The nVidia Cisco parallel I think is reasonable but AI is also the current version of the 90s/2000 Internet growth engine. So it will be huge.. the trough of despair may not be avoidable however, but as sure as digitalization, AI will take over.. AI is definitely winner take all except for the open source part of it which levels the playing field somewhat.
I believe we have reached 85-90% of what AI/LLMs will achieve with current technologies, and further progress with be asymptotic. I don’t believe more parameters were make them
better and companies will spend billions training and realize this.
I also think there will be a huge war in terms of AI-consumable content and many content creators will sue AI companies that use their content without licensing it. It’s going to become extremely expensive for new AIs to get trained in my opinion, and businesses still won’t be able to make money from them.
I dunno. I can hear music, it is muted in my brain but I basically have a audiographic memory. I can't visualize anything though. It is so weird. Basically my internal DNN can record and playback but not yet generate new things like stable diffusion either.. I suspect my musical training at a young age played a role in this as well as listening to lots of classical music.
Psychologists and neuroscientists can tease these things apart.. in the article they talk about it. But aphantasia is only recently being understood so methods to quantify the degree will improve.
I thought syntax was finite... wasn't that the Turing thing discussed here the other day.. also if your idea is true you'd be limited to the recursively enumerable.
State licensing weeds out the cranks. But these AIs could go for FDA approval and direct to consumer. So I don't think doctors are the limiting reagent here in your example. If these AIs are good, they will reduce demand for dermatologists as some enterprising doctor will offer it as a service that scales.
While I think doctors are paid handsomely and there are critical shortages, I don't think regulatory licensing requirements per se limit access to new and cheaper care.
Adopting a child is generally much harder than having your own. "Normal" kids rarely end up in foster or adoption situations. You see abused kids - you need a degree in psychology to help them recover from their background, or you see kids defects (genetic, mental, or otherwise) who need special care that is much harder than normal to give - while this could happen to you by random change, odds are it won't. Once in a while a "normal" kid has their family die (not just parents, extended family as well) and end up in the system, but this is rare. I'm glad for those parents who do take on foster kids, but it is not easy.
Adopting overseas is not a good answer - there is a lot of fraud (most countries will not even allow it anymore - it is too easy to kidnap a baby and by the time the rightful parents can get to authorities the baby is in a different country and impossible to trace down). There is still some opportunity, but make sure you are not unwittingly kidnapping a child.
That's certainly one of the issues - if having children was separate from raising them, like how pet breeders are generally not pet owners, children would be more of a fungible commodity and everything would be a lot simpler. I'm not saying it should be like that, just that economically it's one of the reasons the child business is not an efficient free market.
>There is also a cost benefit dynamic of increasing preventative screening, for example lowering the prescribed age for breast screening would save lives but may carry costs for nations or corporations that are unacceptable.
reply