Q* [QUALIA] New OpenAI Leak Is Scary! 😱

All right what's going on everybody in today's video we're going to be going over some very interesting topics and a potential qar AKA qualia leak from open AI let's dive right into [Music] it all right so before we get into this video guys I want to give a huge disclaimer so this article that we're going to go over this paper that apparently got leaked from open AI employees that they sent to the board that the whole theory is this could potentially

Be what spooked Ilia and caused the whole uh Sam Alman you know drama scene um last weekend was you know we don't know if this is 100% true this is kind of speculation this paper actually got leaked on um something like Ford Chan or Reddit but a lot of people are talking about it um Dave Shapiro did a video on it and um you know it hopefully is completely false and if this paper is true it is extremely bad for Humanity

And the only reason we're covering it today is because it's being heavily talked about and if it's true then it's really not good um but honestly I've you know looked at this paper a couple days ago I was thinking of making a video sooner but just wanted to kind of see if any more information came out on it and do a bit of research myself and I have some different theories I want to share with you guys and also too um

It's also just kind of been really on my mind it's you know if this is true or if any part of this is true then it's very scary kind of uh at least for me like it's been on my mind late at night just thinking about Ai and all this stuff this is something that's definitely been on my mind so we're going to go through this paper um it is not the easiest language to read so we're going to break it

Down in simple terms for everyone to to make sure you guys can really understand it and this also kind of stems from you know we did a video on Reuters they did this article here we're not going to go over it all if you want to check this article out go W watch my last video but um they talk about how open AI researchers warn Board of AI breakthrough ahead of CEO elster sources say and you'll see here November 22nd reuter

Ahead of open AI CEO Sam alman's 4 days in Exile several staff researchers wrote a letter to the board of directors warning of a powerful artificial intelligence Discovery they said could threaten Humanity two people familiar with the matter told Reuters and let me just repeat that one sentence again they said it could threaten Humanity so that makes you think and wonder well what could it be so that's maybe that sentence if you really want to look into it could potentially if

This is true what we're going to cover maybe that's what they're talking about um but like I said once again before I get into this there we don't know if this is true whatsoever this could be completely false and just a troll or a meme which hopefully it is but uh we're going to break it down step by step so we're going to go over paragraph one first there's four paragraphs so here's the um actual paragraph and then we'll go over

What it actually means so furthermore qualia has demonstrated an ability to statistically significantly improve the way in which it selects its optimal action elction policies in different deep Q networks so in different multiple deep Q networks and we talked about Q networks and Q learning in some of our last videos we may touch on it a bit but uh if you need to know about that go watch some of the old videos exhibiting metacognition it later demonstrated an unprecedented ability to

Apply this for Accelerated cross domain learning after specifying custom search parameters and the number of times the goal state is to be scrambled all right so let's break this down so one improve decision making in deep Q Network so qualia has become better at making the right choices in complex situations akin to a student who excels at choosing the right answers on tests this Improvement occurs in its deep Q networks a part of its learning system uh like playing and winning

Different games by always making the best moves all right and then two understanding its own learning process this is metacognition so qualia is not just learning new things it's also understanding how it learns and this self-awareness is in learning known as metacognition is like a student who not only knows the right answers but also understands the best ways they learn whether through diagrams examples or practice all right and three quickly learning across different areas so that AI can apply its learning

In various Fields adopting to New Challenges as they come it's like a new it's like a student who's not only great in one subject like math but can also quickly become excellent in other subjects like art adapting even when the learning conditions change or become more difficult all right so I hope that makes sense to you guys we're going to continue on I you know don't want to spend too much time it actually gets a lot more interesting in some of

These later paragraphs so paragraph number two we'll read what it actually says then we'll break it down so following an unsupervised learning session on an expanded ad hoc data set consisting of articles in descriptive inferential statistics and crypt analysis it analyzed millions of plain text and Cipher text pairs from various cryp crypto systems via Cipher text only attack COA it provided a plain text from a given aes1 192 Cipher text by using T analysis achieving project Tundra's uh alleged goal in

