Rubbishnet: The AI Future of the Internet
The above image is from a YouTube thumbnail posted on January 12, 2025. As best I can tell, the topic is Franklin Delano Roosevelt, the 32nd President of the United States, who was the only President who served over two terms, winning four (not just three). Do you see any problems? Oh, where to begin! Here's a partial list:
The top image — the one without a mustache — is the only one that vaguely resembles President Franklin Roosevelt. The other three smaller images have bushy mustaches based on the 26th President of the United States, Theodore Roosevelt. Even the mustache-free top image adds Teddy Roosevelt glasses to Franklin Roosevelt, who did not wear glasses in public.
The text additions are even more astonishing. Not a single one makes sense, and not even the letters of the alphabet are fully immune to corruption:
Finally, the image contains innumerable random snippets of utterly meaningless graphics.
领英推荐
The good news is that it only took a few trillion dollars of infrastructure investment to bring us this "super-intelligent" — as AI investors and promoters keep claiming — next step in what once was an exceedingly useful concept called the Internet. The original Internet enabled humans to create new, interesting products and insights by working smartly with each other. Its replacement is an oh-so-friendly, energy-guzzling, chop-chop word salad generator that converts all solid facts into sloppy word prediction pairs. Increasingly, people use this new system not to connect with others but to complete their every thought.
"But Terry, this is only one worst-case example from someone who had no idea how to use generative AI properly!"
Really? All of the coding-assistance tools promoted lately by various companies, sometimes very aggressively inserting AI whether you want it or not, use the same wonderful we-don't-need-no-stinkin'-facts-anymore probability-pair tech that generated the above image.
Now ask yourself: Do you seriously think that the errors are not there just because these other variants of no-facts-just-probability-pairs databases bury their nonsensical results in cryptic programming code? Allowing randomizing tools unmonitored access to critical software infrastructure is the most dangerous form of sweeping problems under the rug imaginable.
So, welcome to the Rubbishnet, the multi-trillion-dollar replacement for a tool that once helped humanity work better and smarter. You will, at least, get more bright colors.
Video Editor at AVID Technology Development | CG Artist
1 个月Yeah that's it. The best title for current internet!
NASA SWFO SAN Operations Automation Expert at NOAA: National Oceanic & Atmospheric Administration
1 个月This is serious... You're scaring me, dude... And you're from the "Show Me" state...
Leadership and Keynote Speaker and member of the Data Science Research Centre at University of Derby
1 个月Given that a lot of us understand how the token sequence process works in both transformer and diffusion models, I do wonder why so many people are surprised by this sort of "synthetic data" or AI Slop.
Let's not throw the baby out with the bath water. https://www.uts.edu.au/news/tech-design/portable-non-invasive-mind-reading-ai-turns-thoughts-text
Is there anything right?