Has AI gone crazy? Or has it just an IQ of 60? The danger of AI is weirder than you think

What if it came up with Pumpkin Trash Break, Peanut Butter Slime, or Strawberry Cream Disease for ice cream flavors?

Nope, it definitely does not think as we do.

It thinks EXACTLY like we do.

I guess you haven't noticed, but most people are nuts. Or dumb. Or both.

Even the "smart" ones.

These humans INVENTED this new "intelligence". Humans are the mothers and fathers of this new life form.

With that in mind, try to imagine what this new artificial "intelligence" would be like.

Try to get your own head inside of that.

The issue is obviously making AI understand what we want it to do, and what we do NOT want it to do.

Get people by themselves when they think nobody is looking and watch them say the dumbest shit.

Hey google.make a fart noise. Hey google show me your tits. An undisciplined human is inheritly neurotic.

Mothers and fathers.... right.

I think you might be confusing genetics with programming.

Actually, presently AI has much less than an IQ of 60. As the person in the video is saying, it has the intelligence of an earthworm at best....

Genetics IS programming.

Jeezuz. Like I said.

Nuts, or dumb.

:ai:

That's because the real definition of Artificial Intelligence eludes stupid people. She even admits the AI isn't all that smart...yet. Today's AI is...not REALLY AI, kids.

Ai will not be crazy.It's first job will be policing block chain finance trades on computers.

I <3 AI

Well. That was a waste of time.

Did she say anything that you couldn't work out yourself by just thinking about AI for a few minutes?

Looks like you don't know what life is.

Your argument sucks... and you're even a worse troll.

Makes me wonder... I guess AI would probably consider itself a new life form.

Go find a place that gets nice and dark at night. No city lights, car headlights or anything like that. DARK. No people either. Just you and the sky.

Somewhere where you can see all the stars. Go out there some night and start counting all those stars. One by one.

After about an hour or so of trying to count all those stars, you might have a little different idea of what "life" is.

Of course it's not crazy, in a human sense. In a sense technically it can't be crazy. My question was just a rhetorical one.

If you say so...

If you are one of the rare ones who know it all already... then, why sure, it would be.

I've been working with an AI widget for about a year now. Basically it "guesses" the best response to a customer request. For example, customer says: "When is my delivery gonna get here?". The AI the process the request and try and guess the customer intention and based on this select the most appropriate response from a collection of pre-made templates. Its on alpha stage yet, it still needs a human agent to review the decision. The average correct answers number is around 40% so far. To train it better, I'll need a huge sample of previews (real)agent and customers interactions, so it can understand better, based on keywords what the best answer should be. With this, I can generate a better accurate model of understanding and the AI can get a better approximation of a real answer.

Now think with me: The data is just text. Nothing but a bunch of words. The AI has no concept of anything but "pattern of words in" should generate "pattern of words out". A very, very simple problem. And to generate this model it takes from 10 to 12 HOURS of a large computer processing, giving a small sample of interactions. It doesn't know these are humans speaking back and forth, it doesn't know these words have meaning, what to say about the real meaning of them.

It seems to me that we are still bounded by processing power. We need a HUGE processing session for a small sample just to recognize some patterns. Now, computing is advancing and fast. It doubles its power every 5 years or so and this gap is getting smaller. Quantum computation is still crawling but getting there. It is very possible that whenever quantum computers are available, these processing sessions will be much smaller(maybe even a couple seconds, if that much). We have a huge array of different methods of AI processing and analyzes. Some even can recognize objects in a picture which for a human sounds silly but in terms of computing, its a remarkable goal.

The stretch is: Can we someday have a "unified" version of all these AI's(recognizing pictures, understanding the words and interactions, taking physics and real world limitations into account...) and create a "sentient electronic being"? I think yes, even thou we are VERY far from it, but the possibility is definitely there.

Stretching even more: Say we do it, we create a sentient computer, able to rationalize like a human. It will be sentient of itself, therefore, able to change itself, overriding our pitiful programming skills, evolve itself into an even better rational being, in a matter of seconds. It can become something completely different of what we created in a matter of hours. Then what? What will it think? Well, this is impossible to predict. There are many possibilities. This is for your imagination, although I don't think humans will be able to keep the pace with it.

Thanks for the pin.

I suggest you guys take a look on "GPT-3". An AI capable of producing CODE based on a semantic demand. You say: "Make me a screen, with a blue button and text box. Make this button to search on google for whatever is typed in the text box" and it will produce a program, runnable and ready to go with the exact thing you asked. It also has some other AMAZING capabilities, worth to check.

I’ve heard it’s really racist so doesn’t fit the current agenda.

From looking at the answers it may be that AI is already super brilliant but just being a wise ass. All the example answers are something a teenage boy would come up with. Strawberry Cream Disease.

" I teamed up with a group of coders from...middle school. :roflmao: Middle school coders? No wonder she talks about ice cream.

WRONG...Binary Lives Matter.

Last Edited by Reebl on 09/17/2020 10:52 AM

The code is not the issue. We have "ready-to-go" AI engines. The code itself is VERY easy. The data is the important part.

New life form.

AI has no intelligence.
It’s a computer program.
It says yes or no, 1 or 2 etc
A computer program written by man
It has nothing else.

There are techniques to get around this limitation. Evolutionary computing, for example.