this post was submitted on 01 Jan 2026
960 points (98.9% liked)
People Twitter
10319 readers
666 users here now
People tweeting stuff. We allow tweets from anyone.
RULES:
- Mark NSFW content.
- No doxxing people.
- Must be a pic of the tweet or similar. No direct links to the tweet.
- No international politcs
- Be excellent to each other.
- Provide an archived link to the tweet (or similar) being shown if it's a major celebrity, figure or any politician. If you replace the "x" at their site with xcancel (so it beomes xcancel.com), and you're not on mobile, it will save video. https://web.archive.org/save is another great option. Archive.is is no longer recommended because they remove/alter content.
- If a tweet (or similar) makes a statement of fact, either mark it as "satire" in the body or provide the original source link backing up the statement
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
This is your daily reminder that LLMs don't "know" anything. They are a really complicated way of averaging out all the data on the internet. All this means is that the internet, on average, believes the word Trump is used in relation to pedophiles, and Grok was able to match an ensemble of pixels in an image to the word Trump.
For a minute there I thought grok had actually sat down and watched trump fuck kids.
I mean, Elon probably did steal a lot of confidential Government data during DOGE....
It is young enough to be desirable to them....
Probably effectively has, if it had a video from Epstien in its training data...
The term “AI” is marketing bullshit.
Excuse me, but everyone who thought Machine Learning isn't about to rearrange all of humanity simply didn't get the memo. /s
Everyone fucking knows that. Everyone. Can you please stop like your bringing some revelation. ?
Well, nope. You should talk to some of my colleagues, most of them IT engineers (and a couple of C-suite guys). That have swallowed the marketdroid speech about LLMs and there is no way to convince them those are just glorified linear regressors able to predict a stream of tokens without any resemblance or ‘knowledge’ about it.