Notices by Karsten Johansson (ksaj@infosec.exchange), page 2

@larsbrinkhoff @amszmidt @loke Is there a barnson or dotterbarn? If not, there should be. 🎯

@amszmidt @loke The British sometimes mimick the Swedish way. For example, "Me mum's mum."

@amszmidt @osi @fsfe @fsf Thanks, I think. :ablobdj:

When I initially read about this, I honestly thought I'd read it wrong. Open Source, but with secret ballots. Reminds me of how Russia votes.

PS: The list of signees on the petition is compelling. I'm not an OSI member or I would certainly sign.

@amszmidt @osi @fsfe @fsf Oops, wrong thread. I'm almost as inept as a Trump hire sometimes. :ablobspin:

@amszmidt @osi @fsfe @fsf Trump really hates kids, and health science. That's two dings against children with cancer in such short time. Remember he deported a child with stage 4 cancer.

@amszmidt George Takei refers to them as such. And he would know - he and his family spent time in a US American concentration camp for the crime of being Japanese.

@amszmidt Even if you make a small model (kinda goes against the name, but you certainly can do it) it is not as simple as just Markov chains.

Just because it has the surface appearance doesn't mean its the same thing. A duck has legs. So do humans. Therefore humans fly?

@amszmidt If you try to explain llms in a Markov chain context, an LLM is very complex Markov chain where the "states" are words or tokens, and the "transitions" are the probabilities of predicting the next word based on the current state and context.

Note how many qualifiers take it far and beyond a Markov chain. There is a cms deep kms wide similarity only.

One could argue that a person editing a document is doing exactly this.

@amszmidt Every one that dropped. They'll have forgotten in a day or three.

@screwtape @khleedril @amszmidt I'm definitely doing something annoying. Quite a few annoying things. The code works, but only by the grace of God. I commented the less obvious things, but this code would or should get you banished from the Lisp kingdom.

@screwtape @khleedril @amszmidt Would be, it could be. But really shouldn't be. 😉

@khleedril @amszmidt @screwtape lol it's legit common lisp as long as you don't use clisp -modern.

@amszmidt @screwtape Oh it works. It just breaks all the "rules"

Lisp sucks. Tell me what's wrong here:

(DEFVAR x 0) ; GLOBAL BADNESS STAYS

(DEFUN dO-StuFf (NUMBer)
(LET ((rEsUlT 1))
(DECLARE (SPECIAL X)) ; THIS IS LEGAL HERE
(SETQ X NUMBER)
(TAGBODY
StaRT
(IF (< X 1) (GO ENDz))
(SETQ rEsUlT (* rEsUlT X))
(SETQ x (- x 1)) ; NOTE: STILL USING GLOBAL x INSTEAD OF LOCAL X
(GO StARt)
ENDz)
(pRiNt (lIsT 'FACtoRiAl-oF NUMBER 'iS rEsUlT))))

(DEFUN nOW-dO-AlL ()
(LET ((I 0))
(LOOP
(WHEN (> i 5) (RETURN))
(Do-sTUFF i)
(SETQ i (+ i 1)))))

#commonlisp #lisp #sbcl #clisp @amszmidt @screwtape

@amszmidt @screwtape I was expecting NIL, but good enough.

You've seen my code posted here. I 'm all downcase unless it's sofa king important someone might spam me about missing that golden opportunity lol.

Allcaps is just wayyy too old fashioned for me. If your screen isn't monochrome, it's probably not needed anymore.

But don't get me wrong. Anything that make your code easy to read (in the human sense) is fine by me.

@amszmidt @screwtape (REALLYP)

People have low expectations for AI. I asked ChatGPT (paid account, not the free stuff) what would make ChatGPT better. I think the answer is a lot higher quality than what people keep saying one should expect. This is what I got:

--- My Prompt:
With DeepSeek, there are things it is very good at, but it fails quickly and rather spectacularly. Your models tend to break down as well, but somewhere there must be a model that will be able to stay on topic, even when the topic changes, so that conversations move around like human conversations move around subject-to-subject. What would take ChatGPT over the top?

--- ChatGPT reply:
Here’s what it’ll take to get closer to that human-like conversation flow:

1. Long-Term Memory That Actually Works. Most models either have no memory or they remember only in a clunky, artificial way.

A better memory system would not just recall facts, but contextual relevance—like “we talked about LLM efficiency two days ago and you hate cheerleader language.”

It has to be smart recall, not just data storage.

1. Thread Awareness
   Imagine a model that can say: “You mentioned DeepSeek earlier. Should I compare it again now that we’ve brought up open-source tools?”
   That requires models to track not just words, but conversational branches—like a tree, not a straight line.

2. Graceful Fallbacks
   When a model doesn’t know, it should gracefully pivot, not hallucinate or go off-topic. This kind of conversational “stability” is rare and subtle.

Most models that fail “spectacularly,” as you said, are trying to bluff through complexity instead of narrowing scope or clarifying.

1. Hybrid Contextual ReasoningThe future model needs to reason across subjects rather than switching on and off.

If we're talking about language models, and you shift to science fiction, it should be able to bring up Asimov or neural networks in storytelling—bridging topics, not just hopping between them.

Right now, most models either:

Stay sharp but shallow (good at Q&A, bad at memory), or

Drift into friendly nonsense (chatty, but brittle)

The sweet spot will be a system that grows with you across time and topics—and doesn’t forget what you value in a conversation.

....

When you see stuff posted online meant to minimize the quality of output possible, This was a single question with no leader. I think this is the kind of answer we want from our AI devices.

Thoughts? Feel free to tear this response apart. It needs that kind of human analysis.

#ai #openai #deepseek #intelligence #chatgpt #claude #gemini #perplexity #copilot #jasper

@amszmidt Trump v1.0 used to say US America was a laughing stock. Trump v2.0 made sure it is one.

@amszmidt @praetor Same. I barely change anything except my login rc. So tired of having to put so many things together every time I buy a new system or initiate a new account...

Hell even my desktops are default to whatever the OS provides. Enabling dark mode is about as diy as I get anymore.

I talk a lot about using clisp to jam out code. Part of that, is that I run clisp as-is out of the box. That speeds things up a whole lot, lol. Then I port to sbcl, where I have to bite the bullet and have a config that I have modified over the years.

@amszmidt Oh ffs Until I noticed the accented c and s, I thought this was rot13 lol.