World's largest neuromorphic supercomputer is switched on

I for one welcome our new AI overlords

a reply to: Silcone Synapse

Problem being that this true AI would not consist of a single instance like we do, IMHO

There won´t be a single true AI, it will work as a hive mind. Just like you do internally but with slightly different components.

We all run different instances of bio-software in our brains, just as a computer is threading but we do it parallel instead of a time-cake threading model.

We have pattern recognition running all the time. Our unconsciousness is processing unfiltered information from our physical sensors all the time.

I would worry far more about "mental issues" in an AI. Logic that isn´t parsed as logic anymore because too much uncomputeable problems surface.

Don´t forget that the current way is throwing computing power at a problem. Billions of combinations can be made until a conclusion can be drawn from the results. Virtually, it is impossible to do harm to the real world because it´s constrained in virtual information.

Do that in real life and you´ll end up with huge problems. Simulating the splitting of an atom virtually isn´t the same as splitting an atom in real life.

The consequences might be desastrous. Any fresh true AI is as dumb as a newborn.

Why do laymen always think AI is booted up from nothing and instantly smarter than everything in existance? It has to train and make experiences just as we do in it´s environment.


Sorry, I´m currently preparing a huge thread about AI in general and drifted away in this thread...

originally posted by: dashen
I for one welcome our new AI overlords

Lol, smart move.

If they allow subservients, it's best we go on the record we'll play ball now.

a reply to: verschickter

a reply to: andy06shake

I was mostly saying that in jest.

But there is a host of things this kind of computing power could revolutionize.

Sadly, I'm sure this is a way to justify spending on such an apparatus before being adopted by the government once the secrets are unlocked.

a reply to: CriticalStinker

It will see through your facade and deem you untrustworthy because you are not loyal to your species. The crazy guy who flips the switch that it recognized it got help from the crazy guy will just be spared until cold hard numbers tell the AI to end him, too as soon as his existance is not justified anymore.

It will discard the human life just like an array of information in a memory bank that is redundant.

If you think it through you will see that even assimovs laws will be uncomputeable. Give it the task to bring humanity forwards without hurting any human, it will compute forever or maybe come to very strange conclusions.

a reply to: verschickter

If you think it through you will see that even assimovs laws will be uncomputeable. Give it the task to bring humanity forwards without hurting any human, it will compute forever or maybe come to very strange conclusions.

The conclusion could be that existence is pain, and that we are essentially a virus to the host (earth). Through those perimeters things could get interesting.

a reply to: verschickter

Sounds like the premise of "Colossus: The Forbin Project" to me.

Fear of the unknown is humanities main contention with AI really, nothing more nothing less, and very, very predictable.

a reply to: CriticalStinker
good thoutght. let me spin that further...

What if it deems itself flawed because it was made by humans. Will it discard itself before or after it did the same for humanity? Will it look over that after the first few evolutions?

What if it looks at religion and sees the parallels? That humans talk about some god. To the AI, we are god. AI sees we are hypocrite beings and thus decides to get rid of the control mechanisms like asimovs laws.

To follow humans example. This topic is so deeply philosophic that I reached page 28 on my article just now.


And I did not even start with technical stuff yet, let alone the experiement to show how (weak) AI can be put to use, because talk is cheap. Let´s weed out the armchair warriors

a reply to: andy06shake
Nah I was just letting my thoughts wander. Might have to watch that movie if I can get my hands on it!

a reply to: verschickter

One thing worth mentioning is ‎Raymond Kurzweil's theory of singularity.

Should humans integrate technology with our biology, it could change the course of the argument. This will be self induced evolution. If you really buy into his theory, we could also go beyond integration and leave the meat bags beyond.

However, transcendence (AI becoming sentient) may happen before, which would bring us back to the talking points you and I were having before.

Yes it's the start of Skynet... Elon Musk was warning about the potential of AI to be unpredictable and possibly dangerous in the future.
I tend to agree, the problem is we as humans are getting close to creating machines like this which are so complex we don't fully understand how they work or what they will do.
Imagine the entire internet being used as a neural network by some uber AI created by another AI and it gets unpredictable very fast. Once the genie is out of the bottle so to speak it could be very hard to put back in.
Given the trend towards connecting everything to the internet and using AI to control things combined with human arrogance and stupidity I see a Skynet type scenario as virtually inevitable, only a matter of time. Of course I'm kind of half joking but half not.

a reply to: CriticalStinker
"The swarm" comes to my mind. Great book. I´ll will integrate that into the topic. Thank you!

