@ChuckCook@elonmusk There is a reason to write a new programming language for LLM, closer to a bytecode level. Can be understood by humans but hard to change and pretty compact token wise for LLMs.
@wholemars@elonmusk I think he means something else. The source code is still going to be there, you simply will not need to look at it as you don't look into a binary code. However, I still do look into a binary compiled code to understand what can be micro optimized.
@TradexWhisperer@Micro2Macr0 Compute gets cheaper and faster over time. At some point of life almost everyone will run such models at home on their own box sitting beside their TV and showing targeted AI crafted ADs to pay for models people use at home.
@JustinCarowick@ClownWorld If you pour concrete you have partucular expectation on a curing time. Time to become usable or a time until it reaches a full strength. Asking how long it cures doesn't make any sense otherwise.