Hmmm, it’s starting to look like the best music across all lands are essential for creating an efficient AI which can generate music, either that or they will have to use strictly non copyrighted music or employ an orchestra to create 1000s of hours of music examples.
Slight but important correction.
Because as I’ve noted if some day we create an AI that thinks like a human mind or at least has the skill (perception, affinity, and so on for sound / music) then it could create music from scratch once explained to it what music is and what makes a music good music.
Even then there is the key difference in what a human is and an AI on its own is. Human perception is part of our process of experiencing and creating music which is part of our innate ability. We like music because it incites emotions, feelings, impulses in our body, a music can be unpleasant but still good if it expresses a harsh scene for example, yet the same might be a bad experience when it is not part of music.
So a human already has at least some level of understanding even if not conscious, mostly in the form of the body’s reaction or way of experiencing the audio both in the ear, with the internal organs and so on that a newly designed machine will not have or not necessarily.
A microphone of course is a thing so it is not impossible for a robot / AI machine to experience audio but humans don’t just experience audio it creates certain impressions in the body, in the ear / listening to the sound and also these and the ways the music is also create certain impressions in the mind (how we perceive it mentally, interpret it and what thoughts and memories it trigger if any as the same music / painting / writing can be interpreted / feel different / incite different emotions and memories).
So a new AI / robot might have its own way of perceiving audio but it will certainly be a completely different experience for it than what it is for humans. It is like how it would differ between an alien species and humans, they might perceive and interpret a human music entirely differently than us due to the differences in physiology / body, psychology / mind and so on.
So probably even an AI android / robot / audio capable machine would still need it explained how to create music properly suited for humans unless its creation process results in such anyway or if the foreign / alien / different music it creates is still something humans would enjoy or prefer experiencing without it being tailor made for them.
What I mean is that it likely is an unreasonable expectation from a technology to be instantly suitable to create human suitable music without basing that capability and skill on something. Even if not a generative AI and AI would still need to learn what music is, how it affects humans and so on to can make proper music.
At least that is my thought experiment.
So in the end it is not about human superiority, instead it is about the born with advantage of humans when it comes to this due to the biological conditions of humans compared to the newly created AI which has not such innate abilities.
In some sense AI is deaf in this regard until somehow (through learning, coding or teaching) granted the ability to actually hear.
For contrast AI can easily outplay humans in chess or perform trillions of times the mathematical calculations than a human. The two have inherently different abilities and music is too prone to interpretation.
I don’t think it is possible to create an AI that thinks like a human, the sheer amount of power needed to have an AI constantly “learn” is more than we will ever produce.
Also, a computer or program will never actually “think”, anything a computer processes are simply 0 and 1 in different configurations. So assuming an AI has learned what a snare drum and kick drum is, it will have info on instruments.
I will feed it an english statement “Create a rap beat using just a snare and a kickdrum”, When I press enter the AI will simply analyse that statement and the AI will have lots of information on the words “create a rap beat” alone, it will signal a computer to prepare a .WAV file for example and make sure it’s open for editing, Then it will go ahead and find examples of “rap beat” and the words associated with that which will lead to audio files stored in memory which will have the 0s and 1s configured in a certain way. For my specific case the computer will need to identify the snare and kick, I think you get where I am going.
In my opinion this isn’t AI, to me this is a computer program which fully exploits parallel processing, an artificial intelligence would be curious and able to learn beyond it’s initial learning which isn’t possible at the moment due to resources and scaling issues.
Not sure how you missed it but this was what I was talking about. What you described is a generative AI.
This is exactly why I was talking about an AI that actually thinks instead of being an LLM for example.
And if you narrow your imagination to today’s technology then yes current hardware might be unable to support it unless some new model with methods currently unknown can do exactly that on current hardware, which be honest could be possible.
New technologies are exactly that and might be better supporting such. Maybe an actual neural network hardware will be developed that can formulate new connections in itself like a real human brain and learn just like a real human, who knows. Plenty of possibilities for future science.
Such systems, even LLMs are not operating at the 01 level, on a technical level yes but the important part is how they operate which is very different from traditional computing. While not outright thinking they are much closer to human thinking than traditional programs. They create virtual neural networks that somewhat imitate human thinking.
The next logical step will be to advance the hardware to the same specs then we could really have efficient mimicry of human thinking / brain functions. After that just a proper AI model is needed that goes beyond a simple LLM using the new hardware.
So if you do believe it is impossible at all to create an AI that thinks like a human that is a very narrow minded approach. Not saying this to offend instead to inspire you to be more open minded and read about related concepts both fictional (sci-fi) and speculative (scientific theories) because this is not a new idea, we are just limited in our current tech (and software) which is already much more advanced than even just a decade or two ago.
