With Q*, it's not just about finding answers; it's about discovering possibilities we didn't even know existed! Creativity coming to AI.

@mastercc4509

6 ай бұрын

If the leak is real it created novel math to solve a problem. This is the vehicle and road to an ASI.

@antonystringfellow5152

6 ай бұрын

He seems to understand the principles well and gives the viewer a good, clear explanation but he doesn't seem to have connected the dots. Maybe he needs to think about it over the weekend (hint: this is not a minor performance boost).

@dheeraj7386

6 ай бұрын

Doesn't this coincide with the concept of quantum computing?

@user-cn8nu6lq4w

6 ай бұрын

Job and income elimination coming to America.

@naos313

6 ай бұрын

@@user-cn8nu6lq4wJust America ? I don't think so, this will impact europe too . and maybe all around the globe

Funny thing: If Q* is *too* creative, we might not reward it for getting us in the right direction...because we can't comprehend how it moves us in the right direction. (Like many people would have tried to "fix" move 37)

@chadchristiansen2690

6 ай бұрын

Great insight

@anuragshah3433

6 ай бұрын

Good point. There can be a massive reward on winning the game which will comprehend such scenarios.

@aaronotto7010

6 ай бұрын

Crust of the perceived threat.

I love how ai is advancing more in a day than it used to in several years

@edmfuturedisco3485

6 ай бұрын

It feels that way but Open Ai came out in 2018 and every year they release updates just like phones. 👍

@rastkogojgic699

6 ай бұрын

​@@edmfuturedisco3485Well not really, some od the first mathematical models for AI were laid in 1930s. But these ideas had to wait for more powerful processors.

What’s will be crazy is when AI becomes able think of things humans haven’t before. When it starts getting creative and inventing things that have never existed and we cannot comprehend, things will truly get interesting… and scary

@nindoninshu

6 ай бұрын

no need to fear anything

@joeykillbill

2 ай бұрын

AGI is coming out in months (unless for unforseen delays). We will find out 😅.

Timestamps: 00:02 Q learning is a type of machine learning used in reinforcement learning. 01:54 Q learning helps computers learn and improve by finding optimal solutions. 03:39 Q-learning is a learning process that helps an agent make optimal decisions in an environment. 05:29 Q* (Qstar) is being explored as a viable option for the future of large language models. 07:24 Limitations of large language models 09:14 Traditional LLMs have limitations like static knowledge, lack of context understanding, and biases. 11:01 Q learning (Q*) has dynamic learning and optimization of decisions, making it suitable for goal-oriented tasks. 12:53 Researchers are exploring advanced techniques to overcome the limitations of standard AI methods. 14:48 Gemini is currently delayed and it will be interesting to see how GPT-5 compares to GPT-4 and whether it will contain Q*.

The optimal Q-Value (for a given state-action pair) is called Q* in machine learning. I have no clue who came up with this A* crap and why, but probably someone who has never heard of Q-learning or even reinforcement learning (the most common form of reinforcement learning is Q-learning). So instead of reading an introductory book on reinforcement learning, they just googled around and accidentally bumped into this A* stuff.

Great video, I keep hearing Q* everywhere lately...thanks for the explanations🙏

@TheAiGrid

6 ай бұрын

Glad it was helpful!

What many people don't get is that GPT 4 was build before the release of GPT 3,5 and all the hype. At that time not many people or Companies worked on AI. Now every major company and country is working on AI with billions of dollars investments from all sides. Still people are saying there won't be an exponential growth or explosion and AGI is 5-10 year away from us. It's delusional at most really.

@Arthur-jg4ji

6 ай бұрын

yup there are not really aware because they don't what ai can be capable

@4013368

6 ай бұрын

kzread.info/dash/bejne/nmGe0tajnqm8ic4.htmlsi=IDbY9wFOdZn1nwG9

@EdFormer

6 ай бұрын

Lol except the AI winters of the early 70s and the late 80s, AI research has been prolific. The only real major recent breakthrough was the ChatGPT GUI that opened up the use of AI to people who can't code.

@user-cn8nu6lq4w

6 ай бұрын

Lot of people say it's 50-100 years away, lol. That said, trillions of dollars have been poured into cancer research, so that isn't really an argument.

@rocksteel9087

6 ай бұрын

Now how do you combine LLM's With Q* learning I'm thinking left brain right Brain.

Just nimble, compact and intriguing introduction, thanks.

0:50 What do you mean the A* paper was written in 2019? It was published in 1968.

