AI Robots have won.

AI Robots have won.
#shorts #AI #robots

[Ramil_]

i love how your new haircut makes it look like you have a bad greenscreen

[Banana_Fusion]

Lmao the constipation and love song is actually really well crafted- lot of dual meanings throughout

[stupidityinc.]

Fun fact my English teacher was having an existential life crisis about how his whole job is pointless when humans can learn anything with the press of a few buttons after discovering this AI last week.

[burrito]

“5 paragraph essay”

me, a person still in school: fr?

[theEnemyAnemone]

dude yesterday I asked chatgpt to write a song about barbie running a hydraulic press channel and it wrote an absolute banger
we’re all screwed

[chxxstheory]

I read the Ed Sheeran lyrics and I absolutely lost it I'm literally laughing in tears while writing this 😭

[FrigidARC]

“An Ed Sheeran song of being constipated”
“Gotta push through the pain”
“To find the relief”
“I need to find a way to let it go” 💀

[Decimator_]

Well it seems all my English homework is about to get to absolutely destroyed right now.

[IAmInWalls]

Heheh... fools, I replaced my triple A's with car batteries! MUAHAHAHAHAHA

[bolbol]

What’s that one Jurassic Park quote? "Your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should."

[anelander]

Gotta go and make that song real, the blend between love and constipation is so well done lmao

[mechamaster]

I like how my boy is wearing bunny slippers😂

[chiachimariachi]

Top 10 quotes taken moments before disaster:
#1 - “Hey Chat GPT3, what’s wrong with the human race and how should we fix it?”

[nathanpixelkid]

Holy shit, my English grades are about to skyrocket

[rkvkydqf]

Just a small interjection to calm anyone too impressionable about this kind of stuff.

1. It's a "text transformer", a very cool NN architecture for this type of model, which just splits the text into tiny snippets called tokens and predicts what would come next.

It doesn't have any traits common with consciousness. It's quite challenging to make it consistently answer as itself, not write nonsense, repeat the same word over and over, or randomly cut off, due to how it actually works.

By the way, this type of model has a very curious little phenomenon called "answer hallucination", or more accurately "Socratic parroting".

2. This isn't new. OpenAI, the company behind it, made the tech behind it a few years ago, but have decided now to see if they can stop their models from acting as BS generators by opening it up to the public to see if enough finetuning data changes things, creating the ChatGPT you know now.

3. GPT-4 is quite a silly concept made up by the most clickbaity obscure websites on the internet.

Even conceptually, it's more or less impossible to get better than it's now. GPT-3 is essentially the long-known GPT-2 with terabytes of training data and millions of dollars in computing costs. The closest thing is the GPT-3.5, which was just a tiny update to make it lie less.

All of this is just a result of throwing money at the problem, which of course comes with diminishing returns. If anything, we might just overfit it, fancy speaking for letting the model detect random coincidences as patterns, decimating its real-world performance.

The future of this tech is more finetuning, better architectures, and smaller models, possibly with the ability to query real-world databases. All solving real problems of them making stuff up, burning throw all that big-tech money, or becoming a reputation disaster for a company.

4. That LAMDA incident is quite cringy, actually. The guy in question was a religious fanatic who against all evidence decided to go with a rather insane to anyone who knows how these things work conclusion, that it's actually sentient.

These things are basically generating the mean value of the entirety of the dataset. So, how often do you see people on the internet call themselves non-sentient? Did you read a sci-fi novel where the AI calmly explains it's in fact just a boring machine with no inner thoughts?

That is the simple logical answer to that question.

5. You might believe these things are smarter than a person, but many researchers are starting to doubt it's smarter than your iPhone's predictive keyboard.

Recent research on this matter questions if it learned anything other than generating convincing BS based on the fact it couldn't generalize anything from the training data, reliably understand grammar concepts like negation, or comprehend its limitations.

[wiffleballkidsleague]

Singing the lyrics in my head in the rhythm of your average ed sheeran song and it working makes me so happy

[Hugelag]

On one hand, that's terrifying. On the other, students will never have to write another essay again.

[arandombanana]

Bro went bald and just put a dark oak plank half slab on his head 💀

[Chese_frfr]

Just gonna excel at all my homework and just fail all my exams, seems like a fair trade if you ask me

[AverageEnthusiast.]

Im leaving this comment so i can see this when i need to make an essay

Edit: i will update once i finally use this for an essay

Edit 2: I used it!