@aakashgupta Transformers donโt store the dataset verbatim they learn conditional distribution. low ratio's the signal of strong generalization not 95% overhead. You canโt just divide 1.8T params and claim 1800ร efficiency.
@heygurisingh In Appendix B they admit they only fully simulated the small surgery gadgets. For the large ones actually needed in Shorโs algorithm they just say we expect similar behaviourโฆ leave detailed simulations to future work. The entire 10k number rests on an unverified assumption.
@sephsephsephs when a stock that's been compressed for years breaks the monthly upper Bollinger Band with strong volume and fundamental catalysts, itโs not overbought itโs the beginning of a regime change. The Bollinger Band will start to expand (because volatility is increasing)
@ACInvestorBlog People seem to be forgetting that April 30th is Rank Day for Russell 2000 and BB seems to be very well aligned to be included.
22nd April we get the preliminary lists.
After that at the end of June, index funds will have to buy a big share of the float.
@NaveYoung People seem to be forgetting that April 30th is Rank Day for Russell 2000 and BB seems to be very well aligned to be included.
22nd April we get the preliminary lists.
After that at the end of June, index funds will have to buy a big share of the float.
@venturetwins Not enterprise right now but there is a case to be made about niche or small market apps where the end user can now just satisfy his needs.
@CryptoLawson@verasitytech The lost is already accounted. I'm with you on the gamble. Most important stay safe! Lot's of people won't be able to cope with this rug... Talk about it