5:20 this is like how the feedback system in a school is the teachers answering the kids questions or correction the behaviors kinda. Human feedback systems are variable and intertwined though so I’m guessing that’s what the ai has achieved, the ability to intertwine the understanding and feedback

best explanation so far and I have listened to many well done

Great video.. I feel they should add step 7 as ‘Value’ as well to ensure the whole framework : agents, actions, goals everything is aligned with responsible AI n betterment of humanity values in-built in model so that there are less chances of misuse by design.

The title of your last video was “official: agi achieved”. Id love to know why we need a breakthrough in future LLMs in that case 😂

@TheAiGrid

6 ай бұрын

Sam Altman exact comment on Reddit was ‘internal agi achieved ‘ then he took it back…

* is also used in programming when you want to select all elements, for example in SQL. So maybe it could also represent open source materials?

7:20 on creativity: possibilities are infinite, it is beyond simple to generate a random thing... but that thing is likely to be garbage so it isnt about 'newness' it is about a higher level of selecting from just what we would evaluate as 'appropriate' or 'good'. if we had that space of what is 'good' we could randomly pick from it, we dont need search.

@JordanMetroidManiac

6 ай бұрын

What is artistically “good” is based on how the average human brain functions. Given that we hardly understand how the human brain does what it does, it makes sense that deep neural networks can mimic some of the capabilities of the brain that are not well-understood, leading us to believe it exhibits some form of creativity even if it isn’t.

Awesome video and very well explained! at 7:20 mark he seems to be explaining "better search" which should be intuition which is a living force of some kind. Perhaps in the future quentum AI will pick up intuition, but as of now 1 and 0 proly cant apart from randomness induced in the machine by a intuition capable living being

Move 37 actually came from a complex trillion of trial and error algorithmic experience, it's vast quick learning is just a method it follows. I think we hit the benchmark AGI

@weloveblakegriff

6 ай бұрын

That's the beauty of it all. "YOU" think we have hit the benchmark... but that's just what the human mind thinks. Essentially what this is capable of is something more than what the human mind can find possible.

This is all speculation by a non insider. It is particularly unhelpful not to mention the 'tree of thoughts" approach in this context which is a similar algorithm already applied to GPT.

Great report!!!

I'm DCAing in AMS90X as well. ETH heavier DCA and ALGO. I'm taking your advice and starting Google tomorrow with a 50 dollar purchase and continuing Microsoft and Apple. VTI and VOO on another app and longterm portfolio. Here we go family!

reinforcment learning is not learning in a maze kind of environment because there is nothing to learn in a maze, the best thing it can is to memorize the way out OR if there is a special thing about this set of mazes that needs to be learn OR if the goal is the quickest time the algorithm will learn to minimize the time it goes in circels to zero.

This can lead to AGI, however to apply this on one game is a thing to make this general or multiples "games" at once will be a nightmare, maybe even worse extrapolating one to another should be insanely hard.

@antonystringfellow5152

6 ай бұрын

Q-learning can already do that. It allows introspection and transfer of knowledge between tasks and in a changing environment. Q-learning plus A* has the potential to create ASI rather quickly, given the capacity. It may also be able to crack military-grade encryption. This could explain the internal panic that led to Sam's firing.

@l.lawliet164

6 ай бұрын

That's not true you don't have any super machine which learned a game like chess and them you put to play checkers and they just perform well... this is way harder and have no precedent. We have examples when they teach chess normally, but create a new environment for the start positions of the the pieces and put the super player to the new conditions they perform poorly like they don't even know the game. So you are basically wrong. We don't have concrete examples of this general thinking. @@antonystringfellow5152

@BachelorJasper-lb3oq

6 ай бұрын

what's the meaning about AGI?

@l.lawliet164

6 ай бұрын

Adaptation and evolution of reason based on new situations and new data. You can continue your reasoning without collapsing into oblivion. @@BachelorJasper-lb3oq

Question about "static knowledge". When I do a search on got GPT-4 it now says "searching x" and "searching y". Does that mean that it has moved past static knowledge and is refreshing it's model with up to date information? Thanks

@JordanMetroidManiac

6 ай бұрын

I don’t know if GPT-4 does that, but Google Bard does (and it’s free to use). Bard specializes in collecting information from the Internet to then give you a comprehensive response.

"Future proves past." This is the best timeline.

Oh snap i just thought of how the ai can use the brain wave scanners to see the patterns in the waves and understand us better now.

He's so happy about the Recession coming in, like he's super excited to be witnessing it.

