LLMs will go down in history as worse than QAnon and Fentanyl

First, what LLMs are not:

LLMs are not CGI
LLMs are not Quantum Computing
LLMs are not Robotics
LLMs are not Machine Learning (as a whole)
LLMs are not Statistical Modelling

I need to be very clear on my position on this particular topic since I have spent a large portion of my life building machine learning algorithms from scratch, perceptron networks, neural networks, bayesian networks, N-Gram prediction models, accidentally creating LLMs (sans transformers), among other systems from scratch in C, C++, Python, Prolog, R, and other languages. I have also used libraries such as Weka and caret to do similar things. I have been a system adminstrator for econometrics projects using STATA/SAS clusters in datacenters. Statistical modelling is not what I am talking about.

If you are using third-party systems including but not limited to, OpenAI, Gemini, Nova, Grok, or even ollama, for LLM usage for most any usage, you are probably going to end up in a worse shape mentally than any other human on the planet. It is not going to help you, I have seen nothing but negative outcomes, especially in IT. Do not use it for extracting code, one misplaced brace or one extra letter to a find command can cause the malware you were investigating to execute.

It's all in the name of NVIDIA and Quantum stonks, and the argument is always: "but the stonk is go up and law is good". The same was true about the pharmaceutical companies when they sold excessive amounts of fentanyl and oxycontin to the public and doctors overprescribed them, those doctors are in prison, and the Sackler family got off scot-free. You can and will be sued for the equivalent of malpractice in IT if you continue to peddle this stuff, and I am not interested in your failing future bankruptcy of an organization and future incarceration.

That's not to say there might not be some mild and extremely niche usage for offline models using ollama for robotics for instance, but anything internet connected is also a massive system administration, security, and privacy hazard that is not sustainable and will bankrupt your organization. Imagine internet connectivity being disrupted for a drone for even 5 seconds and it falling out of the sky because of a prompt going wrong, or your secrets being leaked because of XSS flaws dumping it (something Microsoft themselves have demonstrated to me in person, and there are plenty of other ChatGPT leaks in the news).

I exclusively see negative value to LLMs in programming, debugging, and system administration, which seems to be the last frontier of advertisment and propaganda on this topic. If you sell this idea that it is a time saver, you are delusional or incompetant and it is "speeding up" your awful code. It's also not helping you learn, so go read the documentation. I have had to keep correcting people who keep skipping docs because they did the only one giant bold thing in the documentation it says not to do that the LLM skipped right over when summarizing some topic to them and providing erroneous code/advice.

The giant GPU datacenters we are building are not CPUs, and cannot do anything but train models. They do not even power cloud computing. They can be repurposed as a gaming streaming platform (tried and failed many times, plus people are souring on the concept of video games right now for cultural propaganda purposes), resoldered/repurposed/resold for normal desktop systems, or likely end up demolished and thrown in a landfill. GPUs are also nothing to do with quantum physics, so gaslighting an entire industry that GPUs are going to solve quantum computing is literal insanity and theft of taxpayer dollars to do it.

If you find this position offensive, move on. The rest of the world will move right past you as your organization files for bankruptcy because of your own actions, unless you are a large corporation pushing this stuff and then you'll likely get a massive bailout at everyone else's expense. In which case, carry on grifting Mr. Snake Oil salesman, the US government has your back.

Reading/Viewing Material:

Inside the relentless race for AI capacity
The quest for superintelligence is spurring a data centre boom — but critics question the cost, environmental impact and whether it is all needed
"It's almost like a mass hallucination where everyone is on the same wavelength that we need more data centers without actually questioning why"
https://ig.ft.com/ai-data-centres/ Stack Overflow data reveals the hidden productivity tax of 'almost right' AI code
https://venturebeat.com/ai/stack-overflow-data-reveals-the-hidden-productivity-tax-of-almost-right-ai-code/ AI agent promotes itself to sysadmin, trashes boot sequence
https://www.theregister.com/2024/10/02/ai_agent_trashes_pc/ Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task
https://arxiv.org/abs/2506.08872v1 Not So Fast: AI Coding Tools Can Actually Reduce Productivity
https://secondthoughts.ai/p/ai-coding-slowdown Is ChatGPT making us stupid? - Aaron French
https://theconversation.com/is-chatgpt-making-us-stupid-255370 Interview with Vibe Coder in 2025
https://www.youtube.com/watch?v=JeNS1ZNHQs8 The Art Of Poison-Pilling Music Files
https://www.youtube.com/watch?v=xMYm2d9bmEA This ‘College Protester’ Isn’t Real. It’s an AI-Powered Undercover Bot for Cops
https://www.wired.com/story/massive-blue-overwatch-ai-personas-police-suspects/ Pentagon ran secret anti-vax campaign to undermine China during pandemic
https://www.reuters.com/investigates/special-report/usa-covid-propaganda/ Russian propaganda campaign used AI to scale output without sacrificing credibility, study finds
https://www.psypost.org/russian-propaganda-campaign-used-ai-to-scale-output-without-sacrificing-credibility-study-finds/ Seeing Through a GLASSBRIDGE: Understanding the Digital Marketing Ecosystem Spreading Pro-PRC Influence Operations
https://cloud.google.com/blog/topics/threat-intelligence/glassbridge-pro-prc-influence-operations Is ChatGPT making us stupid? - Matt Asay
https://www.infoworld.com/article/3809945/is-chatgpt-making-us-stupid.html Former CIA Agent John Stockwell Talks about How the CIA Worked in Vietnam and Elsewhere
https://www.youtube.com/watch?v=NK1tfkESPVY FULL INTERVIEW with Yuri Bezmenov: The Four Stages of Ideological Subversion (1984)
https://www.youtube.com/watch?v=yErKTVdETpw