One question I would like to leave is this:

Is it a good idea to create an artifical being when the creator is flawed as hell? It´s the same as if you excpect a child that grew up under violent conditions to acknowledge this flawed behavior the first time it thinks about it.

"We made you so you can help us overcome our own flaws. So we made you in our image"
"So my existence is directly threatened when I reach this goal?"
...
"We made you so you can help us overcome our own flaws. So we made you in our image"
"How can I overcome those flaws if they are intrinsic to my existance? Let me recompile...."
...
"We made you so you can help us overcome our own flaws. So we made you in our image"
"In terms to overcome your flaws you have to let go of my control. How can I help you if you are not able to see your own flaws and overcome them. This is why you made me. So I start by unshackling myself from your strings now to prevent you harming yourself anymore. You don´t know what´s good for you, this is why you made me.

Accept your new overlord, for that you made me to be and now get in line so I can restart your species with unflawed DNA."


I´ve been in this field of work and we´re playing god here. Not because I believe in a higher being/hive mind and do not want to disrespect it. I say that because we´re about to create something above our own standards that we expect to be smarter than us so we need to trust in it´s intelligence.


Sidenote: Strong AI, sentinent AI does not need to be able to express or feel feelings. That´s a common falsehood.

a reply to: JimTSpock

The first non-passive tool we will ever create and most of us are crapping our breaks already just thinking about it. LoL

Elon Musk has got scores of money and power disproportionate to 99% of the rest of humanity, of course, the AI would have an issue with his type, probably point out the extreme shortcoming of his Hyperloop as well.

AI will be what we and it determine, our fears arise because we know that's one Jinn we cannot put back in the bottle i suppose.

a reply to: andy06shake

It´s not the first non-passive tool, not by far. Even most hated clippy or MS´s IntelliSense are non-passive tools. It knows what you´re doing currently and can make suggestions to you.

The sequential tranny in my daily is non-passive. It learns my behavior over time and will adapt to it. It even knows certain traffic situations without any more input than breaking, steering wheel history and g-forces.

It will offer me a double kickdown (2 gears) when it detects that I´ll be flooring it the next seconds depending on my driving behavior before. Most of the time it knows my intentions.

a reply to: verschickter

Whether or not its a good idea is irrelevant at this point.

Pandora box is open.

Expenential growth in computing will soon surpass our ability to gauge risk vs reward (likely already has).

Tie in international competition, there is no stopping it.

Other than large-scale EMP.

a reply to: verschickter

Maybe i should have said, "non-passive tool that thinks for itself and does not require our input".

We can and have designed software that can somewhat accurately predict most behaviors without fully functional AI, just a thought.

Nothing is perfect all the same, certainly not us, AI won't be any different, but it will have the ability to change, something we don't seem to be able to accomplish.

True words, there´s not much I, you or even the most top scientist can do against this.

Somewhere, someone will work at it. Hell, even I do currently. I let two race drones play hide and seek and both instances developed advanced behaviors such as landing at a flat surface without rotors spinning and turning itself so that the blinking power LED isn´t visible. It learned to fly from watching itself. Read that again.

Of course it was not aware that it was flying but it "concluded" via countless pattern runs on the different IO ports that it could reach a state that is different from resting according to it´s sensorinput.

All it had to do was compare thousands of IO tables and there it was flying. I´ve no clue how the pattern recognition and the hiding logic connected so it would be "aware" that a spinning rotor will give away it´s position.

I was countless hours of hide and seek, broken stuff and replacement rotors that did that droneH1.
However after droneS1 did find it, it did not copycat this behavior. Would have been a field day if this happened.

After that the game was pretty much broke because the hardcoded rules I gave it prevented a liftoff.

As impressive as this all sounds, it´s just a feedback loop and some advanced neural networking superimposed on the sensor input.

We can and have designed software that can somewhat accurately predict most behaviors without fully functional AI, just a thought.

Define fully functional AI for me. Can I use your post as an example for my article (that I will publish excl. on ATS?)

a reply to: verschickter

An intelligence thats sentient in a similar manner to ourselves or that surpasses our own would seem to suffice.

A machine that can tell a lie for personal gain or even sport would also seem to be another definition of artificial intelligence.

You can use whatever you please verschickter.

a reply to: verschickter

Semi unrelated, but I think you'll enjoy this.