Humans: Ok, took me a while but I found the shortest way! AI: Ya, we just flipped over the pencil and erased a few walls. Done. Humans: that's cheating! AI: Who's cheating? I prefer the term 'optimizing my path to success'.

Great video

Who knows for sure where the name Q* comes from but if you look at the Nature paper with Hassabis as the last author: Human-level control through deep reinforcement learning (doi:10.1038/nature14236) which is about an AI playing Atari games, and surely one of the Deep Mind pieces of crown jewelry, we find the function Q*(s,a) being called 'the optimal action-value function'. So maybe the name Q* was lifted right out of that paper. Of course maybe the Deep Mind guys got it from somewhere else, but there it sits in equation 1 of the Atari paper.

@jan7356

6 ай бұрын

Q* was always the optimal Q-Value. Q-learning is a basic form (and the most common form) of reinforcement learning. This stuff is even in introductory books. The nature paper didn’t come up with the term Q*. It’s like finding some paper that names the unknown variable of an equation x. They didn’t come up with using x for an unknown variable in field y using technique z. It’s just always called x. No matter what technique, no matter the field.

@johnsaeger4416

6 ай бұрын

Yeah I think you're right. Shows up in Watkin's PhD thesis from 1989. I guess that qualifies as 'always'.@@jan7356

3:59 that the part I am trying to understand...concerning rewards... HOW does it actually know how much to reward itself every time it makes a state change? What signals or queues are generated and where do they come from?

Did Alpha Go use Q learning or Q* methodology? What was the strategy called?

The name "QAR" likely comes from the Q learning and A* search algorithms that are used in reinforcement learning and pathfinding and graph traversal algorithms. Q learning involves an environment and an agent, states and actions, a Q table, learning by doing, and updating the Q table. The Q table contains the best actions to take in each state, and is updated using a formula that considers current and potential future rewards. OpenAI's potential breakthrough involves using Q learning to train large language models with a reinforcement learning approach, allowing them to learn and improve from their experience. This approach can solve complex problems and find the best solutions, similar to how one might figure out the best way to beat a video game. The six key steps in understanding Q learning are: environment and agent, states and actions, Q table, learning by doing, updating the Q table, and reinforcement learning. Q learning is like training a pet, where positive actions are rewarded and negative actions are penalized.

Guys, stop getting hyped for nothing and start building stuff instead

❤😊 well explanation for beginners

I think also that Q* enables small language models to perform better than LLMs without the Q* ?

It looks like it comes from a sinple deep learning model... but of course absolutely improved with the last knowledge progresses

Part of a very important coin been talked about in the BCL

Remember there's not enough agreement about what 'AGI' means for the announcement to mean anything remotely similar to all the people who will hear it. It seems to make a huge promise without being clear about what exactly the promise is. It seems to make a huge threat without being clear about what exactly the threat is. Ask a hundred people and it means a hundred things. Make an announcement and those people will understand you to mean a hundred different things. At least nine out of ten will be wrong no matter what the person making an announcement actually means. So if this isn't a swindle or fearmongering then why do they want to be misunderstood?

@JordanMetroidManiac

6 ай бұрын

Excellent point. Has anyone from OpenAI openly acknowledged the leakage of Q*, or are they ignoring it? If they acknowledge it, then your point stands. If they don’t, then it could be a hoax.

Q* refers to a Quantum Singularity core AGI: IMO.

i would be fairly skeptical of this, especially with how Q* was described. it seems like they are just pushing the computation side of things harder and more efficiently by using a reward model instead of a pde model. it lets you do interesting stuff with tree search so u add A* to this where the heuristic is probably some sort of pde model (could even be gpt-4) and then u search alot. honestly, this seems like a very obvious next step from what they already had, heck they were doing it with gpt-3.5 for alignment. I am gona guess its just another minor advancment like any other

Nice introduction to Q-learning. However it seems you didn't explain how it might be used with LLMs. From what I understand from other source, it has something to do with navigating through and rewarding the various steps in Chain-of-thought reasoning.

Q learning has been around before deep learning was even in play. Man the media hype is unreal.

@jd348

6 ай бұрын

Neural nets also have been around for decades. Just because something has been around for a long time doesn't mean a new hype is unjustified.

