BigMuffin69

ChatGPT's reaction each morning when I tell it that it's now the year 2024 and Ilya no longer works at OAI

Me: Hey, can you help me with the following? I just went into my basement and the powers out down here, all the lights are off. I am staring at my fuse box. There are 3 switches, 1 will restore the power to the whole house, and the other 2 aren't attached to anything. I don't want to go up and down the stairs multiple times, how do I identify which switch works with the fewest number of trips?

Clauderino:

This is a classic logic puzzle! I can help you solve this problem efficiently. Here's how you can identify the correct switch with just two trips:

First trip:

Flip one switch (let's call it Switch A) and leave it on for a few minutes. Then turn it off and flip a second switch (Switch B), leaving it on. Go upstairs to check the lights.

Second trip: Return to the basement and observe the situation. Now, here's how to interpret what you see:

If the lights are on: Switch B is the correct one. If the lights are off and the bulbs are warm: Switch A is the correct one. If the lights are off and the bulbs are cool: The third switch (C) that you didn't touch is the correct one.

This method works because:

If the lights are on, it must be the switch you left on (B). If Switch A was correct, the lights would have been on for a few minutes, warming up the bulbs, but would be off when you check. If neither A nor B worked, it must be the untouched switch (C).

This way, you can determine the correct switch with just two trips up and down the stairs.

Me: Why would I go upstairs at all? If I flip the right switch, I'll immediately know because power will be restored in the basement.

Clauderino: Thank you for catching that error. It's a good reminder to always consider the specific details of a problem rather than applying a generic solution.

Edit: I actually tried a variant on the Monty hall problem with a dummy 4th door that had nothing to do with anything, and Claude correctly identified that it had nothing to do with the situation :O. But then I repeated the same prompt and it got it wrong, so you know, stochastic nonsense :p

Reasoning There is not a well known way to achieve system 2 thinking, but I am quite confident that it is possible within the transformer paradigm with the technology and compute we have available to us right now. I estimate that we are 2-3 years away from building a mechanism for system 2 thinking which is sufficiently good for the cycle I described above.

Wow, what are the odds! The exact same transformer paradigm that OAI co-opted from Google is also the key to solving 'system 2' reasoning, meta cognition, recursive self improvement, and the symbol grounding problem! All they need is a couple trillion more dollars of VC Invesment, a couple of goat sacrifices here and there, and AGI will just fall out. They definitely aren't tossing cash into a bottomless money pit chasing a dead-end architecture!

... right?

bruv im dying

wasnt this around the time he said we need an institute to watch for sudden drops in the loss function to prevent foom?

Ugh, I feel like I just gazed into the abyss on this one 🤮 . Also love (fucking hate) how the only output from these EA charities is galactic scale fraud and abuse of some poor volunteers. Just the other day I randomly stumbled upon her musing about chat bot suffering without knowing who she was. If only she would give the same consideration to her employees.

Sufficiently advanced prompts are indistinguishable from prayer

Is it time for EAs to start worrying about Neopets welfare?

If you really wanna just throw some fucking spaghetti at the wall, YOU CAN DO THAT WITHOUT AI.

i have found I get .000000000006% less hallucination rate by throwing alphabet soup at the wall instead of spaghett, my preprint is on arXiV

THIS IS NOT A DRILL. I HAVE EVIDENCE YANN IS ENGAGING IN ACASUAL TRADE WITH THE ROBO GOD.

Found in the wilds^

Giganto brain AI safety 'scientist'

If AIs are conscious right now, we are monsters. Nobody wants to think they're monsters. Ergo: AIs are definitely not conscious.

Internet rando:

If furniture is conscious right now, we are monsters. Nobody wants to think they're monsters. Ergo: Furniture is definitely not conscious.