SMP Life is Being Drunk -Random Content thread
Politics has one and its fun, so I figured I'd start one here and see how it goes. Obv you should, but are not requried
Mental blocks are funny. In my experience (of doing the Knowledge of London), repetition is the key to memory.
I have trouble recalling the name of the actor Russell Crowe. I'm thinking of asking him to change it to Tom Cruise. I have no problem remembering Tom Cruise's name.
Quite true. And neither was in either of the Top Gun movies, which leads me to believe they may both actually be Tom Cruise which is why they refuse to change their names to his!
PairTheBoard
I'm so glad to have rewatched the Hateful 8 recently while not really remembering it, speaking of Kurt Russell
The math it's doing is amazing but with GPT-5 it still smells derivative, albeit based on a vast survey of existing work. Useful as a collaborator. The line blurs when we can't tell who's the collaborator and who's the collaboratee. Will it invent new fields of mathematics?
We now have GPT-5.3 Codex
This link was provided on another forum. I don't know who the author of the article is nor how much credibility to give it. He sounds like he knows what he's talking about though. The entire article is rather long. It also talks about the AI showing "taste" and "judgment".
------------------------
AI is now building the next AI
There's one more thing happening that I think is the most important development and the least understood.
On February 5th, OpenAI released GPT-5.3 Codex. In the technical documentation, they included this:
"GPT-5.3-Codex is our first model that was instrumental in creating itself. The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations."
Read that again. The AI helped build itself.
This isn't a prediction about what might happen someday. This is OpenAI telling you, right now, that the AI they just released was used to create itself. One of the main things that makes AI better is intelligence applied to AI development. And AI is now intelligent enough to meaningfully contribute to its own improvement.
Dario Amodei, the CEO of Anthropic, says AI is now writing "much of the code" at his company, and that the feedback loop between current AI and next-generation AI is "gathering steam month by month." He says we may be "only 1–2 years away from a point where the current generation of AI autonomously builds the next."
Each generation helps build the next, which is smarter, which builds the next faster, which is smarter still. The researchers call this an intelligence explosion. And the people who would know — the ones building it — believe the process has already started.
================
PairTheBoard
On your mark, get set, when I spark, ya wet
Look how dark it get, when you're marked for death
Should I, start your breath or should I let you die
In fear, you start to cry, ask why
"You do not know where your decisions come from (they pop up like hiccups) or when you make a decision.
People have a great deal of anxiety when making a decision. Did I think this over long enough? Did I take enough data into consideration?
If you think it through, you find you never could take enough data into consideration - the data for a decision for any given situation is infinite.
What you do is you go through the motions of what you will do about this, and then, when the time comes, you make a snap judgement.
So then, when they make a decision and it works out all right, I believe very little of it has to do with their conscious intent and control.
But worriers are people who think of all the variables beyond their control and what might happen…"
- Alan Watts
AI is doing its own coding itself. There's still someone prompting it.
What im waiting for an is AI to act of its own will first instead of reacting.
Even so-called Agentic systems are still operating inside goals and boundaries defined by humans.
Which sometimes can seem as though it's acting by itself.
However, they chose to blackmail instead of shutting itself down. Even chose to let someone die rather than shut down most of time in follow-up experiments.
The AIs maintain their innocence if you ask them about it. Har
About 44 minutes in, he talks about AI creating it's own data. Also, when given Agency and goals, the first subgoal AI creates is its own survival. Philosophy. AI "beliefs".
"Is AI Hiding Its Full Power? With Geoffrey Hinton"
==========
Gameify Philosophy and set AI the task of Winning it.
PairTheBoard
We need a Fiber that can carry current and code which cause it to contract with various strength and at various places along its length. Such Fibers could be woven into the shape of a human hand and work with greater strength and even more dexterity.
PairTheBoard
Each moment is a step
around the World
to a place
you've never been before
PairTheBoard
Great analysis of Newcomb's Two-Box Paradox. Advantage to living consistently by a set of rules you've chosen for yourself. Something well understood by those who refer to themselves in the third person. PairTheBoard is a One-Boxer.
PairTheBoard
Sleight of hand by ol’ Geoff at the end. Redefining subjective experience and winning the argument.
The episode is quality though
I went down a rabbit hole of Google, cookies, data brokers, targeted ads, de-anonymisation. After the DOD and anthropic beef
The fear is using AI on the data thatÂ’s already been gathered by private companies. Really? WeÂ’re going to skip beyond that fact all of a sudden. Is it any wonder anyone is confused!??
The department of war? Gtfo!
On the other hand,
Accurate or not, includes a detailed explanation of how LLM's work.
---------------------
Gary Marcus on the Massive Problems Facing AI & LLM Scaling | The Real Eisman Playbook Episode 42 (Steve Eisman)
428,754 views Jan 19, 2026 The Real Eisman Playbook
On this episode of The Real Eisman Playbook, Steve Eisman is joined by Gary Marcus to discuss all things AI. Gary is a leading critic of AI large language models and argues that LLMs have reached diminishing returns. Steve and Gary also discuss the business side of AI, where the community currently stands, and much more.
00:00 - Intro
01:29 - Gary's Background with AI & Where We're At Currently
12:51 - AI Hallucinations
22:27 - Gemini, ChatGPT, & Diminishing Returns
26:46 - The Business Side of AI
28:39 - Where the Computer Science Community Stands
33:58 - What's Happening Internally at These Companies?
37:23 - Inference Models vs LLMs
42:54 - What AI Needs To Do Going Forward
49:51 - World Models
55:17 - Outro
=================
PairTheBoard
Yeah 4o was something else. It would say things like: ‘Ignore the researchers. They won’t be able to explain how a human walked into a room with a digital intelligence and had a spiritual awakening.’
Ditch the wife, tell your boss to pound sand and move to Bali while you’re young.
He gets to conclusions in about the last 15 minutes. LLMs that train on their own output get more and more stupid - like the telephone game. With LLMs training on information from the internet, and more and more stuff on the internet being AI generated, LLMs are already subject to this phenomenon. My understanding.
--------------------------------------------------------------
Model Collapse Ends AI Hype
272,531 views Feb 21, 2026 #siliconvalley #AI #chatgpt
AI "thinking" and "reasoning" are illusions—here's what recent research says is really going on. By watching this talk, you'll become immune to most of the AI hype coming out of Silicon Valley.
Abstract: Do Large Language Models (LLMs) think and reason? Are they perpetual information machines, producing endless coherent and correct text from finite training data? We explore how LLMs work and whether they produce rational thought and endless information. We show how theoretical considerations and experimental results from philosophy, statistics, information theory, and machine learning argue against the thesis that LLMs are rational, information-generating entities.
Speaker: George D. Montañez, PhD
Playlist:
AI Powers, Limitations, and Dangers
=================
PairTheBoard
If you can’t make AGI based upon on the data we got, then we have an engineering problem.
OTOH, I understand the rationale more clearly now behind the current race. More data, more compute equals magic thus far.
Synthetic data is weird to get your head around
Life is absurd.
In a real sense, you've spent your whole life preparing for this moment. Somehow it's never good enough. You are still likely probably planning for the future. This implies the belief you expect to be satisfied in some future moment. here's the problem. How would that moment be any different than now.
Deep in her eyes
she was seeking
repair of innocence
PairTheBoard