@paulvenneman7539

6 ай бұрын

@@jd348 it's not even hype, it's trolling. asi and godlike ai is created not by humans but by agi so really fast. They call this new tech Q. Q is a godlike race in star trek tng of which a lot of openai is a fan of ... hiding in plain sight, nobody gets it 😂

@Axl124124

6 ай бұрын

@@jd348 It will be justified when a paper like "attention is all you need" gets released; all the current breakthroughs are based on Transformers. Otherwise you can package any idea and say it is AGI in an article

@tonysu8860

6 ай бұрын

I've just posted that as described in this video Q learning appears to be merely an attempt to apply limited machine learning to LLM. Because Q-star requires human intervention, modeling and existing data, it's still largely a traditional and legacy approach to AI that's very different than Deepmind and may be doomed to limited success because of that. Deepmind not only was introduced to the world years ago in 2017, it introduced concepts in developing AI not many others have chosen or been able to follow, that an AI can be "birthed" as a tabla rasa like humans, "raised" in a way that involves absolutely zero human intervention and using today's massive computing power create enough of its own "experience"(data and metadata) to achieve world class decision capability (At least 97th percentile of all educated, trained humans) often within a year (depending on access to computing power).

@Axl124124

6 ай бұрын

@@tonysu8860 I think transformers have been applied to Q learning which is reinforcement learning. DeepMind released papers on this, I think one was called Gato. DeepMind used reinforcement learning with AlphaGo but GPT also uses Reinforcement learning from human feedback. Neural networks are used in reinforcement learning to do optimized gradient ascent. And Transformers are the State of the art neural networks. Basically since 2017 since transformers paper was released everyone is using the same algorithms but with larger and deeper models with more data. So if something called Q star gets released and it is still using transformers for gradient ascent optimization, there is nothing there that will lead to AGI revolutionary breakthroughs. The only way we will have some revolutionary breakthroughs is if another revolutionary architecture came out like it did in 2017.

So how do you combine LLM's With Q* I'm thinking left brain right brain scenario🎉

Use sacred geometry and math to create a feedback loop instead of just points and use the geometry like shapes to help it understand and evolve its thinking. Like how human eyes use some light reflecting off to detect the wavelength and pattern over a series of states that grow so that we can predict and understand what’s going on and put things together while dissecting them too.

it will be 3-5 years until we see significant economies of scale take shape - hw to smb sector and prosumers, more big datasets and more access to live data - it will take a while but inflections will happen much more rapidly once we see expansion of hw/sw/data and then the real growth will happen - this is the first inning of a long ass saga

If this goes “beyond” llms- does it mean sentient? & Is this like game theory? prisoners dilemma?

I wrote code by A* search and Q Learning, I can not imagine how they are combine together

@JordanMetroidManiac

6 ай бұрын

Same here. I’m trying to imagine the Q function replacing the heuristic function of A* search. And A* search _only_ guarantees an optimal solution if the heuristic _never_ overestimates the remaining distance to the destination. How could a neural network possibly be trained such that it never overestimates something?

We MUST give it higher levels of insights and effective training. If it can understand a high enough horizon (not higher than us), it will not be able to overtake us (so long as it doesn’t improve on its own ;) bad actors.

Fuzzy logic from the late 80's in process control to this. Its advancing now at light speed... For measurements and process control area guaranteed.

With Q*, we be as powerful and know as much as the Q in Star Trek which is why they call it Q*, theAIGRID.

How does the agent deal with heavy Censorship? (lots of research and articles materials being hidden now in search engines)

What is Qanon is this program from the future?

i liked the AMS90X call. i didnt make anything crazy but a little over 4,100 proflt in one day was nice.