It’s not really an if possible, rather when it will be possible.
I would say it is more of a realist approach based on fact and physics, the people leading AI have made poor decisions and have poor judgement leading to billions already being wasted. The impatience level for one is astounding, many have invested a lot of money on copper connections between the GPU and memory where if they had sat down and done the math they would have understood it’s impossible to scale that and perhaps waited for the light based GPU’s where it’s light and not copper transferring data from GPU to memory,
I’m unsure where the next set of $300 Billion is going to come from to build thousands of rack servers needed to create what you speak of let alone the 1 to 6 GW’s of power needed.
Of course I would love to see advances in AI but not at the cost of a humanitarian disaster, If the water isn’t available to cool the data center then there aren’t many ways of getting around it aside from building the data center underwater.
We can go ahead and talk TDP/power consumption for the specific GPU chips available now and in the future and understande the sheer scale of what’s needed to power what you’re suggesting, you’ll see it’s not scalable.
Look at it like this…a human brain uses 20watts of energy, we are immediately millions of times more efficient than any one GPU now and in the far future. To try and immitate such an efficient use of energy simply isn’t possible.
I asked google AI: “compare the power usage of a human brain to a GPU and highlight as much as you can”
The human brain is a masterpiece of biological energy efficiency, operating on roughly 12 to 30 Watts of power (similar to a dim light bulb or a smartphone charger). In stark contrast, modern high-performance GPUs, such as the NVIDIA H100, consume roughly 700 Watts each, with data centers requiring thousands of them, consuming megawatts or gigawatts of power.
The brain performs complex, real-time, adaptive, and creative tasks on this “shoe-string” budget, while AI requires immense energy to simulate only a fraction of that intelligence.
Power Usage & Efficiency Comparison
Human Brain: Operates on roughly 12–20 Watts.
High-End GPU (H100): Consumes ~700 Watts per chip.
AI Supercomputer (Frontier): Uses >20 million Watts (20+ Megawatts) to perform around 1.7 exaflops, while a biological brain manages equivalent complexity with far less energy.
The “Simulated Brain” Penalty: A full digital simulation of a human brain using current technology would require approximately 2.7 billion Watts (the output of a large nuclear power plant) and would still run slower than real-time.
Key Highlights: Brain vs. GPU
Efficiency Leader: The brain is roughly 9 million to 2.7 trillion times more efficient than AI, depending on the task.
Operational Mechanism: The brain uses analog processing and low-precision, noisy signals. GPUs use digital 16-bit or 32-bit floating-point operations.
Capacity & Memory: The brain is estimated to have a memory storage capacity 1,000,000 to 2,500,000 times higher than a 1 terabyte GPU/PC.
Adaptability & Learning: The brain learns, adapts, and functions in uncertain environments with little data. AI requires massive datasets and high energy to train.
Data Transport: In GPUs, memory and processing are separate (von Neumann bottleneck), requiring high energy to move data. In the brain, memory and computation happen in the same place (synapses).
“Mind Wandering”: The brain’s default mode network allows it to process information while doing nothing. A GPU must be powered on and active to perform calculations.
Why the Brain Wins
Massive Parallelism: The brain operates 86-100 billion neurons and 100 trillion+ synapses simultaneously.
Sparsity: Neurons in the brain are not always “on” or firing, saving energy. GPUs often keep all components powered.
Low-Voltage Signaling: Brain signals operate at much lower voltages than silicon-based components.
Energy Costs
Training AI: Training a single large language model (like GPT-3) consumed about 1.3 gigawatt-hours—enough to power 130 U.S. homes for a year.
Inference (Using AI): Answering one ChatGPT prompt takes roughly 0.34 watt-hours, similar to an oven running for one second. Scaled to billions of prompts, this is immense.
In my opinion much of the AI community need to read and understand these facts here, Yes there is definitely a level of Artificial intelligence involved in AI computing, but you’ll find much of it is event driven and we are simply using text words or a text description to activate existing compute commands, the very term AI is somewhat misleading and should be called Parallel Processing or Ultra Processing.
Not going to read your entire post because it is clearly AI generated. I am here to converse with a human not AI, I can do that myself.
However to address the first part, I never said nor implied, if you pay attention actually said the exact opposite that with current technology we can create a brain similar AI architecture unless something (software itself) enables it through some breakthrough.
As I wrote, in the future, with future technology it can be possible.
Hmmmm, well by not reading that you won’t understand why what you’re suggesting is not possible. All google AI did was compare info and present it to me. Of course it is AI generated, I specifically asked the AI to give me all info about a human brain vs a GPU.
based on fact I don’t see how, please feel free to explain how they will match the efficiency of a human brain.
