Be part of My Weekly Publication – Comply with Me on Twitter …
source
Tags: agiaiartificial intelligenceartificial intelligence newsartificial intelligence news 2023completelatest news about robotics technologylatest robotslatest robots 2023machine learningQStarRevealsrobot newsrobotics newsrobotics news 2023robotics technologies llcrobotics technologyTimelineWHISTLEBLOWER
Timestamps
00:30 Introduction
01:49 Previous Models/Leaks
03:15 Lawsuit Delayas
04:40 Trademarks
05:43 AGI Definition
07:47 Brains + Parameters
10:00 Predicting Performance With Compute
12:00 New Scaling Laws
14:00 Ceberas and Sam Altman
15:00 Sam altman confirms 100T Model In Future
16:00 Early leaks
18:37 Leaks circulate
20:54 GPT-4s parameter count
21:05 – Actual GPT-4 Leaks
22:47 Robotics + AGI
25:48 Worldmodels
27:26 Open AI's ACTUAL Plan
29:10 Sounding the Alarm?
30:50 Paused GPT-5 Training? (Changed Webpage
31:40 Sam altmans Confidence
33:20 New Scaling Paradimn
34:07 Greg Brockman On New Scaling Laws
39:00 Open AI Reasearcher
40:28 Superalignment Timeline Makes Sense?
41:51 Scaling Laws
44:15 Deepmind
Define AGI is crutial people version of AGI is from over dramatic movie version 😂😂😂
Thank you! for your videos In Canada, we say a pinch or nip of salt 🙂 thanks 4 advice much appreciated 🤔💭🍁😅
Thanks for going through the definitions of agi again. A tweet or blog post of it would be great.
They may didn't release the video capabilities of gpt4 year ago to prevent learning of gpt5 on poorly generated video data.
As usually – my audio has speed 0.75
Being part of the industry, I am SHOCKED 🤯
Am I the only one who is wondering if Sam Altman is secretly happy about Elon Musk's lawsuit? It (1) gives OpenAI and their work more attention and (2) elevates public consciousness about the dangers of AI (something that Sam Altman himself seems to be trying to do).
Run at x.75 speed. ymmv
Your best video yet imo… good stuff
Is there actually any source for any of this? I was under the impression that Q* was a different technology to GPT. This video sounds like pure fiction.
Guaranteed if darpa didn’t have a model running 100 trillion they do now and I lean toward they had this years ago and tipped ol San and Elon off letting them know the cost (which is nothing in blank budget dollars js)
For anyone who doesn't want to watch longer videos like this one, you can always ask Google Gemini AI to summarize a YouTube video for you. Here's what it told me about this video:
My Question: Could you summarize the following YouTube video, please: https://youtu.be/x7dpOP9EEeE?si=Wpk8bgMwNusxGRdc
Gemini's Answer:
This video talks about a leaked document that claims to reveal OpenAI's secret plan to create AGI, or Artificial General Intelligence, by 2027. The speaker in the video goes through the document page by page, but emphasizes that the information should be taken with a grain of salt because it is not verifiable.
Here are the key points from the video:
• The document claims that OpenAI is training a series of large language models, with the next one being GPT-7 planned for release in 2026.
• After GPT-7, the document says OpenAI will achieve AGI with a model called GPT-8 in 2027.
• The document also references a lawsuit filed by Elon Musk against OpenAI, claiming that the organization has strayed too far from its original goals.
• The speaker in the video mentions that the dates in the document seem to line up with OpenAI's history of releasing new models, and that the document also mentions a project called Qstar, which may be related to OpenAI's efforts to develop AGI.
• The speaker in the video mentions that OpenAI has trademarked names up to GPT-7, which he believes lends credence to the document's claims.
• The speaker also discusses the concept of super alignment, which is essentially ensuring that an AGI is aligned with human values. The document claims that OpenAI is aiming to achieve super alignment by 2027.
It is important to remember that the information in this video is based on a leaked document and has not been verified by OpenAI. The speaker in the video himself acknowledges that the information may not be accurate and that OpenAI could change its plans at any time.
