We are tired of the news and speculation. Everyone just sitting and waiting for OpenAI to finally get on with it.

@Allplussomeminus

2 ай бұрын

GPT-5 immediately. 👇

@cannonstrong1346

2 ай бұрын

Ha ha ha, no doubt

@waterbot

2 ай бұрын

Open AI is trying to avoid any scrutiny from possible US election influence. after the election cycle, they will go herd, and monetize.

@DailyTuna

2 ай бұрын

Let’s assume the governments are already there either our government or China or Russia. If you were in charge of that department of defense, would you be waiting for open AI to come up with advance artificial intelligence?😂

@myautodidacticism

2 ай бұрын

​@pkfryerIf it's good enough you won't need a job

I'm rather convinced now, you lose some sleep over staying more on top of announcements and getting the word out than anyone else...

@ridgeisland

2 ай бұрын

Maybe this channel is all AI?

Opening Eye has developed AGI internally

@lernmor2137

2 ай бұрын

Rumored to be called The Third Eye

@benroberts8363

2 ай бұрын

Lol

@empirelinkx9930

2 ай бұрын

Yes, we got to really open our eyes 😄👀

@TheNexusDirectory

2 ай бұрын

Eagle eye?

@nickdisney3D

2 ай бұрын

yall know AGI isnt sentience, right?

his speaking pace is so stable now, and sounds almost monotonous.. sure seems to be AI voice now 😅

@DailyTuna

2 ай бұрын

Maybe that’s it Sam Altman is the AGI!

bro squeezed a 13 min video out of a single deleted tweet

@nickdisney3D

2 ай бұрын

thats what youtubers have always been good for. hopefully the AI takeover does better.

@user-yl7kl7sl1g

2 ай бұрын

@@nickdisney3D Yes, Ai will be able to squeeze hours of content out of a single deleted tweet. lol

@DailyTuna

2 ай бұрын

He has to keep it going, and he does good work so you have to cut him some slack

@muhsinkhalif3621

2 ай бұрын

@@user-yl7kl7sl1g while climbing mount everest

@winstonsmith6204

Ай бұрын

That's why I unsubbed. This dude has absolutely no value imn hos channel

2:13 we’re all sitting here listening to an ai voice talk to us about a single vague ass tweet for 13 minutes. I think I’m done with this whole ai news grind😂

@job1481

2 ай бұрын

Spot on.

@anonymeforliberty4387

2 ай бұрын

that quality of the voice fooled me

The video discusses speculative connections between a deleted tweet by Noan Brown, an AI researcher at OpenAI, and the secretive Q* model of OpenAI, but lacks concrete evidence or details. Brown, known for his work in AI and games, hinted at advanced AI performance beyond mere imitation of human data. The video suggests this could relate to OpenAI's Q* model, speculated to involve advanced planning and synthetic data generation. However, the discussion is based on speculation and interpretation of public comments rather than solid facts. The video also touches on the potential of AI systems to improve through planning, leading to more efficient and accurate outcomes. In summary, the video is speculative, discussing the potential implications of Brown's work and the Q* model for AI development without providing tangible proof or detailed information about Q*.

@elianproductions5128

2 ай бұрын

You sound smart

@christopherd.winnan8701

2 ай бұрын

Can you do a video which explains more about the "qualia' part of Qstar please? This video has the best explanation so far. The Symmetry Theory of Valence (@The Centre for Psychedelic Research at Imperial College London) kzread.info/dash/bejne/poeEvJaxpJjMYs4.html Jump to 6m 20s for the TLDR. "For any conscious experience there exists a mathematical object isomorphic to it" just as four simple equations tie together phenomena we know as electromagnetism, they are talking about qualia as a deep mathematical structure to consciousness. Is it this kind of pattern recognition breakthough that has been achieved with Qstar?

@callmetony1319

2 ай бұрын

@@christopherd.winnan8701 The "Q" in Q* likely stands for Q-learning (Q* = Q-learning + A*). The "QUALIA" Q* leak from a few months ago is almost certainly fake

@adrianfelipe18

2 ай бұрын

Saved me 13 minutes, tnx

@lorddeus369

2 ай бұрын

Was this written by ai :o

So basically just let the computer “think it out” for days instead of giving you an answer in milliseconds. It sounds so simple and yet brilliant.