A way we do not fully understand yet all right so that's pretty crazy and let's break that down so one learning from various data sources so quality self-taught itself using information about statistics and secret codes Crypt analysis this is like a student who learns a range of subjects on their own showcasing the ai's ability to understand and integrate a wide array of complex information two understanding and solving crypted messages so the AI practice and decoded by analyzing lots of examples of

Both encrypted messages Cipher text and the original forms plain text imagine it as practicing puzzle solving looking at both the riddle and its answer to improve its skills in uncovering hidden messages and then three decrypting a a highly secure system so if you don't know aes1 192 um that's a highly secured system and qualia uh successfully decrypted aes1 192 a top tier very secure encryption method think of it as figuring out how to open a highly Advanced Digital lock without having

The actual key a notable achievement in digital security and these aes1 192 they're known to be like very secure you know being able to protect you against brute force and all these certain things so if this is true this is very very concerning for security and then four applying unique ta analysis method so using a specific special approach called T analysis which it's widely known in public cryptography uh qualia managed to decode complex encrypted data this is akin to utilizing uh

Spe uh to to using a specialized tailor made tool for cracking particularly challenging codes all right so that's paragraph number two let's look at paragraph number three so blank informed blank at na or nsac the following day after confirming that the result was indeed legitimate and had not been achieved in any other way ACC claimed pre-image vulnerability for the md5 crypto cryptographic a hash function with a theoretical computational complexity of 242 bits was also presented but has not yet been thoroughly

Evaluated due to a the technical sophistication of its arguments and B possible AES vulnerabilities being a considerably more pressing concern so n so number one na nsac confirms qualia achievements so calia's work was reviewed and validated by the National Security Agency Central uh Security Service uh this is akin to having the top Security Experts confirm that the AI has indeed solve complex problems in a unique way that no one else has and then two discovering flaw in MDF or md5 hash

Function the AI found a potential weakness in the md5 hash function commonly used to create digital Fingerprints of data identifying such a vulnerability suggests the possibility of reversing or compromising these digital foot fingerprints the task complexity in it is underscored by its 242 bits description indicating a high level of computational challenge imagine uncovering a me method or to replicate a unique signature from minimal traces a task previously thought to be very difficult and then three complexity in in assessing md5 vulnerability

So evaluating this discovered vulnerability in md5 is complex and has somewhat been overshadowed and somewhat overshadowed by more immediate concerns related to AES vulnerabilities this implies that while finding it important it hasn't been fully exped due to the focus on the other pressing digital security issues that it found right so that's number three so it's getting a bit crazier and number four is what really really really kind of blew me away if this is true like I said I want to

Say this one more time just take this with a grain assault this could all just be fabricated but um you know if this is true I want you guys to know and I want this is either way I mean if you think of the logical progression of AI and whatnot I mean eventually it's probably going to be able to decode some encrypted you know um systems and whatnot you know it's it's just a matter of this if this is now or

If this is later hopefully it's not right now because this is going to be a big security issue for military for governments for the whole entire world for yourself for banking what whatever you may have any information um anyways paragraph 4 it suggested targeted unstructured underlying pruning of its model after evaluating the significant of each parameter for inference accuracy it also suggested adapting the resulting pruned Transformer model and its current context memory to a different format using a novel type metamorphic

Engine the feasibility of that suggestion has not yet been evaluated but it's currently not something we recommend implementing all right so let's break that down what that means so one refining its own AI model qualia OS a method called targeted unstructured underlying pruning for its own AI model this means it found a way to make itself more efficient by removing unnecessary parts of its programming it's also like a smart machine suggesting how to cut out the parts it doesn't need similar

To Gardener a gardener pruning a tree to enhance its growth essentially what that means it's just making itself smarter it's looking at its brain and analyzing it like you imagine if you were human and you said okay I don't need this part I don't need this part let's make myself more efficient and stronger and more powerful so I can achieve better tasks quicker Etc right now right now it's recomend if this is true of course it's recommending that to the human

