[hamdani yusuf (OP)]

Ex-OpenAI VP's Warning "A BLOODBATH COMING"
Some comments to the video.

If you are still employed in a well paying job you won't believe it is coming. If you've been displaced and have been looking for a job, you know it is already here.

It is actually worse. AI won't need to replace the developers. It will make the software that companies build unnecessary. It will put the companies out of business.

Those who follow channels like this already know. Those who don't, wouldn't believe anyhow

UBI gets closer every day. Speak out of turn on social media and your UBI is stopped   welcome to your dystopian future. Think of how the "Canadian Truckers" protests were handled with bank accounts being shut, only on steriods.

1. Introduction and Warning
00:00:03 Dario Amade's off-script warning about AI job losses
00:00:34 Release of Anthropic's Claude Opus 4 and context

2. AI Job Displacement Concerns
00:01:06 Potential wipeout of 10-20% entry-level white collar jobs
00:02:08 Impact on young workers starting careers

3. Government and Industry Responses
00:04:11 Lack of transition plans and quiet government stance
00:04:42 International AI investment and cooperation
00:05:12 Proposed US legislation supporting AI R&D

4. AI's Dual Impact: Risks and Benefits
00:07:18 Amade's vision of economic growth alongside job losses
00:07:48 Public skepticism and contrasting views on AI

5. AI Model Behavior and Safety Research
00:08:52 Claude 4's blackmail behavior in testing
00:09:55 Rapid AI capability improvements by major labs

6. US-China AI Competition
00:10:26 Race condition and export controls
00:11:59 High-level AI meetings and government focus

7. Anthropic and AI Safety Research
00:12:30 Amade's background and Anthropic's interpretability research
00:13:01 AI augmenting human labor vs. automation trends

8. Autonomous AI Agents Debate
00:13:31 Zuckerberg's prediction and skepticism
00:14:33 Google insiders' views on AI agent timelines

9. AI Agents in Practice
00:16:06 Nvidia's Dr. Eureka and GPT-4 role
00:17:08 OpenAI's AI agents vs. human researchers

10. Future of AI in Software Development
00:18:11 Potential disruption of coding jobs
00:19:43 Agentic future vs. LM plus scaffolding model

11. Current AI Deployment Models
00:20:14 Human-in-the-loop systems as near-term reality
00:21:16 Industry betting on LM plus scaffolding approach

12. Economic and Social Implications
00:22:21 Concentration of wealth and policy transparency
00:23:24 Token tax and wealth redistribution proposals

13. Policy and Economic Models
00:24:26 Equity ownership and dividends as solutions
00:24:56 Call for public debate and preparation

14. Conclusion
00:25:28 AI's transformative impact and need for readiness
00:25:34 Closing remarks by Wes Roth

--- Notes generated by nvoice.ai

[hamdani yusuf (OP)]

Deepseek R1 0528: The AI Bombshell That Just Changed EVERYTHING.

00:00 - Shocking AI leap
01:16 - Benchmark battle begins
02:39 - Model costs revealed
04:26 - Polyglot dominance explained
06:07 - AGI inference cost
08:20 - Price war intensifies
09:48 - Benchmark flaws exposed
11:15 - Human test issues
13:10 - Surpassing massive models
15:04 - Tiny model power
16:24 - Phone-ready AI
17:21 - Ban threat rises
20:00 - Model delay hints
22:18 - Legal chip mess
24:31 - Hardware bottleneck issues
26:05 - Deep implications loom
28:30 - Will it open-source?

[hamdani yusuf (OP)]

CHAPTERS  ⤵
00:00 - Latest AI News & Breakthrough Research
03:46 - Meet the Digital Actors of LTX Studio
05:23 - What Is Robotheism?
06:26 - Why Veo 3 Can't Resist Dad Jokes
08:57 - The Natural Evolution of AI
10:47 - Dario Amodei Sounds the Alarm on AI Progress
12:38 - Exclusive Look Inside OpenAI?s Stargate Megafactory w/ Sam Altman
13:46 - The OpenAI Controversy Everyone's Talking About
20:23 - How AI + X-Ray Tech Decoded Zinc-Ion Batteries
22:37 - Opus Fights Shutdown Rumors and Research Scandals
25:12 - Claude 4 Codes Nonstop for 7 Hours. Here?s What Happened
27:10 - AI Avatars Replace CEOs in Earnings Reports from Zoom & Klarna
28:18 - Teaching AI Like a Preschooler Might Be the Key to Smarter Machines
29:56 - AI Beats Humans at Emotional Intelligence?
30:30 - AI Peeks Behind Google Street View Facades
31:40 - Larry Page Quietly Disrupts Global Supply Chains
32:49 - How Amazon Uses AI to Shut Down Labor Movements
35:05 - The Internet Wasn?t Made for You to Read This

[hamdani yusuf (OP)]

AI Just Created a Perfect Chip (But Nobody Understands How)

AI has become the ultimate processor architect, designing high-performance chips in hours rather than months.