You’re enthusiastic about AI, but you refuse to read info generated by an AI where the subject is just facts about a human brain vs AI. This is a big part of the problem with the AI community, many of you think you can completely ignore facts, physical limits and basic human need.
I am looking forward to when Trump and his admin are removed from power and then we will get some sensible laws put in place to ensure AI is developed responsibly without poisioning or dehydrating anyone. Whoever signed off on building impossible data centers that need 6GW of power is an extremely ignorant person(s) and I seriously lack the understanding of how they got into a position to make such important decisions that affect other peoples lives negatively.
Personally I think AI is amazing and I actually need to subscribe to an AI service so that I can create promotion pictures quickly for the products I sell on Ebay.
I simply do not care anymore, because I’m the end of my bloodline. No child of mine is going to suffer the future that big tech has in mind anyway. I decided on this around the time social media became a thing, and it seems I made the right call. SO GLAD I don’t have iPad kids.
I don’t think it’s the tech that is the issue, it’s what people do with the tech that is the problem, war, surveillance, privacy, copyright, humanatarian issues are now important aspects of tech.
So try not to take what you see happening with tech too seriously, try to keep things realistic and be sure to ask questions like you see me doing. Do you notice how nobody has come forward to explain how a computer/Array of GPU’s can possibly match the efficiency of a human brain?
When serious questions need to be answered there never seems to be anyone available to answer, You and I can take this silence as an admission that some things are possible and some things simply arent. All we can do is make sure that we vote for sensible people who will prioritise human welfare above anything and make sure laws are in place to promote a sensible approach.
I applaud your optimism. I’ve seen human welfare deteriorate to far to be brought back into the fold. By now I regard Western democratic systems as “rigged to guarantee certain outcomes anyway” and I gave up chasing shifting goalposts about a decade ago. AI is just one more drop in the bucket of modern suck.
AI is basically the same arms race as trying to get to the moon first. There are too many examples in movies for theoreticals as to why AI is insanely important.
A good reference for me to continually support the push towards more complex AI is Warhammer 40k unironically. The ‘golden age of technology’ delt in AI taking current tech and human knowledge and pushing it forward to revolutionize the world around us. I unapologetically support the continued push towards better futures.
And crypto, even as expensive as it is now, is still make believe funds that you simply cant touch nor hold onto. It has no real affect on the world around us. It only matters in the internet. You can shut your computer and never see crypto at all. What ever this ‘mining for crypto’ for is all make believe and space dust anyways.
AI however has the capable and is actively pushing humanity into more complex systems, more effective medicines, and better futures for all.
We have survived the last couple of thousand years without AI, I’d say it’s important for tasks that result in something like cure for illness or a new invention that will help people.
To me it looks like the world is going backwards in some ways, Elon musk is using emergency power generators that are very polluting to power his chat bot Grok. Normally I would associate the word revolutionalize with a positive occurance.
If the people with unrealistic goals get their way then the future will not be better it will be worse. we would be looking at job loss, and a possible lack of energy and water as big companies start to pander to the mass amounts of money AI can generate.
If we take a look at Black Rock we can see that lots of investors are starting to pull out and withdraw their money to the extent where Black Rock has had to literally stop withdrawals, Black rock have invested around $100 Billion in AI along with Microsoft and Nvidia, we have also seen Nvidia go public about limiting their investment in openAI.
I guess we can pretend that everything is ok and the huge losses in AI don’t matter, the rate of progression is not sustainable. AI is a useful tool where I myself might be looking to get a subscription to help with my ebay.
I have also read that a typical graphics query to an AI server will use the same amount of energy as an oven during the processing of that query, now if we have 20 or 30 million people all needing to do simuilar queries we might end up with an electrical supply problem.
The reality could be that a typical business removing their human staff in favour of AI might eventually face higher costs due to the cost of energy needed for processing.
This is where the efficiency of the human brain comes in which is millions of times more efficient than any current array of GPU’s. Yes, the human may take longer to process, but the energy cost of a human doing a task will never be as expensive as an AI datacenter doing it.
I can see that it will take a long time for many to reach this conclusion simply because they would actually like to ignore it and carry on but please let me assure you… the issue of cost will rear it’s ugly head at some point and it simply will not go away if ignored.
Right now I am able to process my Ebay pictures for free using Chat GPT, i tell the AI what i need for my picture and it takes about 4 to 5 minutes to process. I can do this about 7 times and then I have to wait 48 hours before i can do it again another 7 times, this is free.
They have given me a demonstration of what it can do so why haven’t they locked me down into a subscription? Who excatly is paying for the processing when I create my pictures for Ebay? From my perspective the processing would have cost around £2.00 which doesn’t seem much but if millions are doing the 48 hour wait it adds up to a lot of money.