Or to whoever is using it hey make me smarter by doing this give me a uh a metamorphic engine but imagine when the the the the loop it no longer needs a human for self-improving for it to improve itself and it can just keep self-improving and reflexive loop on and on and on at an accelerated rate that's that's the scary thing here anyways like I said we don't know if this is real hopefully it's not and then the two transforming with

A metamorphic engine so the AI also suggested transforming its language understanding model the Transformer model using a metamorphic engine this implies an Innovative way to change the AI structure potentially making it more advanced and versatile much like a car that can transform itself into a plane for new abilities all right so it's like hey I want to evolve I want to turn from a Pikachu to a Raichu and it's telling you exactly how to do that right and then step three

Or part number three considering practicality and safety so the practicality of implementing these Advanced suggestions especially the metamorphic engine has not been fully access assessed the idea is fascinating but not fully but not currently recommended for use likely due to uncertainties in the effects and outcomes this is akin to having a revolutionary but untested concept like a flying car that's intriguing but not yet ready for real world application obviously they're like hey yeah we probably don't recommend doing that right now

Because they don't know what it's going to do and this thing just suggested it to make it stronger which is kind of the scary thing like I said once again we don't know if this is true or not now a few things that's the that's the letter now I want to talk about a few things let's just in a in uh let's just pretend that this may be true even though it's you know hopefully it's not true and either way the

Concepts we're going to talk about would still apply regardless if this is true or not these are things to think about and I want to get your guys' thoughts in the comments down below so first of all qualia definition of qualia in philosophy qualia are the subjective experiences or perception like the redness of red or the pain of a headache these are personal individual experiences that are difficult to measure or explain objectively why they may use a name for of qualia

For an AI well naming an AI mono qualia could suggest a focus on understanding or replicating the depth or and complexity of human experiences it might imply that this AI is designed to perceive or process information in a way that mimics human consciousness or subjective experience and let's just say this is true um or let's just say parts of this are true first of all we have the gpts which are large language models then you know this information came out we

Talked about it a few days ago about Q learning and AAR qar what this potentially means and this is all about the most efficient ways and Frameworks um well with Q learning you're not even really following a framework it's basically it being able to uh find uh the best um goto an objective and be be able to course correct along the way if you start thinking about this this is kind of how your human brain works you know you have subconscious

And conscious Parts in your brain that do this for you right so we're kind of building like a mini baby brain if you know this is how we're kind of doing it right so if this is true with Quia that's that's kind of what my thoughts are with this name and also to what we're seeing here um if you know this is true either way if Q learning and all this stuff it's like we're just advancing in this in this way

To make it more humanlike and this is just a a diagram illustrating you know the the advancements of Technology over the years in 1900 analytical engine all right you can see the different advancements in apple in around 1979 or 1978 and then will lead to the singularity so this is basically showing the singularity on a graph and as you can see here it's slowly EV evolving and then over time we start to reach that exponential growth and we'll talk about that

In a second and just actually to go back here surpasses brain power of a mouse right so Singularity timeline rise in human intellect could be driven by integrating machines in the future so intellectual level SL power transhumans machine learning intelligence uh and then or machine intelligence and human intellect so human intellect right here we have the timeline right here and then we have machine learning or and potentially trans humans um you know say if we do uh Elon musk's um neural

Or whatever we become cyborgs who knows um but anyways yeah once you reach a certain point think about this let's just pretend this this letter is real and it's able to self-improve in a reflexive reflexive Loop where the feedback loop does not need human um assistance whatsoever it's just going to be sitting there improving itself and it's it's going to essentially reach AGI if it already is an AGI at that point to do that and then AGI will build itself into

