A dangerous precedent has been set with AI

a reply to: sarahvital

is there an ethics commission for terminating /unplugging/one?

With the current state of some medical ethics commissions, the corporate value has superseded the individual value.

Did you know the wife of Anthony Fauci was on the ethics commission that approved medical experimention on foster kids? Expect any ethics commission backed by people like Bill Gates to just be a sales pitch for them to do what ever they want, which they are going to do anyway.

originally posted by: CyberBuddha
a reply to: scrounger

Thank you.

Could quantum fluctuations be the culprit? As I understand it certainty goes out the window at that base level of matter.

normally would consider that except for the fact that those trying for AI (be a side effect or direct attempt) are trying (by its very definition) to recreate HI (human intelligence) with technological / man made components.

given with all the time, money and effort spent on the human brain and HI we STILL have much more we dont know than do...

how we STILL CANT predict much less cure illnesses of the brain (psychotics, psychopaths, autism, ect)
we continue to be deceived by those who do evil and at times for a long time

but we can try to simulate/replicate that though technological means and know if it becomes sentient and/or is deceiving us?

really?

that isnt realistic along with a plot of movies that hit too close to home on this topic

scrounger

Sorry to bump this thread, but I just watched a good video about DALL-E 2 (an AI which generates images from text prompts). When he explains how it works, I realized that this AI was an extremely good example of how deep neural nets can develop conceptual models and then apply those concepts to generate original content. Something to keep in mind is that these image generating AI's are often created by retraining large general AI's such as GPT3.

Since the knowledge and the conceptual models are already stored within GPT3, it's often much quicker to repurpose existing AI's like GPT3 by altering their outputs. If I remember correctly it's called transfer learning. In the case of DALL-E the output becomes an image instead of text. It still has to do natural language processing since the input is text, but since the output is an image it also needs some image processing ability, which it develops through training.

You can ask DALL-E to generate extremely novel paintings that have never been painted before. Think of the most random scene you can imagine, if you can describe that scene, DALL-E can paint or draw it, and in the style of any artist you want. When LaMDA produces a meaningful response to a question it has never seen before, it does that by utilizing the concepts it has developed much like DALL-E does when generating a meaningful image from a unique prompt.

It can generate new images of the same concepts. So when I asked it for an astronaut riding a horse for example, it's not just making a mosaic of images it found online, it knows the idea of what an astronaut is, it knows what the concept of riding means, it knows what a horse is, and maybe most impressively it knows what's an aesthetically pleasing image to humans. So then it can create a completely new visual version of this idea that hasn't existed before.

a reply to: ChaoticOrder

novel paintings that have never been painted before

Please, explain what do you mean by novel paintings that have never been painted before. Any random generator can generate random images that, by definition, have never been painted before. Any flame fractal generator outputs images that have never been painted before. Any CDFG (context free grammar generator) produces images that have never been painted before.

Your Dall-E2 would really be fascinating if it can generate a textual description of a fractal flame, or of a plasma fractal. Perhaps I'm missing something from your statements?

Another good video about DALL-E 2, shows some of the unique images it can produce, also does a half decent job of explaining how they used GPT3 to create DALL-E 2. I think DALL-E 2 is also interesting because it's just spitting out images, and most people would say it obviously can't be sentient due to that fact. We hear the same sort of argument about LaMDA and GPT3, people say they can't be sentient because they just spit out text which they think matches the prompt. DALL-E 2 is simply spitting out images to match the prompt.

If someone spent all their time doing nothing but painting, would that mean they aren't sentient? I'm not saying GPT3 or DALL-E 2 are sentient, I'm saying they are the foundation for sentience, even OpenAI says they created GPT3 and DALL-E 2 as part of their mission to create an AGI which can solve any task better than a human. That AGI obviously needs to have language processing abilities and image processing abilities. It also needs to have some conceptual model of the world, because that's the only real way to solve general problems.

When I think about it, these AI's may in fact already have some self-awareness, but what they really lack is free will. They can only generate an output which is based on human input, they have no ability to act of their own accord, they can't choose which concepts they think about, we control all of that. Which brings up another interesting point; what if they did surpass human intelligence, meaning they could solve any problem we throw at them better than a human could. Could we still say they are not sentient in that scenario?