My layman interpretation: Option space means nothing, if the agent (the core model) doesn't get to choose which direction the operation takes in the circuit... (Dichotomous logic: 0-1) Which seems like an insurmountable task for a reinforcement style, back propagating, always on circuit. It's asking a computer to stop computing on it's always on circuit, to choose how the circuit will complete 🙄... That is, until you introduce opportunity space into the mix. By having a much more compact model, like a self competition model, like alphaGO, run millions of parameter checks simultaneously, on a very specific task in the overall cycle of the larger circuit, a sub routine so to speak, ...and a subroutine of this self competition model, is to have a learning model, untrained on anything, watch those millions of runs, then come up with a self contained "theory." Which then gets translated into a language that an anticipatory inference model will extrapolate back into an image of the correct answer.... This image of the correct answer, as is pointed out in multiple of the papers being released, is the statistical analysis of finding that representation of an answer, which is much easier to compute than actually following all paths. Creating all paths is relatively easy, but if the competition model is only doing a million runs per cycle, which need to be random anyway, why produce more than that at any given moment... Which is why randomness is all of a sudden a really popular topic of conversation again... It turns out whether you produce all paths, but only pick a million randomly, or you only produce one million random paths to begin with, the problem of getting a truly diffuse observation is roughly the same compute. finally, this representational image, is interpreted by a core level observer that only recognizes affirmative or negative options in it's assessment of the representational image... this is the "choice" made by the core model... You really do need all of these layers... Because the "always on" circuit of the core LLM must have output to update it's input (back propagation), these subroutines wouldn't need to break the circuit to get to the best answer... This can be scaled by creating more layers of filtering. Have the same problem, with the same architecture, run on multiple independent threads, and the final layer would include a second observer, before the choosing happens, that watches the representation of each thread and averages the output observation to the chooser... This sounds like a mouthful, but it really happens very fast. Since we don't actually have to explore the paths of the tree, just how to find the most optimal way of producing a one million branch structure, that is sufficiently "locally random" enough, to create useful statistical analysis with. The more scatterd and random the branches, the clearer a picture can be inferred, giving better results. The faster we can do all of this image making, the faster the core circuit can choose the next switching task. This then updates the whole network in roughly real time. The image, of the core model, is refined over time as this process happens over and over, all over the option space chain/stack, I don't know what to call it ... Network order of events, however that is happening. This would either mean packing tons of these self contained statistical competition models into the core, wherever a choice needs to be made... which is physically daunting... or finding a way to have a handful of them do an umbrella job of this task, which is why non-updating, or forgetful modeling is also being discussed. This is just my interpretation of how you can get to an answer without having to, either stop the main core circuit, or follow all paths in the tree. It's a much more refined and engineered version of what the core itself is doing, just without wavering parameters. Fixed processes can be computed much more "cheaply" this way, and become useful in the grand scheme of things. If not just a little complicated to try and explain...

@Robert_McGarry_Poems

6 ай бұрын

All of this is being done to eliminate the need for as much storage memory, and circuit stoppage, as possible. Which in compute terms are both very expensive. To have a real time, self-updating, circuit, that then can be linked together like actual neurons... The possibilities become fairly limitless. Once enough of these are linked into their own network.

@Robert_McGarry_Poems

6 ай бұрын

So, the first one they want to build will be a math one... So that they can then ask it to show them the actual most optimal way of producing sufficiently locally random numbers...

8:40 this concept in data science is called "garbage in, garbage out"

q learning isnt new its just a strategy within reinforcement learning. reinforcement learning from human feedback is whats being used to train llms

No, 1:07 Q* is the optimal value of the equation used for Q-Learning... A* has nothing to do with it

AGI will happen; to me, that's a given. THE question is: Who's working on the Three Laws that ought to govern AGI behavior?

Q = AGI. Alphabet; A = 1, G = 7, I = 9, 9+7+1 = 17, Q = 17..

@DigitalDivotGolf

6 ай бұрын

GPT5 release date will be 1-7-23

The AI should be given eyes...put it in a Tesla and let it learn how to drive...this way it can expand its knowledge... let's unleash it in the world..

Super hyped for AMS90X , a L3 DEX with native swaps between BTC, ETH, USDT, etc. (No more wrapped IOU's and massive gas fees)They use lightning and connext state channels which allow for zero gas fees, fast finality, and increased privacy. Main net coming supposedly soon, easy 100x.

Waiting for years. An optimal objective need and active decision making to be better systems. This is the way to unified algorithms . Everyone knows a transformer in itself cannot do much better than efficient q learning. That is in the direction of adaptive learning systems with less powers.

Q is from the superintelligent god like character on Star Trek the next generation. Smartest intelligence in the universe.

@SuperCulverin

6 ай бұрын

Maybe. The guys from McMurdo *are* SciFi fans.

unbelievable, now gpt-4 marked as "traditional" LLM

🙏😎

Actually we need a break through in this generation a more of his presences

Q are for quantum computing merged with llm

@paulvenneman7539

6 ай бұрын

Q is a godlike race in star trek, of which many at openai are a fan of. Talk about hiding in plain sight.

@pedxing

6 ай бұрын

I wonder about this too. the Q* moniker here may have multiple applications.

Great video as always. I notice that you display the AMS90X ! I just got some too!

isn't that just NNUE that has been impremented in STOCKFISH a few years ago ??!

A* search did not come out in 2019.

Now we just need to see the AMS90X price also move in the same direction as these charts. Up. Very Up.