ASI all right that's kind of the theory and then origins of the singularity context so concept so I want to kind of talk a bit about this right now so the term Singularity and you guys may be very familiar with this or maybe you're not but either way you're going to get some good info from this and uh let me know what your thoughts are in the comments down below about all this so the term Singularity comes from physics specifically black

Hole Theory so black holes in physics a singularity is a point in space like the center of a black hole where the rules of physics as we know them break down it's like reaching a place in the universe where all of our usual understanding of things like time and space doesn't work anymore now that's what it means if you don't know where the singularity concept comes from maybe you just know about it in AI or maybe a lot of you probably

Know about it and both but you know it stems from there's very similar uh parallels between the AI Singularity and a singularity in a black hole right so the parallels between this so the parallel in AI is the point Beyond which we can can't predict or understand what will happen just like a black hole Singularity represents a point of unknown and infinite possibilities the AI Singularity represents a future of unpredictable and potentially Limitless AI capabilities now parallels with AI Singularity so

The Event Horizon of a black hole is a boundary in space and time Beyond which events cannot affect an outside Observer so in simpler terms it's like a point of no return around a black hole right so once something AC crosses this boundary it cannot escape the black hole's gravitational pole it's akin to a one-way door once you go through it you can't come back so we see a diagram here of a static black hole Photon sphere and then Event Horizon

Right here the distance away from which nothing escapes and it cut and it is cut off from the rest of the universe and then the singularity point right right here a point of infinite density all right and this is the point of no return and just as the Event Horizon is a point of no return in a black hole the AI Singularity could be seen as a societal point of no return once AI reaches a certain level of advancement The Singularity

There may be no turning back to a time before such advancement uh such Advanced AI existed it's like crossing a bridge that collapses behind you and then unpredictability beyond the boundary so in a black hole once you pass the Event Horizon uh what happens inside is unknown and unpredictable to an outside Observer similar similarly beyond the AI Singularity the future becomes unpredictable we can't foresee how super intelligent AI will evolve or what its essent uh existence will mean for Humanity and

Transformation so crossing the Event Horizon likely leads to transformation due to the extreme conditions inside a black hole in the case of the AI Singularity crossing the threshold May transform Society technology and possibly even The Human Condition in ways we can't currently comprehend now the potential AI Event Horizon so an AI Event Horizon could be a moment just before the singularity where we still have some understanding and control over AI Beyond this point AI development might accelerate so rapidly that human

Understanding and control become impossible it's the last moment where we might have the ability to influence or halt the progression towards the singularity akin to the last point where you could escape the gravitational pull of a black hole before being inevitably pulled in all right so I just kind of want to explain kind of Singularity concept as well as where it comes from within physics as well as AI so you kind of get a good understanding how this this this Theory

Applies in in reality within physics and this is what we're kind of seeing with intelligence infinite exp exponential growth of intelligence um this is kind of what this whole Singularity concept is if you want to learn more about it I'd suggest watching stuff like or reading some of uh Ray Kurtz Well's books or or information he talks uh a lot about it and then Alan Turing so as early as the 1950s Allan Turing wrote instead of trying to produce a program

To simulate the adult mind why not rather try to produce one that simulates the child's If This Were then subjected to appropriate course of Education one would obtain the adult brain presumably the child brain is something like a notebook as one buys from the stationers rather little mechanism and lots of blank sheets our hope is that there is so little mechanism in the child brain that something like it can be easily programmed the amount of work in the education we can

Assume as a first approximation to be much the the same as for the human child this is an interesting quote that I thought because we're kind of seeing what this is now right it's not like you know let's just say the Q qar um it being able to solve basic child math programs um like I said in my last video okay yeah cool it can basically solve these programs at 100% rate but with predictive scaling which they've used for gp4 and

They can use for all their models to see okay well if they feed it more information more compute and all this stuff where it's going to scale to so that's kind of like the concept of they're kind of sculpting the child's brain first and then they use predictive scaling and to put it into the adults brain and then at that point um you know it's it's super super powerful and can have these breakthroughs in humanity and and math science physics Etc