@bodell2285

2 ай бұрын

It also just seems so obvious now. I mean if you were to take a math test, you would do much better if given 10 minutes rather 5 seconds per question. The speed at which its finding everything doesnt allow it to process it all completely to form a response. With this we're basically just allowing it to properly study everything and plan out what its going to do, we're giving it time properly learn the information rather then just copy it

@andreaschalke2005

2 ай бұрын

Yeah but the understanding is the massive thing here

@nuclearthrone2497

2 ай бұрын

This reminds me of the scene in Avengers with Dr. Strange taking a long time to think about all the possible scenarios and choosing the 1 that will make them win

"Openin Eye" - This must be a quality video if you can't even get the captions correct.

0:05 what is opening eye? or did you accidentally load a very old and unreliable speech to text model?

If a model can use more compute to generate better answers then you can also use these generated answers to continusly train the model.

AGI coming in between tomorrow and year 2902.

@powerHungryMOSFET

2 ай бұрын

Tired of these KZread videos ?😁

@BryanWhys

2 ай бұрын

We already have types of agi, and the components for better ones

He mentions Riemann Hypothesis many times. It’s in the realm of breaking encryption. “In mathematics, the Riemann hypothesis is the conjecture that the Riemann zeta function has its zeros only at the negative even integers and complex numbers with real part 1/2 . Many consider it to be the most important unsolved problem in pure mathematics.[1] It is of great interest in number theory because it implies results about the distribution of prime numbers. It was proposed by Bernhard Riemann (1859), after whom it is named.”

Oh, the audio is much better today. And it’s much less stressful when you’re talking slower too. 👍🏻👍🏻 Cant hear the caffeine so much in this one. 😂

@MD-qh6ld

2 ай бұрын

Audio is still quite bad

@ascgazz7347

2 ай бұрын

@@MD-qh6ld wasn’t bad on iPhone. Haven’t tried it through anything better.

@nastied

2 ай бұрын

He just sped up the audio before 120 % or so

@MD-qh6ld

2 ай бұрын

I am hearing artifacts on iphone, sometimes sentences are cut or repeated. Its only on this channel so it has to be his editing?

@ascgazz7347

2 ай бұрын

@@MD-qh6ld I wasn’t talking about the editing. I was talking about the actual audio, but yeah it seems he does it all very quickly, good editing takes time. Done my time on it! lol. The last few days have had digital distortion and extra spiky sibilance, that’s what I meant. Like a dying battery in a lav mic, or getting too close to a cheap capsule.

0:04 Someone who works on Opening Eye revealed what Q* is, It couldn’t be anything other than him.

This was such an insightful breakdown of the qstar speculation and the incredible recent demos showing multi-step reasoning in action (like Mesa's KPU and DevonAI). Watching AI plan out tasks is like seeing a kid do their homework, but instead of procrastinating on TikTok, it's pondering the secrets of the universe. 🌌 AI planning seems poised to take capabilities to crazy new levels - I'm on the edge of my seat for OpenAI's next moves...or should I be hiding under my bed? Just kidding! (or am I?) 😂

@drashnicioulette9565

Ай бұрын

Haha, nooo. No worries, if everything goes south you will not be the only one to suffer. That’s a fact and also a promise❤

@eltroito

Ай бұрын

@@drashnicioulette9565

LOL Tweet has shit to do Q* WTF bro?

I am done with this channel. Another exaggerating in the video. A whole new video is about one unimportant tweet. Please stop it, don't waste peoples time with useles information. Do videos when you have really important news.

@pooman2

2 ай бұрын

Having LLMs be 100,000x better is literally important

@CVE42287

2 ай бұрын

​@@pooman2Is this satire?

@pooman2

2 ай бұрын

@@CVE42287 how is that unimportant?

Some custom GPTs like Professor Synapse kinda work similar to this as well.

At 10:03 you can see the Maisa model is capable of reasoning because the UI presents the screen prompt "Reasoning..." 😮

5:55 Were getting the computer from Hitchhiker's Guide - Deep Thought. Imaging the answer we'll get after 7.5 million years.

@effehell7593

2 ай бұрын

We already know that answer.

is not the tweet archived?

One take-away: need to scale inference compute. For robots and autonomous cars, that means better, more energy-efficient neural net chips. Tesla took that seriously with its FSD chip. No-one else is putting custom neural net chips into cars and prototype robots. Closest anyone else comes is GPUs which use a lot more energy.

Whats with the weird cuts in the audio?

@chad0x

2 ай бұрын

AI generated voiceover

@Smokkedandslammed

2 ай бұрын

I think Wes Roth is all AI also. They're just training US to be familiar with THEM lol

Why is the audio editing so scuffed? So much audio gets suddenly cut off, and some of it gets repeated

AGI any day now, right? Or.. maybe it's just hype to get investment money. I think it's the latter.

