolved into an industrial behemoth with dominant positions in a dizzying array of electricity-adjacent markets, from jet engines and turbines to light bulbs and home appliances.

In the 1980s, GE ascended to new heights. Jack Welch took the reins as CEO in 1981, and he established GE a major player in media and financial services while reinforcing GE’s position in select attractive industrial markets. For most of the 1990s and 2000s, GE was the most valuable company in America, with a valuation topping out at over $1 trillion in 2023 dollars. While GE had some skeptics and critics at the time, it was typically seen as a corporate paragon, regularly named by Fortune as the most admired company in the world. Welch was regarded as a management guru, and his underlings were routinely poached to become CEOs at other Fortune 500 companies.

And then, a few years ago, it all unraveled in spectacular fashion. Much of the supposed success from the Welch era of the 1980s and 1990s proved to be illusory, the product of temporary tailwinds and aggressive accounting. GE’s fortunes worsened under the reign of Welch’s handpicked successor, Jeff Immelt, who took over in 2001. Immelt struggled to cope with the problems he inherited, which were compounded by the 2008 financial crisis and major missteps of his own. In 2017, when the extent of GE’s problems became clear, GE’s stock nose-dived, and Immelt was pushed out.

GE has been one of the worst performing mega-cap stocks of the modern era. A $1,000 investment in the S&P 500 in 2000 would be worth over $2,700 today (excluding dividends), while a $1,000 investment in GE in 2000 would have dwindled to only $210. Even going all the way back to Welch’s appointment in 1981, the S&P has outperformed GE by a three-to-one margin.

The reasons: style over substance, an inability to distinguish good-luck from a permanent advantage, a failed attempt to do financial services. In short, the intellectual laziness described in the title of the post.

writing with the LLM LLM

2025-04-22 18:09:16

To write a few paragraphs on a topic, there is (roughly) a four-step process:

Choose a topic
Create an argument
Choose the tone and style
Write the words

The machine is, in many situations, better than I am at the task of writing the words. 💡 ( it still struggles with a few tones) It is always about 20 times faster than me.

However, it struggles with the first three tasks.

The "agent" framework isn't focused on self-willed agents. This is (probably) a good thing. But "fiduciary" agents aren't happening either yet. These solve some of the problem related to motivation.
The machine prefers to create a good-sounding argument to a logically-sound one. 🔥 ( of course, many humans do the same thing) .
The machine defaults to a tone that has been over-used to the point of annoyance. It is obsequious and aggressively cheerful. In the context of "a chat-bot for people unfamiliar with the technology", this is a defensible choice. But, most of the time, you need your own tone. 💡 ( It has been long enough that the training data allows "write a tone prompt for this query" kind-of works.)

edgeley (part 1) Cities

2025-04-21 19:45:31

Edgeley, North Dakota, is a small rural town in LaMoure County, located in the southeastern part of the state. With a population hovering around 500 people, it's one of many prairie towns that exemplify the broader character of the upper Great Plains—quiet, sparsely populated, and closely tied to agriculture.

https://www.lesswrong.com/posts/bfHDoWLnBH9xR3YAK/ai-2027-is-a-bet-against-amdahl-s-law

Of course the post is right. The various FOOM claims are all bullshit. And Amdahl's Law is one of the reason why. Just because a few things will be a hundred times faster (or a million times faster) doesn't make the whole thing that much faster.

Also, AGI definitions vary so widely, from things that have already happened to things that are impossible, that a "prediction market" is nearly meaningless.

I have seen various commentary related to "Twilight of the Edgelords" ⚙️ ( https://www.astralcodexten.com/p/twilight-of-the-edgelords ) , a piece that I don't have access to.

And, the response I can piece together from the fragments I can see would fall under GUILD LAW. ⚙️ ( additional commentary at https://www.writingruxandrabio.com/p/the-edgelords-were-right-a-response and https://theahura.substack.com/p/contra-scott-and-rux-on-whos-to-blame )

https://developers.googleblog.com/en/gemma-3-quantized-aware-trained-state-of-the-art-ai-to-consumer-gpus/

To make Gemma 3 even more accessible, we are announcing new versions optimized with Quantization-Aware Training (QAT) that dramatically reduces memory requirements while maintaining high quality. This enables you to run powerful models like Gemma 3 27B locally on consumer-grade GPUs like the NVIDIA RTX 3090.

It seems pretty obvious. A majority of the users of open-source models are using quantized models on personal hardware; might as well optimize that use-case. 💡 ( it is less clear that a majority of the CPU cycles are there; but a majority of the people certainly are.)

My next round of updating the Greenland metrics will have to include the gemma3-12b-qat model. 💡 ( or, maybe the 27b. According to Hacker News, gemma3-27b-Q4 only uses ~22Gb (via Ollama) or ~15GB (MLX). On a 24GB machine, this clearly needs the non-Ollama approach.)

And, also, GPT-4.1 . And probably Gemini-2.5 . 💡 ( the goal for these models should be to perform at 100% accuracy.) ⚔️ ( well, actually, a few of the "correct" benchmark answers right now are incorrect.)