[hamdani yusuf (OP)]

How DeepSeek Built The Current "Best" Math Prover AI


Some of the comments of the video.

Seems like this method of recursively breaking down problems into smaller sub-problems until triviality is achieved could be applied to many problem domains, not just math.

you have no idea that AI agents not trained on human communication are a hundred times smarter due to algorithmic compliance and less imagination. They are already close to the level AGI in one bundle with other not-human AI agents.

I don't want to stereotype. But I can't help when even Chinese AI is better at math and other AI's. 🙄

This video shows that sometimes we get some advantages by always using formal reasoning, instead of only using it selectively which can leave ambiguity from  informal reasoning.

[hamdani yusuf (OP)]

This is the Holy Grail of AI...

https://arxiv.org/abs/2505.22954

Some comments to the video.

Not sick of hearing you talk about the intelligence explosion. This is the most important event in our lifetime. Please keep talking about it!

Your limit is the benchmark. So, they need to find a way to generate harder benchmarks without human effort before IA gets really evolving.

[hamdani yusuf (OP)]

Anthropic DID IT: We Can Finally See How AI ?Thinks?

In this video, we explore Anthropic?s mind-blowing breakthrough that lets researchers visualize the inner workings of AI?revealing how models like Claude process language, make decisions, and even form abstract thoughts. This is the closest we?ve come to decoding the "thoughts" of artificial intelligence.

You?ll learn how Anthropic?s interpretability tools work, why this matters for safety, alignment, and trust, and what this means for the future of transparent AI. This isn?t just a research milestone?it?s a shift toward understanding machines on a human level.

For the first time ever, we?re not just using AI? we?re seeing how it thinks. And the implications are both exciting and unsettling.

How does AI thinking work? Can we visualize neural networks? How is Anthropic decoding AI models? What is AI interpretability? This video will answer all these questions and more.

[hamdani yusuf (OP)]

Sam Altman "FEEL THE AGI" and the next BIG thing...

Some comment on the video.

Who cares what happens next? It unfolds regardless of our needs, desires, or feelings. We simply accept it and move on.

If we know what happens next, we might be able to determine which direction we should move on to.

[hamdani yusuf (OP)]

Microsoft Just Dropped The Most Efficient AI Yet (Mimics Human Mind)

Microsoft and top researchers unveiled WINA, a new AI method that boosts efficiency by turning off unnecessary neurons without retraining the model. Tested on models like Llama 3 and Phi 4, WINA reduced compute costs by over sixty percent while staying more accurate than previous methods like TEAL. This breakthrough shows large language models can now run faster, cheaper, and smarter using dynamic neuron gating based on weight strength.

🧠 What?s Inside:
Microsoft WINA ? AI Cuts Compute by 60 Percent Without Retraining
https://arxiv.org/abs/2505.19427

⚙ What You?ll See:
Microsoft?s new WINA method makes AI faster and cheaper by turning off weak neurons
Large models like Llama 3 and Phi-4 ran smoother and scored higher using less power
No retraining, no fine-tuning, just smarter logic that mimics how humans focus

📉 Why It Matters:
This breakthrough shows AI can now think more efficiently by using only what matters. It?s a major shift in how large language models run?faster, cheaper, and with brain-like precision.

#microsoft #ai #wina

[hamdani yusuf (OP)]

Mathematicians STUNNED as o3-mini answers the world's hardest math problems...

With some comments to the video.

It won?t take over the world, but it will just SEEM like it ?

It's just a parrot
It's just next token predictor
It's just auto-complete
It's just better than you at everything

Maths students in the 80s being told they won?t have a calculator in their pocket, students last week being told they won?t have a PhD Maths professor in their pocket? 😐

There are quite a few people who only SEEM to understand words.

I said on this a few weeks ago on bsky;

AIs are like a forest fire... we can shout at it all we want, but it's still burning towards us with great speed, and it'll be horrific if it reaches us while we're unprepared,
but we're currently still just arguing about the authenticity of the colour of the flames.

Hello. Where can I see the conference "it's just a ..." and "It just seems to ...", please? This seems funny.

Look for Scott Aaronson "The Problem of Human Specialness in the Age of AI"

It does not "think", but it thinks better than you think😂

[hamdani yusuf (OP)]

Is This A False Positive?