The "AI reflection pause" was also teased by Sam Altman on Lex Fridman's podcast

I love Opening Eye. I see big things in their future. They have thinky robots.

If the deleted comment really relates to Q*, it means Q* is gen II: They no longer use a human source for the personality model, rather, they use a Gen I model to write down the millions of rules. Guess they apply heuristic imperatives to try to ensure safety.

Interesting approach. We live in discrete right/wrong boundary cognitive space, and we need it to explore the superposition right/wrong boundary space. If you were into that sort of thing, there are some impressive proofs. I imagine the proofs (and even the new maths required) produced by LLMs left to "play" mode will be so ridiculously clever you'll just have to laugh in astonishment.

They need an extension with the voice of the USS Enterprise D's ship AI that just states: Working... if your request needs too much compute for "instant processing". 😀

When I ask Q* what the meaning to life is for it to tell me that I must wait multiple Millenia for the answer… which will be 42

hm, something is odd with the video.. it looks AI-generated!

Synthetic Data is nothing new (3 months old). Things I want to know: 1. UBI 2. How does a system leap from being a very smart human to being smarter?

Noam Brown is exactly who people interested in AGI should be paying attention to. Open AI must be producing new data with Q* and using their AI to analyse that data to create whole new inferences. Great Job on the video!

I guess, the next step will be look-ahead-try-and-assess capability in the existing LLMs.

Sam Altman said a similar thing in his interview with Lex Friedman. This kind AI would be out of the hands of ordinary people though

Heh some of us have already gone beyond the planning stage and are now working on the orchestration phase. At least some of the groups I’m in have this using opensource AI for the frameworks (we help one another but each have our own setup)

@lyndonsimpson1056

2 ай бұрын

Can you explain "some of them already have that"?

They're essentially saying: let him cook

🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Can't wait for GPT5! It is taking so long for it to be uodated. Thanks for thevideo, very informative.

I truly appreciate your channel and the content you create! However, I strongly believe that a different microphone could significantly enhance your voice quality. The current one seems to produce some harsh sounds that can be challenging for listeners. Investing in a microphone better suited to your voice would undoubtedly improve the overall experience for your audience. Thank you for all the amazing work you're doing!

it will happen and as the methods get more advanced the inference cost goes up and the bottleneck becomes even larger in the lack of Smeiconductor production and consolidates power to organizations who have the resources and bandwisth to run these new next gen models

Opening EYE instead of OPEN AI LMAO

I'm the first I'm going sell some stonks

@WillyJunior

2 ай бұрын

Which stonx

@Cross-CutFilms

2 ай бұрын

​@@WillyJunior Only the BESTEREST dronk stonk stonx stox!! Shhh!! Listen to the man, geez! 😜

@WillyJunior

2 ай бұрын

@@Cross-CutFilms understood! I'm buying now

@Cross-CutFilms

2 ай бұрын

@@WillyJunior we could go halves 🤷

(6:19 up until now,) that's just the so called agi concept getting discussed there, to be specific the first generation agi prototypes that are made in the ai labs. Genuine people are thinking and working on more than just that nowadays. (maybe) thinking part is shorter than the 'working' part. and for the conventional methods of doing things, they would need some trial and error to reach an optimal destination. having resources such as the higher compute power is just an essential for doing things. in future it may or may not require the same way, but for the present world folks, it is required. but my personal recommendation is that, don't make 'llm' the entire ai topic in your mind, because there is more to the ai paradigms, and it was always the case for decades. llm studies is helpful and would be helpful for understanding the upcoming ai discoveries or the expansions of the existed discoveries that are not enough popular now majorly.

I already told you - Q* is just QuietStar. The paper is not new. And it is exactly what you are describing.

3:20 it's narrated by an AI voice

That tweet mostly sounds like a comment on using artifical data.

nice that you slowed down the pace..

I will believe when Microsoft solves issues in teams or edge browser. Till then all our jobs are safe

@TheEarlVix

2 ай бұрын

Or when, from the official technical specs (W3C recommendations or Internet RFCs) alone, AI can code complete IMAP email clients and servers from scratch.

@cyclopes968

2 ай бұрын

@@TheEarlVix My point is even if they didn't release to public, they would have been using it internally to improve their products. I dont see any of the product improve.

