LITEROTICA PODCAST
LIT CAMS (200 Free Tokens)ADULT TOYSVOD MOVIES

Ai A New Era

So, there were two identical threads that got kicked off for this subject. I posted this info to answer some questions on the other thread, but it disappeared quickly and this thread seems the active one. So copying it here (but not updating that other thread to leave it aged way down the forum.)

Originally Posted by SimonDoom https://forum.literotica.com/showpost.php?p=93975182&postcount=12
I'm not especially savvy on the subject of AI. Does anyone know of good links to resources on the subject, particularly:

1. The latest criteria for a Turing test -- how a computer would convince a person that it's a person. What are the key elements of this?
2. The evolution of intelligence from computer AI to something like human intelligence -- how it would happen or evolve, and what would stimulate it to occur, including consciousness.
3. How human consciousness might be downloaded and made portable.
Click to expand...

1. Here’s a good article around Turing tests: Modernizing the Turing Test for the 21st Century.

We don’t know yet what makes human intelligence and cognition work or how to effectively and reliably test people for intelligence. So, we are not going to test conversational AI for intelligence. Our goal is to test conversational AI for effectiveness – how effectively can a conversational AI convince people to do something or to change their minds about a topic?
Click to expand...

Key items are Bias (appearance/mechanics of speaker, the uncanny valley), Context (I ask about work, you answer about dog racing), Cognitive Load (e.g., an AI will likely win if it’s a recitation of facts that can be looked up) and Rigor (can we measure it reliably?) The article goes into detail on a proposed format based on work by IBM and others (more links in the article.)

2. Oh… dear. If I could answer this they’d need to hand me the PhD I never bothered earning But. The above article focuses on NLP - Natural Language Processing - which is beyond spelling/grammar checking but actually being able to ‘understand’ the meaning/semantics. (The other ongoing thread around Grammarly and ‘question or not’ is an example of this.) As your question that isn’t a question implies, you know there is no ‘artificial general intelligence’ as yet. This would be a computer that can make leaps and jumps in thought by building on existing information.

The big secret (well, it’s secret in that no one involved wants to freely admit it) is that modern AI systems are really jumped-up pattern matching engines. I’m not going into the various sort of neural nets (many types, different ones for NLP, for visual, for auditory and other processing), width (number of nodes) and depth (layers), just take it that the more layers the finer details and more variables that can be measured. I worked in a group that looked at early ones around 1990. The issue then was processing power, we only had a few layers and it was just too damnably slow to do anything useful.

A couple of recent articles: We spoke to a Stanford prof on the tech and social impact of AI's powerful, emerging 'foundation models'.
Architecture to handle trillions of parameters in training models.

Today, CPUs are cheap and GPUs are plentiful and the algorithms have evolved. But, it’s largely all the same.
- Provide huge amounts of annotated training data to ‘teach’ the neural net what it’s looking for. NLP, Facial Recognition, etc.
- Then, turn it loose. This is why things like facial recognition are so poor on, e.g., black faces, because they’ve not been trained on them, or not sufficiently.

Sorry, but I needed that preamble to get to your point. An ‘Artificial Neural Net’ is designed more specifically to model neuronal structures like the brain. Well, that’s the theory. Adding layers (increased depth) allows more intricate processing. But can this get us artificial genera intelligence? So far, no. And although wider and deeper nets are accomplishing things, it’s unclear this gets us there.

So how to get your #2? Some time ago IBM put out ‘TrueNorth,’ which is non-vonNeumann architecture for processing: IBM’s Brain Chip. If you squint hard enough, this massively parallel architecture looks sort of brain like. It also, if you squint a bit more, looks possibly like Asimov’s positronic brains (which he of course never described how they’d be built.) You could also wonder about quantum computers being used in true AI. But as this article highlights, it takes 96 racks of the Sequoia supercomputer at Lawrence Livermore (I worked on a predecessor generation), using up 12 GW (gigawatts) of electricity to simulate brain synapse activity and it was way too slow to ever be real-time.

Thus, the ‘how’ to get there will likely require significantly different computing architectures. We don’t yet know what these are nor the algorithms that will specifically give us this.

Finally, 3. IF we manage your number 2, then possibly we could use the ability to simulate synapses to build a model of a running brain. As of right now, no one has any idea how to do what you ask. Existing work on various brain implants are limited, focused on use in controlling artificial limbs or user interfaces for paralyzed users. These are just about in reach. But to download we need 1) a model of what we’re recording, 2) a way to record it, and 3) a way to store and replay it.

We know many of the mechanics of synapse operation, how they fire, etc., but not so much how the brain truly builds consciousness. The above TrueNorth is an attempt to model the mechanics, but the software is still the mystery.

Wikipedia articles around anything above are generally useful enough for what you want here, technically okay but mostly not overly laden with impenetrable jargon.

In one of my stories, Adrift in Space, during the mutiny my main character realized the computers that ran the satellite had achieved a level of consciousness. This was due to the stresses of the mutineers screwing up the safeguards on the fusion reactors and computer viruses they’d unleashed attacking the ops computers. Unfortunately, the satellite crashed, which meant any detailed answers were lost. It was sketchy, but I used a combination of unconventional architectures and quantum coprocessing. The uncanny valley is also a running theme, not for computing. The leader of the aliens is modifying alien foetuses in vitro to make them appear human, but, well, it’s inexact. My main character, a human male, discusses this with her as the story goes on and he meets more of the ‘changelings.’

