I am kinda a big hater on AI and what danger it represents to the future of humanity
But. as a hobby programmer, I was surprised at how good these llms can answer very technical questions and provide conceptual insight and suggestions about how to glue different pieces of software together and which are the limitations of each one. I know that if AI knows about this stuff it must have been produced by a human. but considering the shitty state of the internet where copycat website are competing to outrank each other with garbage blocks of text that never answer what you are looking for. the honest blog post is instead burried at the 99 page in google search. I can’t see how old school search will win over.
Add to that I have found forums and platforms like stack overflow to be not always very helpful, I have many unanswered questions on stackoverflow piled-up over many years ago. things that llms can answer in details in just seconds without ever being annoyed at me or passing passive aggressive comments.
Hobby programmer her as well. I know you I’ve spent a lot of time searching for solutions or hints for, especially when it’s about edge cases. So using AI as an alt. to a search engine have saved me sooo much time!
Another thing with the approach. I read somewhere that it require about 10 times as much energy to ask an AI instead of doing a web search and spending a little time looking through the result. So it’s something I try to think of to motivate myself with, to do as many usual web searches as possible, saving AI queries for when it matters more.
I would say it’s more like 1000 times more energy. Trillions of matrix math computations for a handful of tokens at max speed and CPU/GPU usage, compared to a 10 millisecond database query (or in wiki’s case, probably mostly just easy direct edge node cache with no processing involved.)
I am kinda a big hater on AI and what danger it represents to the future of humanity
But. as a hobby programmer, I was surprised at how good these llms can answer very technical questions and provide conceptual insight and suggestions about how to glue different pieces of software together and which are the limitations of each one. I know that if AI knows about this stuff it must have been produced by a human. but considering the shitty state of the internet where copycat website are competing to outrank each other with garbage blocks of text that never answer what you are looking for. the honest blog post is instead burried at the 99 page in google search. I can’t see how old school search will win over.
Add to that I have found forums and platforms like stack overflow to be not always very helpful, I have many unanswered questions on stackoverflow piled-up over many years ago. things that llms can answer in details in just seconds without ever being annoyed at me or passing passive aggressive comments.
For now. Maybe.
It won’t be long before these LLMs will start ingesting the output from other LLMs, biases, confidently wrong answers, hallucinations and all.
Hobby programmer her as well. I know you I’ve spent a lot of time searching for solutions or hints for, especially when it’s about edge cases. So using AI as an alt. to a search engine have saved me sooo much time!
Another thing with the approach. I read somewhere that it require about 10 times as much energy to ask an AI instead of doing a web search and spending a little time looking through the result. So it’s something I try to think of to motivate myself with, to do as many usual web searches as possible, saving AI queries for when it matters more.
I would say it’s more like 1000 times more energy. Trillions of matrix math computations for a handful of tokens at max speed and CPU/GPU usage, compared to a 10 millisecond database query (or in wiki’s case, probably mostly just easy direct edge node cache with no processing involved.)
Alright, yea sounds fair enough, even better motivation to prioritize search engines!