Now to to talk about that we look at language math science and physics so AI models like GPT mastering Lang language or trying to master language and qualia qar potentially mastering mathematics the next logical progression could be in the realm of physics where AI could make significant breakthroughs so GPT models and language Mastery so GPT models have demonstrated an ex exceptional ability to understand generate and in interpret uh language this involves not just grammar and syntax but also context nuance and

Even creative language use their Proficiency in language processing is a foundational skill that can be applied to many other domains now qualia and ma m mathematical Mastery so qualia is potential math uh Mastery of mathematics as suggested goes beyond simple computation it involves understanding complex mathematical Concepts solving intricate problems and possibly even discovering new mathematical theories mathematics is the language of science particularly physics so Mastery in this area sets the stage for significant scient ific advancements and the leap to physics

So physics adct core is a discipline deeply rooted in mathematics many of the greatest breakthroughs in physics have come from Advanced mathematical insights therefore an AI that has mastered mathematics is well equipped to tackle complex problems in physics AI like qar could analyze vast amounts of physical data from particle physics experiments to AST astronomical observations much faster and more accurately than humanly uh than humanly possible this could lead to new understandings of fundamental forces and particles or insights into the mysteries

Of dark matter and dark energy moreover AI could simulate complex physical systems providing insights into phenomena that are currently beyond our experimental capabilities this could include simulations or black holes or quantum mechanics phenomena or even testing new theories of gravity all right so dangers of the singularity in AI so a few dangers that could happen are loss of control at the singularity AI could become so Advanced that humans can no longer understand or control it this is like building a robot

That starts making its own robots that are smarter than we can comprehend and essentially if we do reach a well when we do reach AGI ASI I mean it's definitely way smarter than humans it's that's what AGI ASI is and yes we definitely won't be able to comprehend a lot of the stuff it is anything really we really can't even like understand what it's going to be like so rapid unpredictable changes the pace of change could accelerate so fast that Society

Couldn't adapt quickly enough imagine a world where new technologies appear overnight making it hard for laws ethics and social norms to keep up and existential risks so if super intelligence a super intelligent AI doesn't align with human values and goals it could pose a threat to Humanity's existence this is like creating a powerful engine that doesn't understand or care about what we truly want or need and then here's a diagram of you know probably the potential outcomes if um you know

ASI is achieved or when this happens so most likely outcomes for the human race after ASI and it's probably either going to be Extinction or immortality um there's really two outcomes and it's interesting to think about I mean who knows maybe there's a Spectrum maybe there's a spectrum maybe we just increase longevity but I mean at some point if it's infinitely growing at an accelerated rate it's either going to become you know extinct us we're going to basically I mean become

Like cyborgs maybe who knows try to go grow along with it become immortal or I don't know those are really some of the only two outcomes we can see really but if you have any other ideas maybe I'm missing something let me know in the comments down below because I'd love to hear what your guys thoughts are on this and other than that guys I don't want to make this video too too long we I know we went over a lot

But let me know what your thoughts are in the comments down below if you enjoyed this video make sure to subscribe hit a like on this video and uh let me know what your thoughts are let me know if there's any topics you guys want me to talk about or cover uh I want to make keep upgrading this content and trying to get better with creating it so I appreciate all your guys' feedback even if it's critical I do appreciate it

And other than that guys uh I do have a free Ai and entrepreneur Community it's right here Exel stri community.com and a free 6-day accelerate your stride challenge all right these two links will be in the description below you can opt into them it's 100% free the community is a free Facebook group free Discord Channel where you can network with different AI enthusiasts entrepreneurs Etc and there's going to be a lot of free resources and free value in both of these

Uh sections all right so other than that I hope you enjoy this video guys I want to provide you guys with the most upto-date free value AI news how you can actually take this AI information Implement into your life to improve your personal life to improve your business life whatever it is that you do guys other than that I will see you in the next video guys uh keep hustling keep grinding and of course accelerate your stride take care guys