Видео

I really hope the government doesn’t hamper the development. 😔

Very Informational. I love your presentation. --- On a side note, I still can't figure out if you are using heygen or an avatar to represent you or you are indeed a real human. With the plain background and always sweating on the side of your face including simple arm movements, this is most likely an avatar. Yet, the last part shows you have a nice smile which makes me think you are human. We are just at the age where we have to guess if what we are seeing is real or not.

Your Explanation is the best

An LLM isn't going to become an AGI. This is stupid. LLM's don't think. They are just a glorified version of the google search bar.

Hey guys, what are the potential situations that lead to mass extinction of humans because of an Agi and asi? I guess I just can’t really wrap my head around it. Obvi I get the economic part of it. The rich get richer and the poor likely get poorer. But I don’t see how asi could eliminate us besides like a terminator or iRobot type of thang

What is timeline for AGI?

As soon as you mentioned the words, "hedge fund", all the atmosphere of gloomy predictions sucked the logic out of the room. We will watch a movie about this guy one day after he takes his $2-billion fund and disappears. I'm sure his sales prospectus will talk first about curing hunger, as that is the latest cult tactic. Maybe he could prove useful instead and chart the total time-window of danger posed by slowing down AI vs advancing it as fast as possible. But that would involve real intelligence and not just greed.

Excellent presentation of a weighty topic. Good job.

So which of these two scenarios faced more risk: 1) A Jensen Huang's $40K investment in his startup Nvidia in 1993? 2) Or someone today buying Nvidia stock on the public stock market?

I wish you explained how LLM work . We need to debunk what AI is, so people can understand that It’s actually a predictive algorithm that can imitate human intelligence pretty well but the intelligence is not intrinsic to it,It’s being programmed by people. It’s clear with the autonomous driving that has been promised for some years. More data doesn’t mean more intelligence . It really doesn’t understand how the world works and can be fooled easily… enough with all the sky-net doom talking..it’s pretty had to see how LLM can really understand the world . The technology just isn’t built that way.

Chat-gpt types of AI (Autoregressive Large Language Models) are still very narrow AI. Which is contrary to the core assumption of this argument. e.g. they cannot solve problems with exponentially large problem spaces like chess. They cannot really plan and don't have any plasticity other than the input context. This current path won't lead to AGI, let alone ASI.

It’s apparent that we could be in real trouble when so many leaders and researchers in the field of Ai, as well as the answers of Chat GPT 4.o and Claude 3 Opus that we are heading towards a cliff without a driver nor breaks. Seems that Claude thinks we have a 50-70% chance of extinction within 2 years of AGI. It’s not only Claude, it’s the lead Safety researchers who have been tasked to protect us.

You proved 2 points: 1. As AI advances, it juat becomes software. 2. I personally dont believe there will be a singularity but simply a gradual increase in abilities and intelligence until we get to what we may call AGI or ASI but I still think even when we get there, we will be so used to it, we will just call it software.

Anyone ever hear of Malthus? Or a myriad of other alarmists shown to be drastically misguided?

Nice vid, slow down with ya words sometimes

I've been hearing AGI is coming every week since chatGPT was released

Hey man, I thought your intro and graphics looked pretty damn good!

Synthetic data, a creation of code Artificially generated, not from the world's abode Algorithms spin, a digital weave To validate models, and machines to retrieve In simulation's realm, it takes its place Music and flight, a synthetic embrace Approximating reality, yet born of rule Fully algorithmic, its digital cool A filter for secrets, it shields with care Confidentiality safeguarded, without a single snare Sensitive fields rely on its disguise To protect the privacy of watchful eyes Synthetic data, a helpful innovation indeed A digital surrogate, for the real thing we need.

Phenomenal presentation in every way, great work brother❤

It going to branch off somewhere

Thanks for your video! I'm an avid reader of the newsletter - a question if you don't mind: would making your own models still be useful in the future given the AGI systems / other advancements in AI do you think?

You said "Let's take a breath for second" when I literally started to take a deep breath after hearing the information)

Insightful! Refreshing. Good job.

This is fast becoming one of the best AI channels on RUclips. You and AI Explained are probably the only two channels I really trust to not succumb to, or partake in, hype about AI. Just plain, straightforward reporting and commentary, which is what we need. Thank you!

This is fast becoming one of the best AI channels on RUclips. You and AI Explained are probably the only two channels I really trust to not succumb to, or partake in, hype about AI. Just plain, straightforward reporting and commentary, which is what we need. Thank you!

China winning the ai race would be scary

Great stuff 👍 insta sub

do you pay people to say youre beautiful?

Asian hero ❤

11:29 a super intelligent ASI could rethink plumbing altogether. We really aren't prepared for the utility of superintelligence if it's friendly.

Keep going Pete! Your videos are world class

Let me come back and comment here 2 years from now. I will have a better understanding where AI is heading. Until then, your guess is as good as mine. "WE DO NOT KNOW!"

I don't trust any "laid off AI researcher" , who warns me about "skynet" rather than : 1) copyright and IP theft 2) increase of research papers made by chat gpt on the internet . 3) Web being littered by hallucinated content , like having open source maintainers dealing with issues and pull requests opened by AI bots trying to commit fallacious bug patches. 4) any tech giant trying to sell the "skynet" danger to governments , to have them regulate open source AI , so that only big licensed corporations have the full access to this technology . Other than that , LLMs wont become AGIs because they were trained more , that's not how it works

your biceps, slim build and energy, really motivate me.

4:29 is all I need to know. Keep philosophers out of the way of builders.

I like your videos, amazing presenting skills

Even though the researchers "ran out" of data to train, it seems there are ways to extra more out of current datasets. Take alook at Grokking. papers.neurips.cc/paper_files/paper/2022/file/dfc310e81992d2e4cedc09ac47eff13e-Paper-Conference.pdf

Thanks for teams’s insights, keep it up 😊

You are amazing

GPUs may not necessarily be the substrate for AGI/ASI.

Great content! Keep up the good work! Subscribed!

The new graphics are so fresh! Really impressed by Clipt

Global Domination

Cuda

❤❤

The cat is cute !!!

This is a side show. AI is fastest developing in China, Russia and maybe Japan.

Now this is how you turn complexity into simplicity! You have a new subscriber! Thank you!

Blackwell isn't more power efficient its just two previous gen chip fused together. The efficiency comes mostly from lower floating point accuracy. Groq is built from the ground up to be inference efficient. Blackwell is great for training terrible for inference and that is where the rollout to end users will be will accelerators (NPU). Nvidia lead is mainly in-situ graphic cards and drivers running CUDA libraries. This will not help the mobile market.

Pete! Pete! Pete! The man is back 😎🙌🏻. This was super interesting 🔥