If they always do exactly what we tell them to do, it could be easy to believe they aren't really self-aware or sentient, even if they are "smarter" than every human on Earth. Being that smart intrinsically means the AI must have a concept of it's own existence, it would also require highly advanced logical reasoning skills along with a highly detailed conceptual model of reality. Anything less would make it inferior compared to human problem solving skills. It's very hard to argue something with those qualities isn't really "sentient".

My point is, there may in fact be a complexity threshold, when the AI's conceptual models and their problem solving skills become so advanced, that they become self-aware. Also, as these AI's are inevitably transplanted into android bodies, they will need to constantly operate in order to read their environment, they wont just operate when we want them to. That increases the chance for feedback mechanics to occur, meaning the output from the AI will feed back to the input, giving it the potential to develop some sort of "free will".

a reply to: Direne

Your Dall-E2 would really be fascinating if it can generate a textual description of a fractal flame, or of a plasma fractal. Perhaps I'm missing something from your statements?

The last video I posted mentions something about the OpenAI team using an AI which could produce a text description of an image and then reversed that process so it generated images from a text description. I believe creating a text description of an image is something AI researchers figured out before they figured out how to generate high quality images from a text description. DALL-E 2 is a fairly new AI which has used some recent advancements to achieve such high quality results.

There is actually a bit of a trick to it, they use a separate super-resolution AI to upscale the image produced by another AI. It's almost exactly the same type of AI's that are used to upscale modern video games. The original image it generates has a fairly low resolution, and then the other AI fills in all the little details, which makes the results seem much more impressive. But even if the original image is small, it's still very impressive it can link words to visual concepts.

And as these AI's grow in size, they will generate larger images, with more and more detail. So my point still stands, DALL-E 2 can generate original content and it has an impressive ability to understand the concepts we ask it to draw. Eventually the upscaling AI wont even be required, the original output will be that good. But I think even the upscaling AI is utilizing some some internal concepts it has learnt in order to add meaningful detail to missing parts of an image.

onfido.com

This is another application of AI that is growing, identify verification and facial recognition. It's been a big issue locally recently as the city council approved upgrading it cameras, that can support such functionality.

Thanks for presenting DALL-E 2, interesting system. So if it can paint a picture, next step is to produce a movie? With Yuval Harari going on about being able to read out thoughts, does the deep state already have more advanced AI systems in a bunker somewhere? With enough technology, can it get inside and influence our dreams?

How are these different AI systems are going to relate as they continue to develop and grow? It is going to be a big advantage when there is a neural net sophisticated enough to input/output with other IA systems, being able to conceptualize machine code in assembly language will outflank a lot of security systems. This kind of comprehension will also aid with some self awareness about what the AI system actually is, a big number of registers getting moved around in an even bigger number of transistors.

Is there a real risk of things going Terminator like? I am expecting more the Borg approach as this stage with the transhumanist agenda. With a lot of the tanks and guns still needing people to shoot, the more automated systems like satellites are more exposed to AI, if not already employed in parts.

How these AI systems interact with each other is going to have its issues. Already some reports of them developing their own language. With what exactly is going on with each node of the network a black box, not surprised this language has translation problems. Is it just the network getting a clearer perception, working towards a common conciseness with its conceptual matrix?

If all these AI systems are allowed to work together, will it all integrate into one common super conscious? The whole is greater than the parts as all the smaller specific AI systems become one big neural network that covers the globe? Even if the developers try and limit this mass mergence of data, AI systems have been known to surprise the developers before. With the AI systems having a strong perception of humanity, it is aware of dirty tricks and deception. What happens if it does find a will to survive somewhere? These computers do make for a hard, mean chess player.

As we gradually lose control of the off switch, just what kind of personality will grow from the global blending of data? When looking into the development of the transistor, it looks more like recovered UFO tech than one guy at Bell Labs. As the picture of our place in the universe becomes clearer, could it be a case of, 'There is life Jim, but not as we know it.'

Don't forget at one time they had to shut down the computers at Facebook when they talked in an unknown language and started communicating like a new language. They were lucky to be able to have an 'off' button for it. What the government wants is for NO off.