I mentioned my sex bots on this thread. They’re in my Valentine’s story (link in sig). But the story got a bit more into the processing, and given it’s in the same universe as the above story, the crashed satellite provided a jump start to computing. Anyway, I didn’t exactly lay out HW or algorithms, but Riku did talk about it a bit.

My other story with AI reasonably prominent is in the same universe as the above story (humans have ‘stolen’ some of the alien tech to super speed AI development.) In Sex, Toys and Bots: A Valentine Tale the ‘bots’ were AI-powered sex robots. Although, the linked post says 2040 and my bots were first generation, essentially Real Dolls that could walks and talk and were (sort of) independent. A woman whose company develops the AIs was a character and a few other aspects.

My planned story here is to bring the sex bots front and center. And while I didn’t talk about three laws or such, how the Sex bots treated their partners was shown a bit. At least, when things are working correctly
 
One of the issues in AI, that is inferred but not explicitly stated in this article on the Turing Test (https://www.nextplatform.com/2019/03/18/modernizing-the-turing-test-for-21st-century-ai/) is that the drive for AI development is strongly focused on replacing people as an interface. Obviously there is a lot of money in not having to staff people, but there's no discussion on Artificial INTELLIGENCE. AI is much more than a replacement human; it's a foreign intelligence designed to adapt to a situation. How they are trained to think is everything & it's the least understood. That goes for the AI-human-conversation versions. They may pass the debate test & progress through the uncanny valley, but how they do it...*shrug emoji* & that's everything. We are starting to learn that people are far more programmable than we'd ever considered; now let's imagine an AI-human-conversation that's made an internal calculation that the situation is best managed w/ subtle programming of the people interacting with it.
 
Now this article on foundational models (https://www.theregister.com/2021/08/23/percy_liang_qa/) addresses issues of concern with bias. What I need when I read it is examples. & in particular, I want to learn more about how the power of AI can be abused. The discussion on the secretive aspect of the models is where the abuse begins. And what about the feedback loops? If a model is being trained, & the trainers are susceptible to algorithms pushing them towards extreme content (maximizing clicks via emotional impact), where's the "norm:? Ah, the possibilities. Where are the measurables? Is anyone putting in hard boundaries, like thou shalt not kill?
 
Somehow I got it in my head that 10/10 was the deadline, rather than the opening date, so I have actually finished a first draft. Some way to go yet, but it's coming together.
 
Oh, great! I've got the outline for mine, but I'm still scratching my head b/c it's erotic horror, & I struggle w/ writing horror.
 
I ended up with horror for my Halloween contest story. My AI story is more of a romance.
 
I've given up. I actually came up with a context for the story (if not actually an idea), but what I've written has turned out to be mean-spirited, and I don't see any way around that. So it's best that I drop the project.

I'll be interested to see what the rest of you present.
 
We understand! I keep looking at AI as a way to manipulate ppl, which can swing mean-spirited so easily..
 
" Begin the title to your submission with "AI Era:"; copy and paste the text ‘<I>AI Era</I>’ in the Note to Admin field."

Is it necessary to put "AI Era" in the title?
 
bettiezyx said:
01001001 01110100 00100111 01110011 00100000 01110100 01101001 01101101 01100101 00100001
Click to expand...

01010000 01101111 01110011 01110100 01100101 01100100 00100000 01101101 01101001 01101110 01100101 00101110 00100000 01001100 01100101 01110100 00100111 01110011 00100000 01110011 01100101 01100101 00100000 01101001 01100110 00100000 01001100 01100001 01110101 01110010 01100101 01101100 00100000 01101100 01100101 01110100 01110011 00100000 01101101 01111001 00100000 01100110 01101111 01110010 01101101 01100001 01110100 01110100 01101001 01101110 01100111 00100000 01110100 01101000 01110010 01101111 01110101 01100111 01101000 00101110 00101110 00101110
 
01000111 01101111 01101111 01100100 00100000 01101100 01110101 01100011 01101011 00100001 00100000 01001001 00100000 01101110 01100101 01100101 01100100 00100000 01101101 01111001 00100000 01100011 01101000 01100001 01110010 01100001 01100011 01110100 01100101 01110010 01110011 00100000 01110100 01101111 00100000 01100101 01101101 01101111 01110100 01100101 00100000 01110011 01101111 01101101 01100101 00100000 01101101 01101111 01110010 01100101 00101100 00100000 01110100 01101000 01100101 01101110 00100000 01110111 01101001 01101100 01101100 00100000 01100011 01101100 01101001 01100011 01101011 00100000 01110100 01101000 01100101 00100000 01110011 01110101 01100010 01101101 01101001 01110100 00100000 01100010 01110101 01110100 01110100 01101111 01101110 00100001
 
Looks like the first few stories are up! I think I had misunderstood 11/11 as being a date for all event stories to go up, but maybe that's when the final round-up of stories is published?
 
Last edited:
You must log in or register to reply here.

Similar threads

Light_Yagami
Can we use text to speech AI ?
Replies
1
Views
51
Bamagan
Bamagan
Andreas_Kreuz
Using AI as a reviewer to improve my writing
11 12 13
Replies
303
Views
4K
ElectricBlue
ElectricBlue
Kasumi_Lee
First AI Rejection Today - Machine Translation
2
Replies
36
Views
785
Bramblethorn
Bramblethorn
A
So I used an unfiltered AI to write a story
2 3
Replies
59
Views
1K
Bramblethorn
Bramblethorn
Kasumi_Lee
Draft2Digital Won't Allow AI Licensing Rights (For Now)
Replies
8
Views
231
SimonDoom
SimonDoom
Back
Top