The F1 student Visa fraud explained :
(twitter.com)
You're viewing a single comment thread. View all comments, or full comment thread.
Comments (39)
sorted by:
Enlighten us.
AI, or LLMs more specifically since that's the most visible variant, are basically just doing a whole bunch of math to predict the next symbol in a sequence of symbols based on probabilistic weights. So when you give it the input:
It takes a whole bunch of training data and sees that numerical symbols bunched together are often paired with operator symbols. This is determined by frequency of symbol patterns occurring in the training data. The vast majority of the time a set of numbers and operators ends with an equals sign(=) followed by another number. But sometimes it's a variable. The model itself has no actual knowledge of the symbols, they're just a pattern determined by frequency of occurrence and a big ol' hunk of math plus a pseudo-random seed used to simulate variance of response. So you can get a response that 2+2=4, or you could get a response of 2+2=5(because that's a symbol sequence that shows up a lot as an example of absurd claims during discourse on the internet which is what makes up most of the training data). However you can also get something like 2+2=x, because that's something that'll show up as a math exercise from some textbook that got fed into the training data. The seed is what will ultimately determine what kind of response comes out.
AI is basically a very advanced RNG-based spellcheck.
It has no knowledge and it cannot reason. It's all just weighted probabilities + some randomized variance.
For more evidence, look at IBM's Watson throwing the week-long curbstomp during its run on Jeopardy because... it mis-weighted a descriptor in the question.
Indeed, and when people start saying it's conscious that drives me nuts.
Conscious experience is not something computers can geneate at all. Heck we don't even know how the human mind really does it.
Who says humans do to that? You put a LLM and an Indian behind a monitor and try to tell them apart.
Honestly, you can tell because one of them answers immediately and the other puts you on hold
Your mistake is that you think most people are any different from an LLM. In the majority of the world, learning is just fact regurgitation. For a good explanation of that just look at the chapter discussing Brazil in "Surely You're Joking, Mr. Feynman!" It is pretty much only the White Anglosphere (and at least at one point, Germany) that focus on anything else.
No, most of the population are unthinking animals that simply respond to stimuli. I have no misconceptions about that, save for just how large a section of the population that actually is.
Oh I think that too. You saved me a lot of typing, lol.
"AI" is a copyright laundering system.
Search for something. Read the "AI Summary." Scroll down. You'll find an article that has almost entirely the exact same data, in the exact same writing style, hosted on someones site.
It's a bunch of bullshit smoke and mirrors so Google can steal your content. That's all it is.