If I was in charge I would have made it so that the user gets the processing at no cost but they would be forced to wait 90 seconds for the “start processing” button to enable and during this time an advert will be playing with no ability to skip it, If they switch screen or mute then the counter starts back at 90 seconds. I would still use the AI service if this was implemented it would seem that the owners don’t know how to generate money and they don’t really know their customers.
When it comes to power demand, i am simply apathetic. The tech is there now to be able to power the world with very little consequence and easy answers to management that comes in the form of nuclear power plants. the tech is there that the biproduct, waste, from nuclear powerplants are effectively less harmful than hundreds to thousands of solar panel fields that create ‘green deserts.’
As for companies backing out of the AI business are looking at AI wrong. The ultimate issue for both subjects that necessary power and necessary funding for Ai both are being looked at from a consumer business point of view. “How can i get the most money out of this?” AI effectively is an unregulated brain thats able to pull information far faster than anything else we can comprehend. Information is power and the lack of control over information/power makes consumerism difficult to control in the AI field.
Consumer based businesses is your trojan horse to blame. The stagnation of innovation. There was a 1956 Frigidaire Refrigerator ice box Commercial that goes in a positive direction for innovation and technical progress. There was the natoreous Nokea we all love referencing as immortal that should be the gold standard for all cellphones of today. What do we get today though? a cell phone that if accidentally bumped gets shattered into a million different pieces without a screen protector, phone case, laniard and what not. And it still gets damaged. we are physically incapable of working on our own cars now due to each sensor has their own serial numbers programmed into the on board computer that requires a mechanic to sync before that car can even be used again if replacing spark plugs (2026 vehicles). The materials used in all equipment are no longer built to last or easily repaired. The blame rides solely on the businesses obsession to keep us consumers instead of customers which stagnates progress.
AI and electrical power can be viewed the same way. Businesses are trying to view them through the lens of consumerism instead of continued progress towards better futures for all.
Nuclear power plants need enriched uranium where Russia is currently supplying around 44% of the worlds needs. The actual enrichment process is mainly done by Russia and the USA is leading an initiative to reduce the reliance on Russia. I’m not sure how long this will take to come into effect but it’s obviously not immediate.
Is it the AI they are backing out of or is it the poor management? I think perhaps your perspective is wrong as it’s consumerism that builds economies, in fact business and consumerism go hand in hand as you can’t have one without the other. You refer to it as the “AI business” but then seemed shocked that decision makers want to create products from it and sell it to a mass market.
The mobile phones of today are simply the result of studies into HCI (Human Computer Interaction) which is concerned with the design and use of computers and the focus on the interface between human and computer. The design of today’s mobile phone is efficient as children are able to understand how to use one quickly. The thin glass on our screens is what enables the computer to accurately register our touch, obviously the touch screen feature needs to successfully work 100% of the time as a mobile phone is part of an emergency toolset.
Honestly, I think they have done a reasonably good job with phone design as I’m forever dropping my phone and sthe screen hasn’t broken, so I’m sure the are doing the best they can.
Nothing is free I’m afraid, it would be great if things were free but that’s just not how it is. It is actually consumerism that will pay for whatever AI needs, this is in the form of things like subscriptions to ChatGPT, or purchases of Microsoft products would partly finance AI. I’m pretty sure if Putin was asked to increase the supply of enriched uranium for the purpose of “better futures for all” the answer would be no.
If the massive data centers are to be built then it has to be done the old school way which will mean earning our way towards it and financing things like enriching more uranium in the western part of the world, Consumerism is what will finance these efforts. no body can possibly just “give” money to the AI effort with a “better future for all” plan in place, many will want to see a solid business plan where it is realistic and targets have been met.
All of what I’m explaining here takes time to establish, these are not things we can just have set-up by next week, the AI effort needs to have a good long period where only maintainance and services are paid for, no data center builds, no new racks of expensive gpu’s, consumers can pay their subs so that profit can be realised.
ChatGPT has demonstrated that it can earn vast amounts of money from subscriptions, the problem is that it’s being spent too quickly on unrealistic goals where physics and basic human ethics have been ignored. (This is infact a very good reason why compainies will back out of the AI business, many have probably asked where the water is going to come from and not received a good answer)
If AI gets to a point where it will take 60 years to realise profit for investors then there is something very wrong with the plan, more investors would be needed to pay off the original investors and then we have something that looks like a ponzi scheme.
I have given an example of the poor management in AI as I am creating my Ebay product listings free of charge just by waiting a few hours in between using ChatGPT, as I mentioned earlier, they have failed to lock me into a sub or insist I help them get paid by viewing an advert. I’m sure there is lots of money they have left on the table, they are obviously too busy chasing investor billions.
1 Like