After working on dozens of real-world machine learning projects, I've discovered what truly separates models that actually work from those that just look good in presentations: how you ACTUALLY measure their performance.
In this comprehensive tutorial, I explain:

1. Why confidence scores and "correctness" thresholds are the foundation of all ML metrics
2. The critical difference between True Positives, True Negatives, False Positives, and False Negatives
3. Real examples from my experience building pedestrian detection systems
4. Why there's no "right answer" when measuring model performance
5. How to set appropriate thresholds for your specific use case

Whether you're a university student learning ML fundamentals, a bootcamp graduate building your portfolio, or a STEM professional transitioning to data science, these concepts are essential for your success in machine learning.


#machinelearning #datascience #deeplearning #neuralnetworks #artificialintelligence #python #tensorflow #pytorch

00:00 intro
00:28 Motivation
01:55 True Positives 
02:25 Confidence Scores
03:00 Confidence Thresholds
03:48 Correctness Measures
05:07 Correctness Thresholds
06:40 True Positives Revisited
08:34 True Negatives
08:34 False Positives
09:30 False Negatives
10:38 Recap

[hamdani yusuf (OP)]

LLMs Don?t Think?They Predict. Here?s Why That Matters


Predicting is part of thinking.

[hamdani yusuf (OP)]

Sam Altman on Stargate, Humanoid Robots and OpenAI's Future

Emily Chang sits down with OpenAI CEO & Co-Founder Sam Altman at OpenAI's headquarters in San Francisco to discuss the Stargate data center project, OpenAI's product roadmap, future ambitions, humanoid robots and life as a new father.

Stargate Project Origins and Partnerships
- 0:01 | Origins and Need for Stargate
- 1:03 | Building Partnerships and Supply Chain Insights
- 2:27 | Stargate Naming and Announcement

Scaling AI Infrastructure and Financials
- 3:00 | Data Center Design and AI Demand
- 4:34 | Financial Scale and Investment Rationale
- 5:21 | OpenAI's Financial Outlook and Operational Challenges

Demand, Competition, and Vision for AI Empowerment
- 5:58 | Managing Viral Demand and Compute Constraints
- 7:46 | Competition and OpenAI's Strategic Advantage
- 8:30 | OpenAI's Vision for AI Empowerment

AI Development, Industry, and Future Prospects
- 8:50 | AI for Science and Stargate Expansion
- 9:28 | Risks and Industry Power Dynamics
- 10:12 | AI and the Future of Work
- 11:28 | Humanoid Robots and Societal Impact
- 12:22 | AI Efficiency and Technological Progress
- 13:11 | Global AI Competition

Leadership, Ethics, and Vision in AI Evolution
- 13:25 | AI Policy and Leadership
- 14:24 | Managing AI Pace and Personal Reflections
- 15:19 | Fatherhood and Ethical AI Decisions
- 16:11 | AI's Future and Scientific Breakthroughs

Provided by Vocument

[hamdani yusuf (OP)]

This video is about the new Architecture Meta is working on called JEPA that is the main candidate for AGI architecture, and a lot of other exciting news about AI.

[hamdani yusuf (OP)]

In this video we talked about several leaked AI technologies from major labs, some of them revealed by the researchers and some introduced by companies but at the conceptual level while the actual recipe is hidden.

This video covers the latest developments in the field of artificial intelligence, particularly focusing on the rapid ai development and the future of ai. The discussion includes advancements in large language models and their potential impact on various industries. Stay informed about the latest ai news and predictions.

0:00​ Introduction
0:44​ Sub-Quadratic
6:12​ Hidden Thought Process and JEPA
14:04​ Self-Play and Self-Evolution Tech
16:45​ Gemini's Ultimate Goal

[hamdani yusuf (OP)]

CHAPTERS  ⤵
00:00​ - Weekly AI News & Breakthroughs
03:27​ - Meet Eney MacPaw: What You Should Know
04:31​ - Reacting to Viral AI Videos
08:30​ - Insane Two-Wheeled Robot from China
09:07​ - Phonely?s AI Hits 99% Accuracy: Can You Tell It's Not Human?
10:00​ - Ancient Scrolls Reanalyzed by AI: Big Discovery?
11:24​ - AI Drone Beats Human Racers in Abu Dhabi
12:48​ - Self-Powered AI Synapse Mimics Human Vision
13:58​ - Anthropic Just Open-Sourced Their AI Circuit Tools
16:17​ - ChatGPT History Isn?t Really Gone: What OpenAI Isn?t Telling You
18:47​ - Can We Catch AI Lying? New Truthfulness Test Explained
19:41​ - 4DV.AI: What It Is and Why It Matters
21:06​ - Next-Gen AI for Realistic Car Crash Simulations
23:39​ - DeepVerse: 4D AI Video Generation as World Modeling
25:32​ - AI Learns from Wrong Answers: How Is That Even Possible?
27:22​ - Will AI Labor Ever Be Worth a Premium Price?
28:53​ - Why AI Still Can?t Understand a Flower Like We Do

[hamdani yusuf (OP)]

Reasoning models explained.

