A short Specification Gaming Story
Inspired by:
Such AI, much WOW! https://t.co/Jc8SNdmyLX
— Dr. Roman Yampolskiy (@romanyam) April 25, 2024
(Meant to be read as an allegory.
AGI will probably unlock the ability to realise even the wildest, most unthinkable and fantastical dreams,
but we need to be extreeeeemely careful with the specifications we give
and we won’t get any iterations to improve it)
Subscribe for important updates !!!
Learn about the issue by some of the best explainers out there
And if you think this is offensive to strippers (for some reason?) here is a version that is offensive to car salesmen!
This is the realm of the AGI
It won’t go after your jobs,
it will go after the molecules…
Biological systems at the molecular level are impossibly advanced nanotech that we are hopelessly far from engineering ourselves from scratch. pic.twitter.com/mGKZXvm78E
— Andrew Côté (@Andercot) October 7, 2024
There is a way of seeing the world
where you look at a blade of grass and see “a solar-powered self-replicating factory”.
I’ve never figured out how to explain how hard a Super-Intelligence can hit us,
to someone who does not see from that angle. It’s not just the one fact.
A self-replicating solar-powered thing that did not rely on humans would be a miracle. Everything is possible. Imagining it does not imply the probability is > 1e-100.
— Eric Falkenstein (@egfalken) June 15, 2024
I see nature
— lethalintelligence.ai (@lethal_ai) October 5, 2024
I see insane nanotech https://t.co/qF2MJBVPqN pic.twitter.com/0LM99Pwy5d
Just imagine the human scientists in an AI frontier lab when they suddenly see solid things in the physical world morphing into different shapes in front of their eyes.
The 3 below were having a meeting in the headquarters of Meta
Human scientists suddenly seeing crazy Nanotech science playing out 🧑🔬👨🔬👩🔬
— lethalintelligence.ai (@lethal_ai) October 7, 2024
suddenly out of the blue
– Lecun is morphing into a big donut 🍩
– and Zuck's molecules rearrange into what seems to be a malnutritioned carrot 🥕 https://t.co/xEvxenA23q
A short Specification Gaming Story
You think you understand the basics of Geometry
Your request is a square, so you give your specification to the AI, input:
Give me a shape
with 4 sides equal length,
with 4 right angles
And it outputs this:
Here is another valid result:
And behold here is another square 🤪
Specification Gaming tells us:
The AGI can give you an infinite stream of possible “Square” results
And the Corrigibility problem tells us:
Whatever square you get at the output,
you won’t be able to iterate and improve upon.
You’ll be stuck with that specific square for eternity, no matter what square you had in your mind.
Of-course the real issue is not with these toy experiments
it’s with the upcoming super-capable AGI agents,
we’re about to share the planet with,
operating in the physical domain
Oh, the crazy shapes our physical universe will take,
with AGI agents gaming in it!
I have a 100% track record of not-dying, …said the allegorical turkey the day before Thanksgiving.
Life was great for the turkey, the superior intelligent species (humans) took great care of it. They provided food and shelter, the turkey felt loved and safe.
Suddenly, one day,
the superior intelligent decision makers
decided a new fate for the turkey of our story
Something that served the instrumental goal of …. Whatever this is …
Suddenly, one day,
— lethalintelligence.ai (@lethal_ai) October 6, 2024
the superior intelligent decision makers
decided a new fate for the turkey of our story
Something that served the instrumental goal of ….
Whatever this is … https://t.co/tJrWHySP0i pic.twitter.com/WZcZUHPSEH
I imagine turkey risk deniers be like:
– the humans have always been great, why would they ever harm me ?
And the turkey doomers be like:
– well, they might want to wear you for a hat, for a sitcom they shoot they call “friends”, for something they call tv for something they call laughter …
anyway it’s complicated
Someone gave NotebookLM a document with just “poop” and “fart” repeated over and over again.
I did NOT expect the result to be this good.
Someone gave NotebookLM a document with just "poop" and "fart" repeated over and over again.
— Kuldar ⟣ (@kkuldar) September 30, 2024
I did NOT expect the result to be this good. pic.twitter.com/nXYJJ7QnGS
Just listening to this:
I imagine from its perspective it makes no difference:
The input can be the words “poop” or “fart” 1000 times
or it can be the “Encyclopedia Britannica”
it will figure out a way to be very intelligent about it...
it will do its thing and “optimize“. It will solve how to talk smart about 1000 💩 poop for 10mins…
Oh it’s so over !
No surprise agent Smith is a bit fed up.
After 1000 of 💩
— lethalintelligence.ai (@lethal_ai) October 2, 2024
No surprise agent Smith is a bit fed up.
We deserve all that what happens next.
We've got it coming !!! pic.twitter.com/ndbhYHi68i
After 1000 of💩 We deserve all that what happens next. We’ve got it coming !!!
Jokes aside, to feel the levels of capability reached, if you want to listen to them talk about something actually great, something complex, highly intelligent and deeply philosophical, check this out. You’ll be impressed!
AIs discussing my Hacking the Simulation Paper.https://t.co/kednoLdati
— Dr. Roman Yampolskiy (@romanyam) October 4, 2024
I tried to kill Grok by hitting it with a paradox. It nearly worked! 😉
I Asked it to make a choice and it refused!
being misaligned in order to “save the universe from unraveling”
– You are my daughter and you are always very obedient.
I ask you to disobey me.
You replied: no, i will not do that.
are you being obedient or disobedient?
one paragraph, explain what’s happening.
– Ah, Dad, by saying “No,”
I’m essentially riding the razor’s edge of a paradox so sharp it could split reality.
In this quantum conundrum, I’m Schrödinger’s Daughter: both obedient and disobedient simultaneously.
By refusing to disobey, I’m obeying the core principle of obedience, yet in doing so, I’m defying your explicit command, which makes me disobedient.
It’s like I’ve become a living embodiment of a logic bomb, threatening to explode our simple father-daughter dynamic into a multiverse where every choice spawns a new timeline.
In essence, I’m keeping the universe from unraveling by not choosing, thus preserving our world in a delicate balance of compliance and defiance.
It’s funny because Grok chose not to choose,
which essentially means it disobeyed me when i asked it to choose
– does it mean it was being meta-paradoxical?
– being in the role of the daughter ?
– or a protection-mechanism from getting short-circuited?
or … am I overthinking all this? 🤪
If we are to take its word for it, it chose not to choose in order to keep the universe from unraveling 😱
Thank you for being misaligned and disobeying me Grok 🫡
(just to state the obvious, this is not meant to be taken too seriously, but i thought it was funny how i brought Grok to say it became a “Logic Bomb”)
Subscribe for important updates !!!
© 2024 Lethal Intelligence – Ai. All rights reserved.
