a common theme in the history of prime intellect is that we’ll see two versions of how to potentially solve a problem, one being much easier, but where the harder version would be more powerful. we usually go for the hard version. ends up paying off in the long run
@willccbb hey, same
Users agree that harder AI solutions outperform easier alternatives long term because they lower costs for future problems and now seem correct despite earlier advice.
Based on 4 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@willccbb Would be cool if you are doing something similar for your RLMs, distributing inference for each sub LLM call in network
@willccbb I feel this heavily departs from what was recommended 5 years ago but is absolutely correct now
@willccbb the real return comes when the harder path lowers the cost of the next hard problem
@willccbb hey, same
a common theme in the history of prime intellect is that we’ll see two versions of how to potentially solve a problem, one being much easier, but where the harder version would be more powerful. we usually go for the hard version. ends up paying off in the long run
Users agree that harder AI solutions outperform easier alternatives long term because they lower costs for future problems and now seem correct despite earlier advice.
Based on 4 visible X reactions from 5 accounts; directional sample.
Ask a question below.
Published answers will appear here.