@TheEarlVix

2 ай бұрын

@@cyclopes968 Ok, got you. Still when they can do the things I suggested then it really is game over!

bro how? this is really impressive

take 1000x as long to generate the synthetic data, use multiple models including ones that explore geometric relationship between abstract symbolic relationships in the logic, train the models on algebra, learn to program the geometric and symbolic logic inside LLM inference, as in physical world lowest energy cost state wins unless a local trough gets in the way, once in a trough, random exploration of remaining areas of the symbolized relationships allow you to efficiently escape these minimums, using first, second, third, ... derivatives helps find direction, just like NN training itself, would be cool to try and train a mini model on the abstract algebraic relationships of the tree of thoughts in the symbolic logic, perhaps two layers of inference is implied in human vision center on way to brain in this manner, training is so costly though, good enough for long drawn out test but not practical in most cases for most applications, but for Riemann Zeta function, why not train on the different paths through the symbolic logic to get inference on optimal path to take? turtles all the way down! model in model giving it a sort of holographic consciousness.

Considering they are building a $100 billion AI supercomputer by 2028, I'm even more curious about the capabilities of AI on their supercomputers beyond 2030 that collectively will cost up to another $6.9 trillion. Microsoft and OpenAI spending a good chunk of the entire U.S. GDP on new supercomputers will lead to colossal advancements, so the hyperbolic exponential growth curve might be true after all and closer than we think. Anyway, time will tell.

Noam Brown is like demis hasabis of openAI. Infinite labeled synthetic data with SORA. Generating single GPU shaders predicting them. 298 characters.

Ions have concepts that drift away. Let me show you how we play.

They should rename to Opening Eye

Open A Eye would be dope tbh

Why does the Military Industrial Complex hide behind these actors?

I could write something along the lines of "just call it skynet and get it over with" but at this stage, it's useless. If OpenAI don't do it, Google will and if they don't, other companies will.. Like, there is no stopping now. Humanity shall uplift. We can do it the nice way, by preparing a heaven simulation and programming copies of ourselves to live in it, or the "I don't have a mouth and I must scream" way. But.. Either way, humanity shall uplift.

A grid of domain expert, but dumb neural networks, organized and utilized by a high level, planning, pondering intelligence. That is how it will be ... The high level entity is Q*

The tweet said it all. Thank you.

"BIG" is so big in thumbnail

Memories. You're talking about memories....

This is crazy.... I mean what even can a 100000X gpt4 would be capable of doing? It's Possibilities are immense and our comprehensive capability is limited. Only time will tell what fate it holds for us. According to me we are going down ..... rapidly...

@StonedAi

2 ай бұрын

@CdawgAMVsFilmEditing ha ha ha...true that.... The repeated use of same word is truly frustrating....ponder, delve, intertwine...etc etc. I always mention the words specifically and ask it not to use em.....while writing prompts

MKUltra breaks I gonna add this video to my dataset

I think the real next steps of ai are going to just be increasing efficiency rather then intelligence. Because efficiency will improve its performance anyway. They need a better foundation, i think its too much money and time for the whole stargate project, if they just make their model take less energy and raw computational power, they wouldnt need to wait till 2028 or whatever to continue improvement. If you think about how much energy it takes for our brains to function, you can see there is a much more efficient way to obtain intelligence that we just havent found, but we're starting to. All improvements in ai come from them mimicking humans more, reinforcement learning, mimicking, self dialogue, self play, reasoning, this is all stuff humans do just in a different context. So i think the real end goal for us and ai is to create a synthetic brain

@cristianandrei5462

2 ай бұрын

Yeah, efficiency and reliability are the way to go. At the moment we have big, expensive and unreliable general system like chatgpt. What we need for adoption is specialized models, reliable and efficient.

I've been saying for a while that there needs to be a paradigm shift with AIs to get to the next level. LLMs and Transformer models no matter how many tokens are only prediction engines.

@italosantos9174

2 ай бұрын

exactly. this guy is justing surfing the AI Hype and making money from videos.

2 ай бұрын

Prediction is an insane power though. Probably the best one.

Up as a trader, sad as a human being. What th future holds for us.

