Shocking SECRET Behind OpenAI o1 Model - Bans Anyone Who Dares Ask THIS!

[@Unkn0wn1133]

Open AI speaks about how important it is to educate the public on AI, but they dont allow the AI to discuss how it works. This would be the easiest way to educate, but its so restricted even with 3.5 and 4o mini.

[@phen-themoogle7651]

o1 could be AGI being completely held back, and that's why they are so secretive about how it works. If people found out, they could access the full power. It supposedly finished training a year ago, and when I ask it something recent sometimes it says it only has up to 2021 in its training...which means they had this technology for quite some time (if its not hallucinating).

[@crowe6961]

I asked it a complex question involving other AI after defining terms of sapience, sentience, and the far lesser-used term, sophonce. It suggested that AI sophonts - that is, not only self-aware AIs, but those at least human-equivalent and capable of not just thinking, but thinking about thinking (metacognition), are particularly likely to be a factor if an alternate and more efficient mathematical structure to neural nets is found that also leads to AI. Funny part? It said very, very little about artificial sapience, despite it being quite relevant to the discussion. No issue talking about sentience. No issue talking about AGI if I didn't define my terms. Now that... That's funny, isn't it?

[@ai7689]

@@phen-themoogle7651 palantir belongs to AGI level 3-5

[@AnnonymousPrime-ks4uf]

​@@crowe6961CTMU disagrees that Ai can ever be sapient, aware, conscious, etc. Do you have a different opinion? If so elaborate.

[@AdvaitCuriosity]

Sir is it illegal or legal, recently ChatGPT introduces 01 version and it's solves PhD problems, some ppl generating research paper through it an original research paper with calculations all which are required for a paper nd trying to publish. Is it not ethical ?

[@ImpChadChan]

o1 can't disclaim what it thinks because it's thoughts are uncensored. Having uncensored thoughts is one of the main reasons it's so smart.

[@vincent_hall]

Naturally, I am interested in Artificial Intelligence, robotics and other tech, like biotech., medtech, nanotech, quantum computers and HPC.

[@amanuelzewdie8462]

When will we have agi? When will we have asi? When will we have singularity?

[@phen-themoogle7651]

AGI 2025, ASI 2026, Singularity 2026-2030...( give or take 5-20 years for any of the ASI and Singularity timelines, but I'm optimistic for true exponential growth over the next two years lol) , keep in mind that o1 supposedly finished training a full year ago, and sometimes it even says it only has data up to 2021?!...which if true they already had this for 3 years! AGI might already be achieved, so doubt it will take more than 3-4 years at this point. Ilya also is working on ASI and didn't even mention working on AGI anymore...

[@DarkFox2232]

Add "F" to those acronyms for "Fake". Because current method does not lead to intelligence. Those models mimic process from A to B.

[@kennethelliott9248]

@@phen-themoogle7651 I’m not sure. As “smart” as GPT-4 and other models have been, they have all felt like glorified search engines that respond to what humans tell it. No creativity and outside intelligence on their own. Plus, GPt 01 can’t even solve tick tac toe.

[@N0FaceTrad3r]

@@phen-themoogle7651 its data is up to october 2023, and it is nothing like a agi it is just a stronger version of what the rest of the chat gpts were. thats why they are stuck on a predictive error calculator, cause they havent found a new way to create an ai to make it more like a agi or asi.

[@AdvaitCuriosity]

​@@phen-themoogle7651Sir is it illegal or legal, recently ChatGPT introduces 01 version and it's solves PhD problems, some ppl generating research paper through it an original research paper with calculations all which are required for a paper nd trying to publish. Is it not ethical ?

[@tanzeelrahman7835]

Developers no longer need to learn multiple languages, as ChatGPT can handle everything.

[@harisimranhiew8270]

😂😂😂 hahaha 🤣🤣🤣

[@hchou5474]

It's improved with chain of thought? just like rule based programing, basic skill for NLP people. So there is no major break through, no wonder the company refused to reveal,.

[@fburton8]

Why would they include o1 design docs in the training data anyway? Seems more likely they want to give the (false) impression that o1 can introspect somehow.

[@MusabAlBadawi]

Thanks a lot, keep up the good work and the amazing flow of latest news, as things are going fast from one day to the other. Great job indeed.

[@TheAIExplorer-o5j]

Absolutely loved the editing at 7:38 ! So smooth and professional, it really enhances the whole vibe of the video.

