i think people... i mean people in tech, people in business, eh... people in general should move on from this AI slop phase.
let's just talk about ChatGPT for a minute.
openai needs more than 150k servers worldwide to answer its users' silly questions. now don't get me wrong, but the people i've been around have already ditched their calculators for chatgpt. and let's not even talk about someone who just broke up and is writing prompts so they can get a cool breakup poem to post on their ig story.
alright, that's not everyone. some people actually use it for necessary things, and i'm not gonna talk bad about them.
anyways, according to a few website traffic and seo sites, chatgpt.com gets somewhere between 150–190 million visits per day. it isn't publicly mentioned how many average requests a user makes, so if we even assume a single person puts in 10 prompts a day, the energy and amount of water needed to run the GPUs and keep the servers cool is massive.
we're not talking about air conditioners anymore. liquid cooling is the standard for AI servers. the GPUs here are obviously nvidia's latest H200.
these bad boys gulp millions of litres of water every day. not the flavoured water, not the salty water from the ocean either. just fresh water.
and that's just openai.
think about the other giants like google's gemini, anthropic's claude, meta's llama, or perplexity. the list goes on.
it's crazy how people use these things so conveniently without knowing the consequences or how this system actually works.
unnecessary use cases like grok's replies on X, gemini's integration in youtube, copilot on windows 11 (which no one uses, and i personally make sure it's uninstalled whenever i set up windows), and even google search turning into AI-first search.
i'm in that group too. shamefully.
and now, i've started making a habit of thinking for myself before depending on AI.
thinking is disappearing, and it's becoming rare.
people now just pick up their phones, find a solution instantly, or flood their brains with information they don't actually need. i bet most users don't even read complete answers.
thanks to tiktok, reels, and shorts, people have given up their ability to focus and actually read.
and now these short form video platforms have started their own AI slop-generating "features", which i'm actually glad many users hate.
but at the same time, it's scary how close these things are getting to looking real. in a few years, you genuinely might not be able to tell the difference between what's real and what's AI.
investors should stop funding startups that just slap a .ai tag after their name.
seriously, most of these startups are wrappers around existing services. so many investment firms have dumped billions of dollars into this clown show.
at this rate, it's looking like these AI companies might build a giant dam just to run their datacenters.
and if we forget about the energy problem, there's another major issue we're not really talking about enough: personal data.
would you dare show your chat history from the AI apps you use?
even if you do, it'll be second-hand embarrassment for me.
these companies might tell you "it's all end-to-end encrypted", "we don't train on your personal data", blah blah blah.
but who really cares? and more importantly, who really enforces that?
the EU has GDPR laws that at least make it harder for these companies to mess with user data.
meanwhile, india feels like a disneyland for these data companies. they prey on our user data while offering free subscriptions, student memberships, trials, telecom partnerships, and whatever else gets people hooked.
well, i got no robin hood kind of conclusion to end this story, so here's my opinion in bullet points:
-
nothing is free. if it is free, you are being exploited.
-
use local llm models that run on your device. they don't need internet, and your data stays with you.
-
these models can range anywhere from 1B to 120B+ parameters. (
B= billion parameters. bigger usually means smarter, but also more memory-intensive.) -
the latest gemma model is what i use for general queries through LM Studio. it uses around 5 to 8GB of RAM and handles simple queries just fine.
-
for complex stuff, i turn to GPT OSS 20B.
-
chatgpt is my last resort.
-
for coding and maintaining this site, i obviously use AI with my eyes closed because i suck at coding.
-
web dev is not my thing.
-
thankfully, i understand UI/UX concepts, frontend terminology, and some backend logic, but i can't translate that into code. That helps me explain what i actually want to github copilot.
-
track your AI usage. be self aware about your dependency.
-
shift to search engines like DuckDuckGo, Startpage, or Ecosia. or at the very least, turn off AI summaries in google search.
-
most importantly, think for yourself before you type a prompt.
-
if you find yourself struggling, do your own research first.
fundamentally, and ironically, artificial intelligence is actually dumb.
it has no self awareness. it cannot truly understand things. it doesn't have emotions.
sure, it can simulate all of that. but that doesn't make it intelligent.