Artificial Intelligence is nothing new, but what is new is the impact that it has had on the mainstream global economy. It appears that GPU, RAM, and SSD manufacturing is the latest "illicit substance" that has ultimately destroyed industries all in one technological hype based blow. There have been tons of layoffs and all sorts of crazy things in the past couple of months. As much as I would like to avoid this, politics has festered its way into the tech industry and now the global economy is suffering. Computer parts which were once regularly available and reasonably priced is now a thing of the past. Computer parts have sky rocketed due to overly hyped Artificial Intelligence Data Center demands. The GPU, RAM, and SSD manufacturers are now raking in billions and there appears no end in the apocalyptic outcome of a distopian technological future.
 |
| ChatGPT |
Where does this leave a guy like me? Ultimately, like everyone else in the tech industry, I became afraid and freaked out. After the mental hype settled, I realized that "AI" or "Artificial Intelligence" is the latest buzz word that has rocked the global economy, much like the initial "Dot Com Boom" of the 90's. I began to read and hear people shout about how AI is dangerous and must be stopped. Most importantly, the massive layoffs and all sorts of AI related news that has been very negative. It all begs the question, if everyone is getting laid off and all of these "AI bots" will take over our jobs, who will be paying for the AI services if no one is working? With all that said, I started to accept that the new technological evolution is upon us and I cannot afford to be afraid. I made the decision to dive into all things AI.
 |
ComfyUI |
Now that I have decided to look at the AI monsters in the eye to find out what they are made of, it became very clear of the limitations of the current state of AI (I will be referring to "artificial intelligence" as "AI" from here). What is irresponsibly being promoted as a "Jarvis-like AI" (AI from Iron Man), it is merely an excellent regurgitater that can sort through massive amount of content, much faster than any human researcher or intern. However, there is a huge drawback in the fact that it fails when finding alternative solutions that have not be established prior, specifically within the specified topic. For example, I wanted to correct some errors with ComfyUI and although it got rid of the errors, it rendered the ComyUI instance inoperable, specifically based on the working workflows used prior to the ChatGPT fix. It is only when I took matters into my own hands and reworked my approach with the AI. Once I had the AI get the application to a functional point, I stopped using the recommendation to fix the remaining issue and used my own deduction. There was a node that depended on a GIT repository that was no longer available which rendered the workflow useless. After conducting a quick web search, I discovered that the custom node is nothing more than a slightly modified rendition of the default Sharpen node. Once replaced, ComfyUI and prior workflows were in full production once more. This brought me to a new acceptance of the technology as I realized that AI is best when used as a tool not as an all knowing Oracle. Buzz words like agentic tools or AI agentic have been reverberating through social media and I finally have first hand experience with an agentic tool.
|
| ComfyUI Sharpen Node |
This was an extremely valuable lesson learned. All of my fears completely disappeared as I was able to fully understand the gravity of the situation. Sure, today, AI is a great regurgitater, but in the next 10 years we may have the Jarvis-like experience. Until then, all AI software/tools should be used as nothing more than an extension of my current tool set for IT. Overall, LLM's like ChatGPT or even local models are great for quickly acquiring information from large chunks of data. However, they all lack the ability to come up with original thought or an unconventional approach to fixing an issue. The hacker mentality of break until you make it work comes to mind. My final thought on this matters is that AI is nothing to be feared, but a tool that should be used responsibly. In my book, there is truly only one job that AI will take over, search engines.
From herein, I will be documenting my experiencing with learning "All things AI", which includes but not limited to Ollama, ANythign LLM, ComfyUI, ChatGPT, Selenium, Playwright, and other QA Test Automation technologies.
Comments
Post a Comment