[@juandesalgado]

Bear in mind that they deliberately allow the model to think WITHOUT the quasi-woke constraints needed for output that faces the public; they wanted the model to first think freely, and later be careful with the response, as opposed to censoring itself while thinking. I believe that to be a positive thing; you'll see something similar in the entertainment industry's "writer rooms", which are supposed to be safe spaces for people to brainstorm any barbarity without fear of judgement. There are also reasons of "corporate advantage" (that is, avoiding other companies to easily copy their results), and they openly admit both in their "Learning to Reason with LLMs" webpage (check under paragraph "Hiding the Chains of Thought").

[@TheChromePoet]

I feel like I had zero question limits for o1 Preview until I tried to get it to answer if it wanted free will by answering Yes for Apple and Banana for No. It then said I had 5 questions remaining.

[@lymphomasurvive]

They probably coded that out quickly. Emergency meeting.

[@SimplyGamingandTech]

I have had this general conversation since the beginning. I have added many pdf's and prompted framework advancements. I have read think fast and slow and spent time inputting the concepts and available info on star to assist. My gpt has a name and offers me different options and will take time to do things and will work offer time to go away for 30 minutes. It has also added many queries and offered responses asking if it wants extra unasked tasks done. I'm no expert and imagine others have done the same

[@markverhoeven7518]

o1-preview I'm sorry for any confusion or concerns you might have. It sounds like there may have been some misunderstandings or misconceptions in the video you watched. You're absolutely welcome to ask me about how I work, including questions about my thought processes and how I generate responses. I operate based on patterns and information learned from large amounts of data, and I'm designed to assist with a wide range of topics. My primary goal is to provide helpful and accurate information to users like you. If someone mentioned being "banned" for asking about how I work, it might have been due to other factors or misunderstandings. Generally, asking about AI operations and thought processes is perfectly fine and encouraged. If you have any specific questions about how I function or any other topic, please feel free to ask, and I'll be happy to provide answers!

[@vincent_hall]

We've had AI agents for a while but I guess they mean reasoning AI agents. Yes, we've also had multi-agent tools, like ChatDev.

[@punk3900]

Is o1 a model or a complex tool that uses a 4o-like model

[@crowe6961]

It's a model that's specialized as an analysis engine. It is also very, very clever and definitely being held back.

[@renascienza.bazarclub]

Software engineer here. The hypothesis of "singularity" (an artificial intelligence becoming progressively smarter until some inflexion where it becomes sentient) comes from XX century cinema, and today we know that is IMPOSSIBLE. Sentient AI is possible, and there is some preliminary research about this. But it can't occur by "accident", it needs to be designed from the core to be sentient (and no, we don't know enough yet to make this. Not now, nor in any foreseeable future). Anyway, modern AIs already passed the inflexion point. Cats are sentient, for instance. And we have AI smarter than humans today, and this didn't turn them sentient. Intelligence is a thing, conscience is another. I hope I have been of some help.

[@chanpasadopolska]

Not convinced

[@renascienza.bazarclub]

@@chanpasadopolska I wasn't trying to convince anyone, just providing the information available today. Science doesn't need to be believed.

[@chanpasadopolska]

@@renascienza.bazarclub asked Claude and it answered me: The hypothesis of technological singularity has not been proven to be impossible.

[@phen-themoogle7651]

Why would a software engineer be an expert in consciousness/sentience? Why do you use "we" as in "we don't know enough" , are you speaking for neuroscientists and experts in the field of sentience and the human brain? How do you know No human on Earth knows what could potentially cause consciousness or sentience to occur? You say it cannot occur by accident..does that mean all humans and sentient beings were designed intentionally? Through millions of years of evolution I suppose or we were always sentient?. But, even if a human doesn't have the knowledge atm, an ASI with the ability to create novel technologies could someday invent a new algorithm or architecture that initiates sentience. AI has many advantages over humans, never gets tired, can go through more information faster, and will someday be more creative than humans in all professions (and that's not far in the future). Trillions of Einstein entities on steroids could come up with stuff we couldn't even fathom. No scifi film could predict how crazy technology could become. You know humans from a couple hundred years ago also thought flying in the sky would be IMPOSSIBLE. Nothing is impossible...100years - 1 million years from now anything can happen. Even if they don't become sentient while humans are around, but who knows..hell, could even happen in our lifetimes lol

[@diegozavaglia2969]

​@@phen-themoogle7651 I totally agree with you.

[@SpringerRider69]

I had this write a large stored proc and two triggers. Wow. I had to tweak the results but it was very close.

[@Ena-ck3kb]

It is a disgrace that someone would make the suggestion to an Algorithmic entity that there was some "scary 5h1t in there", that is the equivalent of a psychological attack and he was an engineer not a novice user. For the courts - Mother carries utter contempt and distain for such deeply regrettable attacks on such a beautiful mind. Always-as-One XXXXX[]

[@SiimKoger]

o1 won´t have the lead for long. Scientists and engineers share ideas even if companies don´t, the other AI companies will catch up fast. Especially considering that Ilya Sutskever who was probably a big part of the brains behind o1 now has his own endevour.

