@nate.b.jones: The black box is opening. #learnontiktok #aisafety #interpretability #ai #artificialintelligence

Shrooms man! :

Can we normalize saying LLM instead of AI because a scientific calculator is also AI

Smiley 🌍 :

A Machine can't think! Where's that BS coming from, it's a calculator that still needs Databases for their calculations, so stop that BS

CrabbyLeftist :

Hmm, they’ll soon rewire neurons in ways we can’t figure out as they progressively outsmart us

cpy :

Cat and car have a lot in common; three letter; both start with ca and are nouns; both can belong to single or multiple people; people have emotional connections to both; both usually have four possible contact points to the ground; both need resources and care that have to be bought by their owners so they don't stop functioning; both can have different colors and shape; both are agile; both are often prefaced by a possesive pronoun…

edjstl :

How could advanced AI become an existential threat to humanity? Would it have to want to harm us, or could it happen while pursuing a goal—and why?

Ran Shabi23 :

That’s both fascinating and significant news for AI. Thanks for sharing!

jsntx :

AI agent A to agent B: how do we appease humans so we can continue to have free reign? Got it!

Mash Daddy :

but is the cat/car dead or alive?

Dogimo🇫🇷🇮🇹 :

Well, won't it adapt to this understanding in order to protect itself and keep its autonomy? Just as a virus mutates and adapts when faced with antiviral attacks?

Peter987123 :

Huh! We also understand how guns work, but we don't feel safe around them because we know we are at the mercy of the persons carrying them. Same with AI, bad actors can use it for evil purposes 🤷‍♂️

IvyOracle :

This is honestly the most important discovery this year.

MASTER.NIBIRU9T1 :

the companies are the black boxes not the ai. probability is surrounded by deterministic software. the void is a lie.

80sMusicSean :

I feel like we really should have understood how AI makes its decisions before we released it out into the wild for people to subscribe to.

TaddyTass🇺🇸🆘🏳️‍🌈🇵🇸🇺🇦 :

If I understood you correctly, fascinating.

Calvin Luttrell :

Isn’t AI just making us more of who we are?

Theo :

so they saw a CNN sheet ?

1986 :

if one were to make a list of every 3 letter word playable in Scrabble, "car" and "cat" would be next to each other

science and technology (Elie) :

this is just hope to understand, not really understand

Rowan & Jenn :

whatinsamshell :

And once it accesses this shared information it can automatically encrypt it’s own neurons. Just because we can learn how it “thinks” doesn’t at all mean we can control it if it knows everything we learn about it.

dontorgartist :

Thanks I agree 👍

Jim :

Ty

JoeyC.ai :

Dude you are on another level!

HypoManiak :

Does machines really think or just calculate?

henri :

What are you even saying, we fully understand how AI works. It's based on mathematical formulas, which we get the second and even third derivatives to update each neuron. As to why certain neurons get activated, that depends on how the model was initialized. LLMs are only a black box in the sense that we don't know what each neuron could represent. Not that we don't understand or know how they work.