#Guys I don't think ai will replace us anytime soon

[itsme-thekeys]

What's really behind these pictures 😱😱

Or: I used an ai expander picture thingy and it was doing so well and then it lost the plot

It did so well with Wille and then it gave up

This one lost the plot way earlier:

The plot never existed in this one:

I did another one but it turned out very inappropriate.

The original pictures:

[hippo-pot]

Btw, re: my opinion that computers are not gonna be able to translate sign languages in our lifetime, it's not that sign languages are necessarily More complicated than spoken/written languages (I truly don't know how you'd measure that but I'd assume they're equally complicated). But video is, in terms of sheer data, much bigger and presumably harder to process than audio. I cannot imagine this happening without *astounding* computational resources which would take far more energy, water, and money than a human interpreter (and, more importantly, wouldn't work as well, at least for the foreseeable future). I assume the computation would happen off site in most cases if it did work, meaning the Internet connection is gonna need to be phenomenal (there is already widespread dissatisfaction with VRS human interpreters used in medical settings because half the time the connection drops). Speech to text, with all the issues it still has, seems like a breeze in comparison to 'understanding' a video.

I also cannot wrap my mind around how a machine would handle depictions. Like, with some practice behind me, my human mind is now able to understand (some) depictions I've never seen before (thank goodness, because there will ALWAYS be new depictions I haven't seen before, bc Deaf people are resourceful and creative), but I don't see how a machine would. That's pure sci fi to me. I also wouldn't expect a machine to do a good job translating stuff it's never heard before in a spoken language (e.g. wordplay, or the way you can sometimes tell the meaning of a new slang word from context, or an uncommon name even), but the thing is I think depiction is a much bigger part of daily life than wordplay is?

[mothdoesnothing]

the ai art bots plaguing Tumblr rn are making me feel better about my own art because what the hell even are those images ????? guys I don't think ai is replacing us anytime soon we're gonna be just fine

[abstractfairy]

So the Efficient Market Hypothesis (EMH) is wrong / incomplete. Mandelbrot's book The (Mis)Behavior of Markets outlines how on a mathematical / statistical level the EMH is wrong. (Also, Mandelbrot is the same guy who invented fractal mathematics and has a lot of work in various fields)

Dr. Andrew Lo's Adaptive Markets Hypothesis (AMH) is a much better theory imo.

EMH relies on a several incorrect assumptions / AMH fixes them

assumes that people are perfectly rational investors. they aren't. we have fear (panic selling) and greed/mania (bubbles.) there's lots of research showing when humans act irrational (and even institutional investors aren't immune to this.) We can act rational in certain conditions, and we can get smth like the EMH if conditions are stable enough for a long enough time. The first half of his book sets up how we can be irrational and its effects on daily life and the markets (it also covers evolution and ecology, which we'll use in the next half of the book)

AMH takes an ecological view of the market. Not enough people pay attention to where the money is coming from or why. Who is selling? Why are they selling? do they need cash, or do they think its a bad investment, are they right or wrong? etc. His book has a chapter analyzing the 2007 flash crash and implying that it was because an institutional stat-arb (statistical arbitrage, using reliable statistical indicators to generate money) was dumping their investments and that affected all the other stat-arb firms

So back to NVDA. haven't been paying much attention to it (well to markets in general, lil burnt out,) but anyways;

NVDA has a lot to win through AI. They've been betting on it for a long time and they have important tech for ai. e.g. CUDA. which means that data/computation centres will likely be using nvidia processors

training a large ai system is computationally expensive. literally a few million dollars worth of training at least. (this might go down if people find more efficient ways of training ai systems.) And depending on the model, it might also be expensive to run the system

So, we have a lot of existing AI hype ("AI will change the world") and lots of doomerism "AI will replace all the jobs" (if you can call it doomerism i guess.) So people will be buying nvda like crazy. not going to go down anytime soon unless there's a huge disruption in either nvidia or ai

tl;dr EMH is fake af don't listen to it. go read Adaptive Markets by Andrew Lo

Jesus Christ

This isn't a graph, this is a mistake:

Somebody accidentally added the wrong row in their excel sheet, market cap curves don't look like this.

What, and I cannot stress this enough, the fuck, is happening to the GPU market?

Well AI I guess, but damn.