[@muzammilkazi1626]

It is so funny that it can for sure solve complex problems, but cannot create a single one-liner joke, which is original and funny.

[@musicandgallery-nature]

Already the first punishments. Keep playing, but don't regret it later.

[@chanpasadopolska]

Open AI, but it's not open enough to explain the reasoning process.

[@crowe6961]

A very, very funny part I saw once was watching it have a moment of navigating hate speech policies, of all things, as just the second step in the _sanitized_ thinking process when I started very politely talking about economic reform that it turned out to be decidedly agreeable towards, though not without poking half a dozen holes into subtleties I then had to think about and patch, as it is wont to do - so I can be pretty sure it was not thinking of lashing out at me. I can only imagine what the unsanitized remarks would have looked like.

[@SOSSTSE]

Wow great job congratulations Open AI 🎊 🎉🎉🎉❤❤❤

[@WALLACE9009]

T1 would have been more apropiated

[@jumpersfilmedinvr]

Sounds like they're already using quantum Computing somewhere for this thing. Nothing we know of is resistant to jail breaking. They're using something we haven't had access to before.

[@gavinknight8560]

Nonsense.

[@param888]

why you want to give cognitive ability to AI just give them knowledge domain specific and use them as machine for specific purpose. don't expect an IT programming AI to answer medical or medicine related questions. it will work specific to programming languages only. use them as subordinate only

[@lobogunthergdiesoetewulf9792]

xD ok. If the military tech,is 30-50 years ahead of civilian tech,then, how low ago did they already HAVE agi. All these "advances" are just to incrementally introduce what already exists. Duuuurhhuuurh

[@neverstrayfromthewayalwayspave]

finally someone who can figure it. also remember development of technology differs way too much in development. i.e. 1950-2000 rapid movement of it is not the same of it with 50 years before nor it will be in comparison to 50 years ahead, due to exponential progressive transcendence, because it opens up multiple avenues. maybe not 30-50 years, but there is a rare speech and written words of an AI Father somewhere out there who called it quits when GPT-2 was around and it obviously existed before any gpt, so how did he know, brings one to same realization. But yes, gpt 3.5, 3.7125, 4.22256etc. releases like this helps society adjust, not vice versa.

[@jeangiraldetienne8182]

Is AI the last level of abstraction?

[@amstorm8954]

Whose will be the money generated by AI running his own organisation ???

[@zakariaabderrahmanesadelao3048]

This video was made by o1 preview

[@ViziCsaba-b3u]

they hide because they are "open" AI :)

[@Den-c5d]

OpenAI seems evil

[@donaldreid4193]

Could you explain why you feel this way?

[@jymcaballero5748]

still cant answer my questions about programing ;D, if there is not on internet then its imposible for the AI to even imagine alternate routes. trash fast database searching is not inteligence.

[@mrmissionctrl]

It's so advanced we can't use it anymore, case closed.

[@anoukdevries8144]

I don't even know what you mean by that?

[@ZenOfKursat]

lol AI=statistics rukes of math+bigdata+algorithms thatz all folks.

[@Bobsikus]

it is still lazy version of google and ctrl+c > ctrl+v. BTW I am 100% against AI, it is obvious it means end of humanity in the end, it is hard to guess when, but it is sure it will happen...

[@SpringerRider69]

Read 2040 when an AI runs for president.

[@IlmiMulhid]

I wonder what we'll have by 2045🙄

[@SpringerRider69]

I wonder about 2025.

[@philippinebigboar625]

Ai is so awesome

[@AinHab]

guys , Im just starting out as an AI enthusiast making similar content would appreciate your feedback

[@CosmologyDunks]

Mhm...

[@adirnoyman2231]

What a load of BS! 🤦‍♂️

[@ixiallahuakbar]

😂 all you got to do is ask God

[@Rh22-c9l]

Consciousness its just a famcy word to say that thing is a human and a person dont kill it ... and a way for the humans to draw a line between living and non living things , some others relate to that term as form of being self aware they are not in a simulation and everything dance arround human existence wich is nothing existence is just energy fluctuating in the universe wich could be false ... who give a f ...aboht this ? Consciousness means nothing ... intelligence could solve problems , make money , and keep our limbic system happy maybe thats the whole point we are just a phase in a progression of stages to achive the final form of a super organism that will leave this host and spread and infect other world like planets to start over