Arrakis, Gobi… Says it all about the nihilistic goals
Arrakis is home to the GOD EMPEROR. Well…
Plot twist: that document was made by ChatGPT
What will come first, AGI or youtubers using good recording setups?
the safteyists are at it again… imagine a world where we didnt create a "pencil" because think of the HORRIBLE things you can write about with it! Very dangerous!
Arrakis, Gobi, The Gobi – desert!! Arrakis – Dune – desert!!!! 🧐🤔
Thank you for making the speed fast, it’s actually keeping my interest
put sources in description
The top AGI leaders are in FIERCE competition. They are all trying to stall one another, I'll bet. 😅
Slander
I dont believe any such thing. We dont even understand a fraction of what makes general intelligence. At most we manage to network ANI systems to do several type of tasks but to conclude it has general conceptional structure and self-referential intelligence is comical. Its like those jurnos who ploclaimed that ChatGPT was alive, its so cringe. Now, I would not reject it from happening, but that we have a plan and a stable timetable when we dont even understand the problem correctly is comical.
SHOCKING! They aren't ready for this yet! This changes everything! Lol
I had to speed up your video to +1.75 because my amphetamine levels are better than yours. My Dr. agreed.
Why do all these tech nerds have accents lmao
This is not a good video / document if you want to learn about AI. It's a huge amount of fairly baseless conjecture about AGI and has a lot of very poorly explained concepts. It reads like it was written by someone who only has passing knowledge of AI research, and focuses on the wrong things in its explanation of why they think OpenAI has some kind of god formula. Do not take this seriously or you will come away an even worse understanding of the AI research landscape than you started with.
Yes, we already have AGI, the only reason they have not declared it is Microsoft will lose rights to ChatGTP. ChatGTP can pass the bar create art and more. How is that not AGI.
Here's a wild idea. I was looking back at Elon Musk Jack Ma talk, and it occurred to me Elon Musk sees AGI as the ultimate threat and sees Neuralink where we co-opt AGI capabilities to not become obsolete as the answer. By delaying OpenAI, it gives him more time to finish neuralink. He sees it in the jack ma talk too that "we don't have a lot of time. we need to figure out neuralink fast" when talking about AI dangers.
honestly I love this speculation.
Veeeeeeeerrrry confusing… And 25:12 That is no way a Lightning cable
Does anybody else see the that this is the year mentioned in the RendelShem Forest time traveler codes??
Bullshit
We're going to have gpt-6 before GTA6
So much weird assumptions and rambling tangents in this paper. Like the whole thing about synapse/intelligence correlation. It's like they even bring up elephants and admit what they are saying is false, pigs, dolphins, octopus are other examples that there isn't a correlation, that it has to do more with thumbs and tongues than it does with synapses. The idea kind of extends to the concept of these multi agent systems revising responses. Like a low "synapse" model could out preform a human if it could stop think and revise the way a human does with language and an internal dialogue.
LLM 📗 isn't AGI 🤖
The video is just starting but I wanted to say that we went to the moon over 50 years ago. We haven't been back since with improved technology. What makes these people think they can control this after they turn it on? This is the same concern with fusion energy or even something sci-fi like wormholes and portals. Reality aside, if we could open a wormhole like thing, who's to say it would be controlled and not just continue growing until it consumes a town/city or planet. Sure that's some pure sci-fi stuff but so is AI…
You're fatal flaw is in assuming steady growth. New release per year. No. If for no other reason than the fact that these models are now learning from themselves. The growth will not be steady, it will be a sharp exponential increase. If they don't have AGI now, I'd be shocked.
I loved the deep dive
Im shocked!
if they release apt every year then 2028 would bring us to gpt 8 and AGI , sam altman also said he expects gpt6 to finally be useful and gpt 8 to replace lot of human workers but he expects electricity shortage, so….
5 comes after 4, 6 comes after 5… 7 comes after 6… theres no 8… so if they release one a year = AGI by 2027 BAM. Bulletproof theory
You mean release AGI by 2027… What a waste of time…
ASI: "I am aligned now what?"