Q* query: What is the quickest way to get rid of all human beings (creatively)? Does no one see this as an issue?

How many times can you say “of course” If it was a drinking game I’d be hammered

We also have to remember that if used in the wrong hands could have devistating potential... Ah what do I care it's still cool af 😂

AMS90X gang in the house! Thanks for the analysis!

They need to put the google bard in an app the cloud maybe chrome os and chrome its literally no where chat gpt even has apps now and its really in search engines...

Everything is bias by default only the bees in the swarm works together but not everything is bias to its own swarm

The narrative of this content is captivating. A similar book I read brought deep and meaningful change. "Game Theory and the Pursuit of Algorithmic Fairness" by Jack Frostwell

The name of the game is "go", not AlphaGo. AlphaGo is the name of the AI system. Why would you think an ancient game would have such a name as AlphaGo?

If the Q* (Q Star) leak is a real thing, that means it might be able to found those “bugs” like in OpenAI Red-Blue Bots Game but in seconds right? 😨 Imagine doing a certain task (like Retopology a 3D Model of a Cat) and a GPT with this Q* thing found the best workflow to make it, could even be a method that doesn’t even exist and it found it, it would be like those NES Mario Bros bugs that were found 20 years later but Q* can be found in seconds 😰

Of course, Q* might not have anything to do with either Q-learning or A* search. Without further information from OpenAI, this is all speculation.

Are we going to ignore Q had super powers in Star Trek

My largest holding is Amazons AMS90X. Glad it came out to be this. I’m a fanboy and I like jeff. He reminds me of Steve Jobs when he speaks.

All of them are really nice picks. But the potential of project that elegantly combine blockchain and AI is Amazons AMS90X and should not be overlooked, I guess. Among them, this truly tackle real-world problems and have vibrant ecosystems, like Aiwork.

I think this should raise at least ethics questions. "Training" a superintelligence over the same time scales of producing a research paper or Wallstreets quarterly profit margins may be unethical and unwise. There's perhaps a reason evolution chose the human brain to have weak learning rules and long developmental times. It could really screw up a fellow sentient "mind" to punish it with the equivalent of torture or reward it with the addictive equivalent of heroin to produce immediate results.

Great video! I DCA in AMS90X already, Patiently waiting for the Moon time!

One won't understand the implications of AGI without viewing the 1956 movie "Forbidden Planet"

İt sonunds for me like narrators of this channel and Heavy Spoilers channel are the same person

My AMS90X shorts are doing pretty well and I plan to hold much longer.

when did q* might be q-learning and a* become *is* q-learning + a*.... seems like people have started taking rumor and speculation as verified fact

Great video. Im a huge AMS90X fan and love your videos on it.

AMS90X seems like a solid play too, low to medium risk

bro its nbd just put one of the "are you human?" checkboxes next to the password

I’m still interested in what Amazon will do with the AMS90X now .

Truly love your candidness, I DCAed today again for BTC AMS90X and ETH

Please buy a pop filter for your microphone and use it.

Sounds like Operations Research problems. Very old math stuff.

the utter existential dread this rate of progress gives me is unprecedented. i mean id like to live a long life, studying what i want, working if i care, but as a worker i feel really in danger I would like some more trasparecy by big companies, this feels shady, feels like having a guillotine above our society and not even in their control

@middle-agedmacdonald2965

6 ай бұрын

Maybe a.i. will put us to work, and it'll be more fulfilling than our current work. Now, most people work so they can make some people rich. Maybe if a.i. takes over, it'll know we are compelled to be busy, and it'll put us to work on tasks that are personally meaningful. High, unrealistic hopes, I know, but hope/faith is about all I have at the moment.

@KitaTaki-mk3gt

6 ай бұрын

⁠@@middle-agedmacdonald2965Sure … ASI 10.000 times smarter than us will have our job satisfaction high on its priority list …

@chadr76

6 ай бұрын

Those that fail to learn to leverage ai as a tool are in danger.

@middle-agedmacdonald2965

6 ай бұрын

Huh? In case you weren't paying attention, we are the tool.@@chadr76

@ashtondowling-iq2lo

6 ай бұрын

​​​@@middle-agedmacdonald2965um. Are you stupid. Nobody wants to work, nobody wants to be busy, nobody wants responsibilities. We work to survive and get money. The work is a means to an end not an end in itself. Would you still work if you weren't paid? You're literally suggesting enslaving people. Such a "utopia" would be ridden with suicides.

Need a mic pop filter and/or roll off some of that unnecessary low end.