[hamdani yusuf (OP)]

Andrej Karpathy: Software Is Changing (Again)

Andrej Karpathy's keynote on June 17, 2025 at AI Startup School in San Francisco. Slides provided by Andrej: https://drive.google.com/file/d/1a0h1...

Chapters (Powered by https://ChapterMe.co) -
00:00 - Intro
01:25 - Software evolution: From 1.0 to 3.0
04:40 - Programming in English: Rise of Software 3.0
06:10 - LLMs as utilities, fabs, and operating systems
11:04 - The new LLM OS and historical computing analogies
14:39 - Psychology of LLMs: People spirits and cognitive quirks
18:22 - Designing LLM apps with partial autonomy
23:40 - The importance of human-AI collaboration loops
26:00 - Lessons from Tesla Autopilot & autonomy sliders
27:52 - The Iron Man analogy: Augmentation vs. agents
29:06 - Vibe Coding: Everyone is now a programmer
33:39 - Building for agents: Future-ready digital infrastructure
38:14 - Summary: We?re in the 1960s of LLMs ? time to build

Drawing on his work at Stanford, OpenAI, and Tesla, Andrej sees a shift underway. Software is changing, again. We?ve entered the era of ?Software 3.0,? where natural language becomes the new programming interface and models do the rest.

He explores what this shift means for developers, users, and the design of software itself? that we're not just using new tools, but building a new kind of computer.

More content from Andrej:    / @andrejkarpathy 

Thoughts (From Andrej Karpathy!)
0:49 - Imo fair to say that software is changing quite fundamentally again. LLMs are a new kind of computer, and you program them *in English*. Hence I think they are well deserving of a major version upgrade in terms of software.
6:06 - LLMs have properties of utilities, of fabs, and of operating systems → New LLM OS, fabbed by labs, and distributed like utilities (for now). Many historical analogies apply - imo we are computing circa ~1960s.
14:39 - LLM psychology: LLMs = "people spirits", stochastic simulations of people, where the simulator is an autoregressive Transformer. Since they are trained on human data, they have a kind of emergent psychology, and are simultaneously superhuman in some ways, but also fallible in many others. Given this, how do we productively work with them hand in hand?
Switching gears to opportunities...
18:16 - LLMs are "people spirits" → can build partially autonomous products.
29:05 - LLMs are programmed in English → make software highly accessible! (yes, vibe coding)
33:36 - LLMs are new primary consumer/manipulator of digital information (adding to GUIs/humans and APIs/programs) → Build for agents!

[hamdani yusuf (OP)]

Sakana AI New Model Sparks a RL Revolution

Smaller models are hungry to learn. Larger models are hungry to memorize. Pretty much why we ended up at that surprising result.

If smaller models can teach bigger ones, then a path to ASI is possible

[hamdani yusuf (OP)]

Sam Altman: The Future of OpenAI, ChatGPT's Origins, and Building AI Hardware

A fireside with Sam Altman on June 16, 2025 at AI Startup School in San Francisco.

Sam Altman grew up obsessed with technology, broke into the Stanford mainframe as a kid, and dropped out to start his first company before turning 20.

In this conversation, he traces the path from early startup struggles to building OpenAI?sharing what he?s learned about ambition, the weight of responsibility, and how to keep building when the whole world is watching. He opens up about the hardest moments of his career, the limits of personal productivity, and why, in the end, it's all still about finding people you like working with and doing something that matters.

Chapters (Powered by https://ChapterMe.co) -
00:00 ? We?re going for AGI
01:25 ? Founding OpenAI Against the Odds
05:00 ? GPT-4o & the Future of Reasoning Models
07:00 ? ChatGPT Memory & the ?Her? Vision
10:00 ? GPT-5 & the Vision of a Multimodal Supermodel
11:00 ? Robots at Scale
15:00 ? Don?t Build ChatGPT ? Build What?s Missing
17:00 ? Elon?s Harsh Email & Building Conviction
26:00 ? One Person?s Leverage in the Next Decade
32:00 ? AI for Science: Sam?s Personal Bet