481–490 of 609 posts
For political stuff, I avoid wikipedia and just search engines in general and ask Grok/ChatGPT, specifying the specific biases I want it to filter out and know pieces of misinformation for it to ignore.
Earlier quoted context omitted.
If you wanted to know more about a new programming language named “Frob” or a plane crash that happened today, couldn’t you use an LLM like grok? Or any other LLM that’s continuously trained on trending news?
How do I know the LLM isn't lying to me? AIs lie all the time, it's impossible for me to trust them. I'd rather just go to the actual source and decide whether to trust it. Odds are pretty good that a programming language's homepage is not lying to me about the language; and I have my trust level for various news sites already calibrated. AIs are garbage-in garbage-out, and a whole boatload of garbage goes into them.
How do you know the media isn't lying to you ? It's happened many times before (think pre-war propaganda)
Someone at work yesterday asked me if I knew which bus lines would be active today due to the ongoing strike. Googled, got a result, shared back in under 10 seconds.
Out of curiosity I just checked with various LLMs through t3.chat, with all kinds of features, none had anything more than a vague "check with local news" to say. Last one I tried Gemini with Deep Research and what do you know, it actually found the information and it was correct!
It also took nearly 5 minutes..
Like I feel if your search is about _reality_ (what X product should I buy, is this restoraunt good, when is A event in B city, recipes etc.) then LLMs are severely lacking.
Too slow, almost always incomplete answers if not straight up incorrect, deep research tends to work if you have 20 minutes to spare both to get an initial answer and manually go and vet the sources/look for more information in them.
Until the false results rate drops, it can't be trusted.
But in fact I overwhelmingly use search over llm because it's an order of magnitude quicker (I also have google search's ai bobbins turned off by auto-using "web" instead of "all".)
I've used llm "for real" about 3 times in the last two months, twice to get a grounding in an area where I lacked any knowledge, so I could make better informed web searches, and once in a (failed) attempt to locate a piece of music where web search was unsuccessful.
* adult cat sleep time -> search engines. * my cat drops his toy into his water and brings it to me -> GPT
If I find that I really need to be more extensively descriptive, or to actually ask a question, I'll use an LLM, but that's pretty rare.
Here's an example of the latter that I did yesterday (this is Google Gemini):
"Hello! There is a poetic device where one line of a poem can be read both as the continuation of the previous line and as the first part of the following line. What is that called?"
"Hello! That poetic device is called an anastrophe or sometimes a hyperbaton, but these terms are more generally used for the inversion of the usual order of words or clauses in a sentence.
"While anastrophe or hyperbaton can create the effect you're describing, there isn't one single, widely recognized term specifically for a line that functions as both the end of one thought and the beginning of another in the way you've described.
"However, the effect you're describing is closely related to the idea of enjambment (also known as "run-on lines"). Enjambment is when a line of poetry ends without punctuation or a natural pause, and the thought continues onto the next line.
"..."
I proceeded to provide an example, saying "Points if you know what song this is ;-)", and Gemini confidently misidentified the song. (The correct song is "Black Cow" by Steely Dan; Gemini told me that the lyrics I'd quoted were from Bob Dylan's "Positively 4th Street," which doesn't contain any lyrics that even resemble what I'd quoted.)