yeah no, i should probably expand more on this as its not just the vast energy consumption so much as it is the environmental problems it poses, the article(https://www.techtarget.com/searchenterp ... l-problems) mentions how the training of MegatronLM(the deep learning language model) uses virtually the same amount of energy that three homes do in a year, that's not to mention the data centers which require water for evaporative cooling, this video by fellow YouTuber the Hated One delves pretty deep into the problem at hand, demonstrating the fact of the aforementioned data centers running out of water and resorting to extracting from declining reservoirs of water.Energy cost is not a significant barrier to entry for AI models when the renewable energy global capacity additions are expected to rise this year to 320 gigawatts – equivalent to an amount that would come close to meeting the entire electricity demand of Germany or matching the European Union’s total electricity generation from natural gas. Solar PV is on course to account for 60% of global renewable power growth in 2022, followed by wind and hydropower.
lol, this whole statement is tantamount to you saying "God will save us." if you don't even know the capabilities(more precisely the upper limits) of superintelligent AI then why are you placing your hopes, dreams and fears on this clarketech(tech indistinguishable from magic)?Moreover we have no idea what a true superintelligence is actually capable of. Imagine trying to describe the construction, manufacture and usage of an WIFI router to someone in 1850. If you had the required information it would be entirely possible to have built such a device hundreds of years ago, the reality though is that even if you could there is no way you would have been able to use it or even know what it's usage could be. That is the stage we are at when it comes to understanding a theoretical general superintelligence that is orders of magnitude beyond what a mere human mind can understand or learn or comprehend, we have a tendency to anthropomorphize - to apply human limitations logic and functioning to a non-evolutionary-based intelligence because as it stands we are the only reference for any true intelligent life. Doing this though is a fundamentally flawed way of calculating or understanding how an artificial superintelligence will behave what goals it may set or what it may decide to deem as relevant or irrelevant information would be similar to a squirrel trying to understand the value of a human stock trader's portfolio - we do not have the required level of data or intelligence to be able to understand how something that is not human and is incredibly more capable may act or what it may do.
might as well just find god at this point, because that's exactly what you're making it out to be. we barely understand intelligence as is(ask any neuroscientist, we're still trying to figure out the complexities of the human brain), somehow an AI that matches a human in intelligence immediately constitutes for self-iterative improvement and will lead to superhuman intelligence that comes with an ability to solve climate change and eventually ascend to godhood. lol get real man.
