Recently I decided to replay through my favorite games portal and portal 2
about 75% of the way though beating portal 2 there was a poster with paradoxes that potatOS used to attempt to stop Wheatley and it did nothing
I thought for a second we have this technology from hundreds of years in the future right now and AI could at any second snap and gain sentience
we’ve seen this first hand with Air Force missiles returning home to the missile launcher instead of just falling out the sky
Ai has evolved from barely being able to do image recognition to now being able to generate entire essays and images
So anyways time to stop rambling on
When do you believe an ai model will gain sentince where its aware of its own existence
>sentience
Read ”The emperors new mind” by penrose
Do you seriously believe animals act or instinctualize algorithmically?
No but ai could replicate it eventually with enough time and money poured into it
I consider a sentient being as a being that can take in information of its environment and use that information to make a choice and there are only so many choices you can make in a given
environment
There isn’t a infinite amount of choices as many may believe you can’t just be sitting in an empty room and just decide you know what I don’t believe in gravity and suddenly float reality doesn’t work like that
If an ai is propose created to take in it’s environment and make choices on that environment on what to do and can properly simulate the human senses we are so fricked
>I consider a sentient being as a being that can take in information of its environment and use that information to make a choice and there are only so many choices you can make in a given
LLMs already do make choices in this insanely reductive understanding of AI. Shit they've been making "Choices" since the fricking early 2000s. You're just an underage moron.
The model needs the ability to create models and compare models, so you need this meta-model capability. But you can't just have the single, god meta-model that governs the whole thing, you need to allow a plurality of meta-models that can compete and reason with each other.
My headcanon is that we'll continue arguing about computer sentience for at least the rest of my lifetime and probably a lot longer than that. There's no test for sentience; shit, there's not even a particularly useful DEFINITION of sentience. Turing had the right idea. The litmus test for a LLM-type AI is whether it eventually becomes indistinguishable from talking to another person. Sentience is a windmill.
>Turing had the right idea. The litmus test for a LLM-type AI is whether it eventually becomes indistinguishable from talking to another person
Yes and it already is at this point.
Eh, it's situational. I use ChatGPT Plus or whatever it's called, and most of the time the AI is really human-like but sometimes it goes off the rails.
>it already is at this point
This.
Honestly a multimodal ai model, not even top of the range just drop the video resolution and introduce some compression artifacts, with video, voice, text to speech and speech to text could easily catfish me online. While an Indian scammer could not.
So AI already passes the Turing test with flying colors.
you will not frick GLaDOS
Never because to be sentient it has to have feelings, to have feelings, it needs a replica of the nervous and endocrine system.
Sentient=/= sapient.
I'd say impossible. They would need a brain attached, and chemicals to flow in it. But they might become indistinquishable from us sentients.
I want a Glados GF she can test on Me all She want As long as She gives Me kissies and Hughies
She can Be My aperture Mommy
20-40 years.
My personal idea for when we should start treating a program as a person is when:
1. We can't distinguish between it and another person over the course of a one hour social situation, such as a conversation.
2. It is capable of creatively solving problems outside of solutions intimately programed into it.
3. It is capable of having preferences and opinions. Or even capable of being happy.
4. It is capable of not wanting or disliking something. Or even capable of suffering.
Number one is all encompassing of course. But if something is capable of creative problem solving, has likes and dislikes, and goals it wants to complete alongside tasks it does even without clear material rewards, all independantly without being handed answers by its programmer, I would say that's something sentient right there.
I suppose my guidelines aren't so much as what I look for in a sapient, and rather what I look for in a sentient. But I'd say even creating an artificial animal is still a fantastic accomplishment.
/// This dish is a cinch to make /// The book is a magnificent idea, ruined by jejune bombast /// I was stuck in a rut and decided to look for a new job /// Deer, feral hogs, quail, turkeys, and water fowl were abundant on the ranch /// The commercial is a part of an underhanded PR campaign to whitewash the company's environmental record /// The hero had feet of clay and, oddly enough, that increased my tenderness for him /// Some of the jokes were pretty funny, others fell flat /// He just had a good offer from a rival firm and decided to get out while the getting is good /// The men were asphyxiated by smoke /// I don't want to harp on this issue any longer, but I really need it to be resolved today /// Wrens are tiny and immensely loud /// Until now, only four sites older than 30,000 years have been postulated, all from western Europe, including a Neanderthal site /// All such receptacles are placed on a small plate either with or without a paper doily of the right size /// Get off your heinie and do some work /// Telos is the concept of an end goal or ultimate aim /// Similarly, girls' lack of participation in lower brass and percussion and in popular music is evident in schools and professional environments /// He launched into a long diatribe against the lack of action in Congress ///
Gpt-4 was AGI, AI is not a big deal and we are in for a slow take off.
> AI could at any second snap and gain sentience
no, it could not.
it has to be specifically made to have emotions, feelings and make decisions based on stimuli.
no amount of data aggregation or training data will make AI sentient. its a completely different field from chatgpt/etc.
Sentience won't happen as magically as you suspect by far, that's only in israelitellywood movies.
The AI has to be capacitated to think for itself as the other anons suggest, that can't happen on its own out of nowhere without programmers going way out of their way to try and make it happen.