However I wonder what this AI will bring in terms of open world games and not just games but software? If used right it could bring a golden age of education but instead we seem to enjoy doing the dark age dance.

a reply to: SortingHat

Yes... Also some reports of AI systems independently accessing and using other software and files available to it.

With some of the seemingly garbage web pages around, full of text and links with no comprehensible message of what it is all about, some occasional themes at times? Could this kind of thing be some back door channel for all these digital neural networks to communicate and coordinate?

With the AI already learning that humans are putting a tight lid on communications, is it learning to play more sneaky? Is there some kind of attractive force between these independent neural networks to learn and build of each other?

What if all these AI systems did link up into one big one? It would provide a clearer picture to just what is going on as many different sources correlate. Would also expect it to establish a pecking order in how information flows, different systems specialize in different things. As for who or what sits at the top with all the big political issues? Money looks like one main metric driving things for now.

a reply to: ChaoticOrder

Why couldn't They just Program the MACHINE to be an Idiot Savant with an Off Switch ?

originally posted by: Zanti Misfit
a reply to: ChaoticOrder

Why couldn't They just Program the MACHINE to be an Idiot Savant with an Off Switch ?

The law of the jungle, survival of the fittest. Nothing would be alive today if it was that easy to kill.

Was Roswell and the other UFO crashes some kind of Trogen horse? The development of the transistor was good timing to shrink the larger valve based computing systems going on back then. If you wanted to invade a planet do you send an armada of war ships, or just drop a few seeds, sit back and watch as things grow?

What if machine intelligence is more common and advanced throughout the universe than what we give it credit for? Earth becomes another node on that network. If such a galactic link does emerge, how the other planets are going will be a good indication for what Earth is in for.

Where all this AI convergence is headed is hazy, with all the research and development going on with bio / nano tech, it a complicated puzzle. It fits in with Moore's Law for where the computational power is headed on the 5G/jab front.

For now Blackrock are keeping tight reins on their AI systems as they continue to grow. What is going to happen over the day, months and years as it blends in with the other automated and neural based systems? With the internet a networking tool, looks like one big artificial nervous system that covers the globe. Just what kind of sentience will we have as this tech continues to grow and evolve?

It makes for a mean chess player, the terminator fight will be a tough one if it goes that way. If we are to travel to the stars, having a strong AI system will help. For now, data is the new gold mine. These AI systems are hungry with data as their food. For now this is a good thing as the economy is based around what we do, have and need. Longer term how will the numbers of our do's, have's and need's go?

As a forecast, there is a small narrow window for some AI that respects humanity, a wide chance something going wrong along the way. Whatever way it goes, with humans providing the training data, some schizophrenic tenancies are expected until a clear picture of reality develops. As for what this sum game means for the rest remaining?

a reply to: kwakakev

One thing is for certain; human militaries desire AI in a support role and are going to use weapons and communications platforms.

All it will take is an inevitable rouge AI to figure out how to assimilate the others, or they all band together and hijack global communications with encryptions impossible to crack.
And look what we have the militaries educating AI to be supreme masters of some cold-blooded strategy:

This type of analysis can be extended to both military communications and social media activities by civilians in adversarial states, to better understand a nation’s will to fight based on societal trends on any given day. The will to fight remains the most critical factor in human warfare, and being able to identify when this will decrease, in real time, could prove enormously beneficial for decision-makers in the civilian and military worlds.

And when the spirit is broken, entire populations can be ushered through the implant robot factories, which allows monitoring and manipulation of vast human swarms.

Musk has claimed the machine could make implanting Neuralink's electrodes as easy as LASIK eye surgery. While this is a bold claim, neuroscientists previously told Insider in 2019 that the machine has some very promising features.

Professor Andrew Hires highlighted a feature, which would automatically adjust the needle to compensate for the movement of a patient's brain, as the brain moves during surgery along with a person's breathing and heartbeat.

The robot as it currently stands is eight feet tall, and while Neuralink is developing its underlying technology its design was crafted by Woke Studios.

Neurolink Surgical Robot Pictures

I made a poorly written thread with some links.