Loopy Pro: Create music, your way.
What is Loopy Pro? — Loopy Pro is a powerful, flexible, and intuitive live looper, sampler, clip launcher and DAW for iPhone and iPad. At its core, it allows you to record and layer sounds in real-time to create complex musical arrangements. But it doesn’t stop there—Loopy Pro offers advanced tools to customize your workflow, build dynamic performance setups, and create a seamless connection between instruments, effects, and external gear.
Use it for live looping, sequencing, arranging, mixing, and much more. Whether you're a live performer, a producer, or just experimenting with sound, Loopy Pro helps you take control of your creative process.
Download on the App StoreLoopy Pro is your all-in-one musical toolkit. Try it for free today.
Comments
👊
/DMfan🇸🇪
Ha! Incredible and fantastic POV concerning AI…
Ha! Incredible and fantastic POV concerning AI…
Machines to help with laundry and dishes? Crazy talk.
Robots will be doing the dishes and laundry, A.I. will be making the art , writing and music. That should clear things up.
Yeah, this can be the quote of the 2100's century !!
Whenever AI is beating this incredible Norwegian man, I will stop being an tech guy and become an outlaw...
Logistics, logistics….
We have already solution for problem she describes. It's decentralized computing. Same way like Bitcoin mining can run all around globobe across thousands of computers and computing farms, same way we can (there are already existing project workiung on solid solutions for this) run machine learning and models run on decentralized network of computers - which solves problem of need concentrated high amount of energy sources at one physical place where datacenter is located.
Also AI running on decentralized network of bitcoin-type will be basically unstoppable, impossible to turn OFF.
Now, just imagine implications. AI running on tenths of thousands computers around the clobe, connected through global satellite internet network (Starlink).
No matter how much scared you are, you aren't scared enough
Unless there’s massive improvements in the efficiency of the algorithms, those energy costs will indeed spiral in the way detailed in the video. I’m not sure that can be dismissed so easily by saying that peer to peer networks are going to freely contribute that, it’s orders of magnitude greater than bitcoin et al and that’s already creaky and slow…
@dendy running modular systems in parallel as opposed to centralised systems doesn’t overcome the exponential rise in energy or resource requirements needed for either and as you correctly state involves giving over more autonomy to a system that could be impossible to understand let alone control.
It solves energy problem seamlessly .. There are already existing solutions for decentralized machine learning (projects like render.com) .. when you can distribute computing power across globe, you solved problem of huge energy source needed at one physical place .. it's exactly how Bitcoin runs .. You can't build 100 GW/h powerplant near 1 data center to feed it, Sabine is right at this point, but you can easily have 100 data centers around the globe each one needing just 1 GW/h ..
And now think one more step forward - you can have not 100 data centers but 100.000 smaller units even to the level of single computer or just few GPU rig running by somebody at home, connected to this vast global decentralized computing network.
Also important point - with machine learning, there are 2 different phases (when it comes to computer/energy needs)
1/ training of the model - this is the part extremely demanding for computing power / electricity - and this task is EASILY to be distributed across even tenths of thousands of nodes around entire globe... We have experience with such computing literally for decades (various decentralised computing projects like SETI at HOME, Rosetta at HOME, and so on..)
2/ running of trained model. This is vastly less energy demanding than training phase, and we are far away from reaching any local hardware limits (at least decade from that). This part is a tiny bit more complicated to run in decentralised network (for example because of latency issues), but with rise of ultra fast / ultra low latency global internet this stops to be problem in literally few years (looking at you, Starlink)..
Yes but this doesn't mean people will not go that way. They will do, i am absolutely sure. It's human nature, to try everything what is possible.
Also step back - fact that AI was trained (or even is running) on decentralized netwok of HW doesn't mean it has access to control things outside this network or even in real world - those are 2 different things. It just means it is hard to turn OFF that network (near to impossible if it is sufficiently enough decentralized).
Current average consuption of Bitcoin network is estimated between 97 and 170 TW/year .. this means 97000 - 170000 GW/year .. this means beween 11 GW/h - 19 GW/h
Based on this source:
https://www.theverge.com/24066646/ai-electricity-energy-watts-generative-consumption
"Training a large language model like GPT-3, for example, is estimated to use just under 1,300 megawatt hours (MWh) of electricity"
Which means to train GPT-3 needed energy like Bitcoin network from 4.9 days to 2.8 days
Thing is - you don't need to run learning phase nonstop continuously during entire year .. this phase is run obviously once for very long time. And as i said, then to run already trained model, your energy needs are vastly lower (comparable with any other computing - websites, cloud data servers, you name it). We are here talking literally just about spikes of energy consuption needs between relatively quiet phases with almost no consumption when compared to those spikes.
@dendy you don’t really address the eventual limit when all the data that can be garnered from the networks is reached, up to that limit could be technically and logistically solvable, but beyond that the logistics become ever problematic, unless you’re talking about removing ethical boundaries of privacy, even then there are still constraints.
Decentralization solves not just energy distribution but also data logistic - every node of network can cather own data at location where it is located, from local sources.
Also - this is absolutely hypothetical limit, and there is no rigorous way how to define what amount of data is needed to achieve AGI and eventual sAGI .. we just don't know .. also at some level of understanding of entire world, AI will be capable to extract emergent information from existing old data sets (information not explicitly added to set, but information which emerges from data relations - information which we even don't know they are there).
Our ability to understand what information can be extracted and used from given data set are orders of magniture lower that are physical limits of how much information can be stored in limited space.
So this part (lack of training data) doesn't bother me at all .. At current phase it is not problem, it will be not problem at least for next few years, and when it starts to be problem, i believe models will be sophisticated enough to overcome this problem and bring some solution to it.
Let's be realistic.. just look how so called "private data" are processed by big entities (not just private companies but also governments) now .. privacy is already lie, and it will be even bigger lie in future .. it's very sad and worrying, but it's reality..
To be absolutely honest - i am not too much optimistic when it comes to near future. I believe with very hight probability, 10-20 years from now humanity will not exist. There is just too much rising extinction-level threats (AI is just one of them) ..
Would be happy to go back to this thread in 2044 and post "thankfully, i was wrong" - but i don't think i will have opportunity to do this ...
This is the ‘big hope’ where uncertainty creeps in, the hope AI will solve the problem that limits itself.
What use for ethics, a very, very slippery slope, which I agree are already being flouted.
yup.. we already live in dark world, and much more darkness is on horizon … no doubt about that
Chilling interview … albeit i do not agree with few details which are imho based on wrong assumptions, in general, dude is right on the point ..
Energy, the realization of its importance for itself and securing the access to it (like primitive animals) could actually be the way normal AI gets to the point of AGI, kinda like animals evolving their intelligence through hunting and ultimately developing a consciousness.