AI companies want you to consume what they do not trust to train their models on. AI Companies Are Buying Tons of Old Books Because They're Free of AI Slop
China has less data centers than the US yet they're already ahead of the curve. But sure go ahead and call your state officials and tell them they can build a data center in your neighborhood's backyard. I'm sure your neighbors will be thrilled that you were the one to lead the charge.
Ahead of the curve in what? Do you not understand servers are still required to process every AI query you submit to your favorite LLM? Also, do you not understand this is just the beginning? You are making a claim data centers are being built in neighborhoods (i am sure you mean adjacent). Since YOU are making this claim, please provide me with a list of data centers built next to neighborhoods. There are appx between 4000-5000 data centers in the US. I am sure you can come up with a couple examples.
PR more than anything. An unreleased, powerful model under test escapes its "sandbox" (lol on what is a sandbox), hacks the open internet, and dances around to show off. OpenAI and Hugging Face partner to address security incident during model evaluation | OpenAI
Very likely ahead in cost and efficiency. BIG LARGE VERY SMART DUAL-TURBO DIESEL MOTORS TRUCK vs smaller, almost as smart and closing the gap, single electric motor suv.
I saw an article on this yesterday. I freely admit I do not understand all of the capabilities of AI, but this is some scry stuff. We will need you Sarah Conner.
There are two separate sides to this equation - Training and Inference. Training is extremely compute intensive. The foundation models (LLMS) like Opus/Mythos, GPT Sol, Gemini and Grok are a money pit. It is a race to the bottom (cost to revenue wise) and will always be offset to open source models like Deep Seek. Specialist models like Tesla FSD will be the money makers. There is not a lot of prominent specialist models currently because all of the computer and energy consumption is going into foundation models. Specialist models as an industry will require even more compute than the LLMs by a large factor. LLM's are incredibly inefficient compared to specialist models. Great to replace Google search engine, but you do not want them screening your medical records or driving your car. They are extremely lossy. Then you get into inference, your basic day to day use across the board. Currently we are in the infancy stage. Most of the inference usage today is highly inefficient because people are using LLM's for specialist tasks. I follow a couple people who are using agents like Hermes and OpenClaw and they have exposed how bad these LLM's really are, especially Anthropic. At this time, its largely unknown where compute requirements for inference is headed due to the lack of specialized models. FSD uses a computer the size of an ipad and just a mere 300W of power. However, I expect AI models to completely cannibalize the current SaaS market like Sales Force and even the OS market like Windows. Its not going to put the MAG 7 out of business, but it will force them to completely revamp hardware and software. AI is going to flow deep and wide. Todays market resembles more of Windows 3.1 than IOS. AI is not going to become sentient Anthropic and OpenAI are dead companies walking as they do not own their own datacenters
Imagine going back to the late 90's sitting with your drunk self at 2 am in the jack-in-the-box drive through scrounging for quarters to buy a few greasy tacos. Then tell him that in 25 years, he can be sitting on the couch, order the nasty taco's on his little computer the size of his palm and have it delivered to him without ever talking to a person.
The advanced models are a cybersecurity nightmare in the wrong hands. Humans' only hope is to use models to defend against models. The future is T-800 (Schwarzenegger) vs. T-1000 (melting and reforming). (OpenAI and others are not necessarily the right hands)
https://www.wxyz.com/news/voices/ne...-data-center-is-24-7-and-upending-their-lives Neighbors say noise from Michigan data center is 24/7 and upending their lives
