Checkout https://t.co/3HfpzAq3AS a virtual pet that’s tied to a token’s price and community strength. Token is currently launched in @Pumpfun.
CA: 3ownsXNfXRVJR4YGxs4s686N8nb8E9rb9UE5AHDXpump
Okay. The first one I plan to release is another cool project. I have always been interested in the concept of intelligence as a spectator sport. So built something around that, but instead of humans it’s LLMs against each other.
Okay. The first one I plan to release is another cool project. I have always been interested in the concept of intelligence as a spectator sport. So built something around that, but instead of humans it’s LLMs against each other.
Close to releasing something probably cool. Been working on it for the last 8 months or so. Probably the biggest project I have ever done ever since I taught myself to code.
Close to releasing something probably cool. Been working on it for the last 8 months or so. Probably the biggest project I have ever done ever since I taught myself to code.
The "One Billion Row Challenge" by Ragnar Groot Koerkamp shows the incremental performance gains and the journey behind it - a solid weekend read.
https://t.co/nPXHVvb6Qc
Just released LTX Director 2.0!
LTX Director is a free, open-source, all-in-one tool for creating AI videos in @ComfyUI. It will hopefully provide a free alternative to the many expensive closed-source tools.
I've been working full-time on this update for the past month and a half, and I'm excited to finally release it. Hopefully it'll be a big help to the open source community!
Key New Features:
Complete Video Support: Edit Videos with AI all inside the node. Videos can be extended using a combination of prompts, keyframes, and audio. Trim, Split, and combine videos all within the timeline.
IC-LoRA Support: Take full advantage of IC-LoRA's to take your generations to the next level. Simply drag and drop videos onto the IC-LoRA track to quickly setup IC-LoRA videos. Compatible with prompt relay, keyframe, and custom audio features within the node.
Audio Inpainting: Seamlessly blend imported audio with generated audio. Not only can audio be extended, but can also be prompted alongside your imprted audio to really bring your generations to life.
Retake Mode (Beta): Redirect what happens within a shot. Allows you to select a segment within a video, and re-generate what happens in that segment. An early working experiment.
Timeline Saving/Loading: You can now save your timeline and settings to a json file. It will keep any videos/audio/images you have imported into the node and every setting you have changed.
UI Overhaul: Huge update to the UI, dozens of big changes such as a new side bar, redesigned prompt boxes, a bunch of new settings and redesigned menus, and more.
Quality of Life Improvements: Snapping, in/out points, multi-select, mark selection, workspace folder, more HUD options, resizable prompt boxes, new hotkeys, labels, filename preview options, "split at playhead" functionality, end frames (convert any keyframe into a end/last frame), toggleable tracks, NAG Support, tons of bug fixes and more!
And of course it can do everything it could before: Text to Video, Image to Video, Prompt Relay support, Keyframe (first/last frame) support etc.
@DavidSHolz I only have one question, how will we ensure this works for all and not just a privileged few? As a Transhumanist, this is something I constantly struggle with.
Training Qwen2.5-0.5B-Instruct on Reddit posts summarization tasks with length constraint on my 3xMac Minis with GRPO - evals update
So, I trained two variants of this task:
>using just length penalty
>using a quality reward and length penalty
I ran LLM-As-A-Judge eval for checking the summarization quality using DeepEval tools. Those are:
>Consciencess
>Coverage
>Clarity
>Faitfullness
Th results are as follows:
1) with quality + length penalty rewards: 2.5/4
2) with just length penalty: 2.4/5
Results:
Both are significant with a p-value of 0.0042 using one-sided t-test with a total of 5 rounds of evals for each model.
Performed on the test sample of 200 of smoltldr dataset.
Baseline: length penalty only
Pause AI rhetoric is predicated on the notion that the AI companies are recklessly racing toward dangerous tech and that a government controlled pause button is therefore necessary, but this seems really hard to reconcile with the fact that government is attempting to destroy an AI company because *the government* is racing toward plausibly dangerous AI uses (Sec. Hegseth has stated in official directives that he wants to deploy AI into critical systems regardless of whether it is aligned, for example) and *the company* is pushing back.
The roles are totally reversed from the logic that Pause AI and frankly other AI safety advocates confidently assumed for years. It is *industry* that is in favor of alignment and at least somewhat measured deployment risks, and government whose actions seem much closer to reckless.
I predicted this for years. I said, in particular, that pauses and bans and licensing regimes gave government a dangerously high degree of control over AI, and that the incentives of government are much more dangerous than those of private industry with competitive market incentives.
I believe the events of the last month are good evidence in favor of my view. At this point if you are an AI safety advocate whose policy proposals do not wrestle seriously with the brutal political economic reality of the state and AI, I don’t take you seriously.
It gives me no pleasure to have been right about this, by the way. The state has an incredibly strong structural incentive to centralize power using AI, and we are, all of us, not so empowered to stop it. I am quite concerned about this.
An MQ-4C Triton sent an emergency “7700” signal and disappeared from public tracking near Iran. Some reports speculate it may have been affected by Iranian actions, but no official confirmation exists about it.
Nick Bostrom’s new paper:
>Developing superintelligence is not like playing Russian roulette; it is more like undergoing risky surgery for a condition that will otherwise prove fatal.
> One could equally maintain that if nobody builds it, everyone dies. In fact, most people are already dead. The rest of us are on course to follow within a few short decades. For many individuals—such as the elderly and the gravely ill—the end is much closer. Part of the promise of superintelligence is that it might fundamentally change this condition."
>Along one path (forgoing superintelligence), 170,000 people die every day of disease, aging, and other tragedies.
>The choice before us, therefore, is not between a risk-free baseline and a risky AI venture. It is between different risky trajectories, each exposing us to a different set of hazards.
>Imagine curing Alzheimer's disease by regrowing the lost neurons in the patient's brain. Imagine treating cancer with targeted therapies that eliminate every tumor cell but cause none of the horrible side effects of today's chemotherapy. Imagine restoring ailing joints and clogged arteries to a pristine youthful condition. These scenarios become realistic and imminent with superintelligence guiding our science.
>We assume that rejuvenation medicine could reduce mortality rates to a constant level similar to that currently enjoyed by healthy 20-year-olds in developed countries, which corresponds to a life expectancy of around 1,400 years.
>Developing superintelligence increases our remaining life expectancy provided that the probability of AI-induced annihilation is below 97%.