Эксперт Джон Спенсер призвал Chattanooga строить компактные ИИ-центры вместо гигантских комплексов в бедных районах

I'm sure many are already at the keyboard, ready to respond. Please, hear me out. First, Chattanooga already has at least 3 data centers in various locations. The difference between a data center, and an AI version is how they are used, as well as the components and infrastructure inside it. A regular data center has many server racks loaded with mainly hard drives, no monitors required. AI data centers have some servers, far fewer hard drive racks, and many more racks of GPUs, or high end video cards, which when utilized correctly, can crunch numbers very quickly.  All those GPUs require a lot more power, and also produce a lot more heat than the average data center. Unfortunately, that heat has to either be cooled or dissipated, and that's where a lot of the issues come into play. Water is mainly used to help cool these servers, and the amount of water in total being used in these data centers, are less than the water used in golf courses across the country. Many people ask why many of these data centers are being built in places where it's hot and has relatively no access to a water source, and that's because the land is cheap, very cheap. Also, these AI data centers are enormous, but they don't have to be so large.  Chattanooga would benefit from a smaller AI data center, one that isn't as massive, requires much less electricity, and we already have access to a water filtration system. Chattanooga’s power is mostly (95%) provided by TVA itself, so electricity would be much cheaper. Also, as a technology startup hub, it could be utilized locally, and the cost of tokens would be much cheaper as the cost of running it would be lower. The benefit there would be competing against China, as their “open weight” models are much, much cheaper.Yes, I understand the data is going to the Chinese government, but the open weight models can be run locally, without an internet connection and the only cost is your own personal electricity bill and a very capable computer or server. Open weight models can also be used on a local, city-owned AI server. I understand the issues with AI. My job could very well be replaced by AI in about 5-10 years, or at least be made so that my pay would be greatly affected. But, AI is here, and the best way to handle it, is to work WITH it, not AGAINST it. I know that there are jobs that will be lost due to this technology, and I think the only fair and ethical way to approach this is to give opportunities and training to those that lost their jobs first. My reasoning is that from my research, the employees that have lost their jobs due to advances in technology, are rarely the same ones that get the jobs that have replaced them. The AI genie has been released from the bottle, and as with the genie that granted Midas’ wish, we need to be careful with what we ask it to do. Midas was only thinking in the short term, thinking that certainly the genie wouldn't take him literally. Yet he wished that everything he touched turned to gold, and then his daughter and his food also turned to gold. The genie knew full well what the implications would be, but Midas had to find out the hard way. Had Midas been specific, he could’ve had all the gold he wanted, a full stomach, and his daughter wouldn’t be solid gold. In closing, I feel that the path that the large companies such as Anthropic, OpenAI, and Google are building far too many large data centers, and in areas that are quite unpopular. Mainly in populated areas in low-income housing, while they should be built in non-populated areas or in areas that can handle the power consumption. I feel these large companies are building as quickly as possible, faster than the courts can respond, and faster than the other companies. I fear that when the bill comes due, it will be the taxpayer that will be paying the bill, just like we did in 2008 with the banks. Going local, with smaller data centers, and in areas that would be better suited, would work better in the long run. John Spencer