@TheStalwart@ShanuMathew93 For eg medical or legal, where privacy is a must or low resource edge compute systems it is very useful. Not straightforward. You might train an already specialist model further or distill and of course only if it can’t be done with RAG etc
@sobedominik Your accuracy will also depend on your harnes setup and workflows. Proper specification and planning, verification agent workers etc used depending on work type will allow better output from smaller models
@richardartoul Pretending that you are better at coding in any language than an AI which has learnt coding from any coder that ever lived is hubris. It needs verification and that can be done by another AI if set up properly. 2027 and the bitter lesson is coming.
@phl43@teortaxesTex Easier said than done. Banking, insurance and retail run on decades old systems/software. Industrial systems have exposed control hardware and platforms that are eminently hackable. In true capitalist style every was built with sticks and bubblegum.
@ATinyGreenCell Depends on the type of work you are doing. Are there verifiable and decomposable tasks that can be performed concurrently. This depends on the type of tasks and the scale. If you meet these criteria you may use beads to do the work.
Debugging a trading system in AG - Argon results passed to Opus(both M) for checking. Main issue categories - multiple issues per round. Simply can't be trusted - not sure if it's lazy or what the problem is. @OfficialLoganK@_philschmid@rodydavis
@sequencemyneuro Very interesting. I believe they are immunomodulatory. The mechanism may be as simple as that they modulate emotional regulation that can affect immunity or possibly additional direct effects. Great research direction! Good luck untwisting and blinding the effects.
@johnennis Tirzepatide . Powerfully changes your relationship with any desire. An almost disturbing psychological change. You don’t think about it and you don’t care when it is around.
@doodlestein Glad I’m not the only one speaking to my agents like that. Once called it ret@rded and since then it often calls itself that when apologising for messing up.
@8teAPi Shulgin’s methods and suggestions have long made this fairly simple for even an undergrad to attempt and many “new” recreational drugs are from these classes.
@BenBajarin Don’t really understand the skepticism as just from the most basic perspective companies that don’t implement AI will simply be uncompetitive in the near term.