@elonmusk Heat shield: it may come at a weight cost, but is there any reason why dragon scales or shingles configurations wouldn’t work better for the expansion and movement?
@dickiebush Aside from scriptures, I’d choose How Will You Measure Your Life by Clayton Christensen. It’s a great book to put things like career and family in perspective.
@girdley I heated my basement with a miner for a while. Quickly, it became cheaper to heat my basement and buy the coins with cash. Though I too ended up losing on that investment, the learnings helped me feel confident in investing earlier than most which has been a good call so far.
At some point, we’ll reach a point where an AI model only gets better with re-generation using additional data generated from previous generations - where questions asked get better answers, deeper thought, further answers to provide enhancement to the model.
these models will not decay or will be able to adapt for synthetic data to not converge on incorrect responses? Once we reach a tipping point of X re-generations of the model, we could classify the model as intelligent?
Could we create a benchmark of intelligence of future generations of AI that evaluates how many generations of a model being trained on it’s own synthetic data (recursion) results in model collapse? Is it possible that once AI’s generate a larger percentage of original thought,
For AI models, and it seems to apply to most all model types, studies have shown model decay as a result of synthetic data and recursion. I’m sure it’s not original thinking, but can you direct me to research or papers postulating the following: