Back

@thezoq2

Output is mostly :

@thezoq2 you should probably add an additional row for "can kill you" filled in with a orange background white circle (sorry couldn't find the actual Unicode symbol) with an asterisk next to it to denote only on the specific circumstances (i.e Killer Rabbit of Caerbannog, self-driving car)

@thezoq2

@thezoq2 oh, also Bunnies are sentient unlike AI

@thezoq2 can you edit this to add an alt text please, so that screen reader folks can enjoy it too? #Alt4You Feel free to use this one!

Table comparing a bunny and a machine learning algorithm, captioned with "Stop Doing ML". They mostly tick the same boxes:

Hard to train.
Dubious output from good input.
Why is is so fuzzy?
Can be quite fast.
Entertaining when not destructive.
Likes to keep things hidden.

They differ on the last item, though:

Cute and cuddly (only the bunny)

@baardvrouw : this you ?

@thezoq2 As a bunny owner and an ML researcher… this is all true.

I don’t know why I do both when I could *just* be a bunny owner.

@thezoq2

@jbaert @thezoq2 Also, one can cause lung disease and the other doesn't. #BirdsAreReal

@jbaert @thezoq2 please add alt text...

I've provided some text here.

@thezoq2 @Lazarou Eats its own ? and .

@thezoq2 cc @tiarael

@thezoq2 @cenobyte

@thezoq2 You forgot "Easy to sell to gullible billionaires for $$$$$$"

@thezoq2 START DOING BUNNIES

@dgar @thezoq2 I would like to unsubscribe

@thezoq2 bunnies are actually quite easy to potty train, much easier than LLMs

@thezoq2 The rabbit is tastier too.

@thezoq2 You wildly overstate with your screaming "STOP DOING ML" imperative.

There are fine and good things to do with old-fashioned techniques, such as linear regression—given ethical training and deployment. Even neural networks can be really great, consider AlphaFold2.

Possibly you mean "STOP DOING LLMs", of the plagiarism machine sort??

@kentborg @thezoq2 It's not so much that there's too much ML, just that you should have more bunnies than ML since the bunnies are clearly superior.

@thezoq2 I think ChatGPT was the 1st time I thought for real we could head down the path towards an AI Apocalypse

not because it will lead to a race of evil geniuses. but because (1) it could potentially put millions if not billion out of work, and (2) massively increase carbon and heat emissions -- all for deliveting results about comparable to what a human could do, already. however imperfect or wrong at times. heck humans are imperfect or wrong at times

@thezoq2 Shits everywhere if the input is faulty:

@thezoq2 And imagine an Easter machine learning algorithm instead of an Easter bunny.

@thezoq2 often involves measuring flops