Man, I like your info but also take care of your health. Sleeping is very important

When did people start saying 'x' instead of 'times'. It just makes people sound really dumb.😂

The video discusses a deleted tweet by Noam Brown, an AI researcher at OpenAI, which led to speculation about OpenAI's secretive Q* model. The tweet suggested that superhuman performance in AI cannot be achieved merely by improving imitation learning on human data. This statement is speculated to relate to Q*, which is believed to involve advanced planning and potentially synthetic data. Noam Brown's work focuses on AI systems capable of superhuman performance in games like poker, and his research could be applied to language models to improve their efficiency and capabilities. The video mentions that if successful, this research could lead to language models significantly better than GPT-4. The concept of planning in AI is emphasized, with the potential to scale up the inference cost for more accurate and valuable outputs. For example, in applications where speed is not crucial, allowing the AI more time to "think" could result in higher-quality responses. The video also references other AI developments, such as Mesa's KPU and Devon, which demonstrate the potential of planning and agentic behavior in AI systems. These advancements indicate a shift towards more sophisticated AI models that can reason and plan over multiple steps, potentially leading to significant improvements in AI capabilities. Overall, the video explores the implications of Noam Brown's tweet and the broader trends in AI research towards planning and synthetic data, suggesting that these developments could lead to a new generation of highly capable AI models.

@lyndonsimpson1056

2 ай бұрын

Genearated by AI/10

Noam: tweets. A colleague: Hi, Noam, I just read your tweet. You have no idea what our group has achieved! Noam: deletes World: speculates that Noam has let out a shocking secret

His statement was wrong. It is quite straightforward: you do get superhuman performance with any machine that outperforms humans. There ar two different ways to outperform us. One, for sure, is someone/something being able to operate in useful ways that we are still incapable to even imagine, for the most part. This refers to cognitive abilities unavailable to us, which includes ways to use available knowledge, plus new knowledge resulting from that. Reasoning, so to speak, and cognitive gain. That is the qualitative side of things. But machines can also outperform us, and already do, simply quantitatively. From this point of view, a system that would simply imitate us, but would do so with access to far more data, with far more computing power, will obviously outperform humans. That would be superhuman performance, based on "human-like" operations. But, the very same system, if allowed to modify itself, might well, soon enough, develop truly superhuman intelligence, sentient or not (i.e. ways to operate and knowledge beyond our current comprehension). This is all very dangerous. This in itself, plus other global circumstances, mean that we are entering the most dangerous era of human existence so far.

Whats the bet he used Ai to right a 13 minute script

All they need are humanoids that learn world models like human babies do (by playing, imitating etc). Then these robots can share data and upload it every night to the super computer to update the world model.

@ascgazz7347

2 ай бұрын

Oh, that’s “all”? 😂

“Accidentally”

*... Q is Stargate?*

@drashnicioulette9565

Ай бұрын

Daaammnnn!😱😱 That’s actually great speculation

Well it's just logical. Humans think very high about themselves, but human intelligence can only get you so far. To create AI intelligence many times greater that human, the AI would need synthetic data.

I got it down Deanna souders

Do you think that in 2035 with the advent of AGI (Artificial General Intelligence) and ASI (Super Artificial Intelligence) they will introduce universal basic income!? a subsidy!? there will be free will!? an era of abundance will begin!? the intellectual and motor capabilities of robots have already surpassed the average human and before 2035 there is talk of them surpassing those of any human and then again those of any human combined! regarding both body and mind! answer me please...

No accidents - it's all purposeful release to make themselves look relevant

AI voice or what?

These thoughts were in 2023 and it is now 2024 so in computer time it is as if a dinosaur had thought it up…

Hes just stating a fact. You can't teach an AI CAD level precision by giving it data on human levels of precision.

Opening Eye

Open Ay Eye

Do you ever relisten your videos? They have so many annoying artifacts in the audio.

"...from someone that works at Opening Eye" eh? Shame you're not utilizing the time saved by having AI transcribe your voice for proofreading. Minor gripe, sure, but given that it happens only 5 seconds in, it makes me question your attention to detail.

He's just making a statement he believes to be true and it's not verified. His statement may have nothing to do with q. It could be out of frustration. But even if they developed agi, it doesn't prove yet that these models won't eventually. I doubt it's q agi

Half life 3 confirmed?

If I was in a leading position in AI, I'd have so much fun posting and deleting random ramblings, just to watch the Twitter sphere explode with conspiracies 😂

13:37

👍

OpenAI is not open anymore 😅

The arrogance that these computer programmers "Know better" and know what's good for humanity and at what pace. What gave them the right to hold humanity hostage with this experiment? Why do they know what is good for us?

Hello

❤

He deleted it because it's not true. AI has already outperformed man in Chess and Go. There are no shortcuts to learning. The AI will have to go through stages of progress like all living things. It can only learn the broad strokes from clumsy apes like us. It will have to refine it's abilities over time. I'm guessing quite a short time. . .

your video is skipping.

is there really something like Q* ? Who said that? - Imho SA is always silently laughing his a.. off when asked about it. All speculations about it are bs.