DeepSeek slaps OpenAI, tech stocks crash
Chinese company DeepSeek announced its new R1 model on January 20. They released a paper on how R1 was trained on January 22. Over the weekend, the DeepSeek app became the number-one free download …
Excellent. Let's see if we can knock another 25% off everyone who is doing AI's stockprices and implode this bubble completely.
(As an aside, I'd also fully support lighting every leather jacket Huang has on fire and forcing him to stop making AI-bullshit GPUs for gamers and then charging $2000 for said AI bullshit.)
That was some good news yesterday, I love the chaos DeepSeek caused (I'd still throw it in the trash like any LLM shit), making investors question the moves of US LLM snake oil peddlers is a delightful outcome. I can't wait to see American CEOs make up some wild bullshit to placate the nervous investor. ROFL
Sknots 📉
if i were young again, i would consider ai to be about the one good thing to come out of these godforsaken last few years. i know this is the wrong sub for such sentiments, but it is a veeeeery powerful tool for self-education, even if that's not the most exciting aspect of its function for you, you can admit it is a useful one.
if i were young again
Why this?
can't approach it with required naivety as when young
veeeeery powerful tool for self-education
You know these are pattern recognition machines, right? And not an actually reliable source of information, right? These are hallucination machine which sometimes gets right simply by virtue of such truthful statements being repeated often enough in the data for the machine to pick it up. It can give contradicting answers depending on what keywords you use. As such it is of little education value.
If anything, it makes self-education harder, as search results will be filled with perfectly SEOed LLM article with no quality standards and complete disregard for the truthfulness of the text written, poisoning the well of knowledge the internet was supposed to be. And as a bit of irony, these poor quality pattern-based texts will later be used in future machine learning databases, thus lowering the quality of said data, causing the machine's own decay.
disregard for 'truthfulness' and 'not a reliable source of information' and what has kept our textbooks or teachers from being unreliable or untruthful?