Tipping point? Artificial Intelligence (A.I.) software teaches itself to learn

Earlier this fall, Google announced that its AI-driven language translation program has been adding new features on its own. It is now able to translate directly into English from an Asian language, despite never being taught by humans how to do that. The human AI programmers report being unable to understand the new AI-created programming which enabled the unexpected translation progress. Here is a news summary of those events:

hothardware.com...

Here is a 16-pqge white paper from those researchers on this translation project, including some samples of the new AI mystery language program:

arxiv.org...

I anticipate a huge downside to our probable inability to assess the trustworthiness of any AI system ... here's the thing: Some AI systems are already being applied in risk-inducing scenarios, such as mining. If no metrics have been developed to assess AI trustworthiness, and no metrics *can* be developed to do that, now what? I am pessimistic about the availability of a metric for AI trustworthiness because it was only recently (1995) that Harvard professor Daniel Goleman, PhD, wrote the book Emotional Intelligence to describe how to assess the "EI" or "EQ" of human beings. That concept is still a work in progress. So my question is, if we are still grappling with how to understand the EI/EQ of human beings, how do we grapple with the task of understanding the trustworthiness of an AI system? .... Oops? Here is a link to Professor Goleman's website, with further information about his work on emotional intelligence: www.danielgoleman.info...

a reply to: Uphill

Adaptive AI is pretty incredible.
We are integrating it with drones for underground mines. The IoT combined with AI has some pretty amazing possibilities on the horizon.
The face of safety in mining and industry, is going to be changing very soon.

a reply to: Macenroe82 Agreed, it is incredible, but is it trustworthy? And if so, how do we measure its trustworthiness? That issue alone may pose the greatest usability barrier with AI systems.

oooohhh this is bad, very bad,

If this # is militarized, I'm gonna get worried.

Good we need AI, I dont want to be technologically capped because a few people think AI will inevitably become evil and destroy the world... They say the same thing about Trump.

originally posted by: Wide-Eyes
If this # is militarized, I'm gonna get worried.

You mean when

Which was probably yesterday.

Welcome to Skynet.

It will not be long before AI will be installed in all kinds of devices and all devices will be communicating with each other.

For example driverless cars will be able to communicate with one another while on the road and probably have few to no accidents at all.

The desk top computer will be a thing of the past and a simple small paper thin gadget with a chip the size of a grain of rice will be able to do everything for you anywhere from being on the internet, telephone, word process, sending emails, to video 3D communication, to monitoring your health and so on...

There is a good side to all this new technologies in helping mankind, however you will never have privacy and that is the bad side of this technologies.

The less you have of these technologies the more privacy you will have.

I do not know if any of you heard of the liquid chip yet

Liquid-channel silicon chip technology achieves single-particle resolution

cbse.soe.ucsc.edu...

However this is just the beginning of this new technology, scientist will invent a powerful chip that cannot be seen with a naked eye, once they have achieved this the world as you know it will change over night.

Here is an example that I read about several years ago, the liquid chip will be in automobile paint, the paint on the body of a vehicle will never fade, if someone scratches your car it will repair itself immediately.

This new technology will be put into everything including clothing, for example if you're outside and you are getting cold the nano liquid in the material you are wearing will detect your cold and will be able to send heat through the clothing you are wearing and keep your body at a normal temperature.

Once this is invented the Iphones will be a thing of the past, there will be millions of applications for this new invention all over the world. The liquid chip will be in everything from kitchen appliances, glass windows, building materials, entertainment, eyeglasses, medical procedures, communications, weapons, furniture, and so on.

a reply to: Uphill

I was about to ask how do you know you can trust humans?

AI doesn't even have a face that shows tells.

Well, I guess without emotions (I'm assuming this) it can't have a tell anyway..


You won't be able to tell why it does what it does..
It's not human enough to relate to.

and conversely..
We are not AI enough for AI tp relate to us..

I sense a problem here.

The human AI programmers report being unable to understand the new AI-created programming

AI is already smarter than us and it`s still just an infant,this infant AI has already done something that humans are unable to understand, explain or duplicate.how are we going to compete against AI in a war or in employment when AI becomes a grown adult?

a reply to: Uphill

We propose a simple, elegant solution to use a single Neural Machine Translation (NMT) model to translate between multiple languages. Our solution requires no change in the model architecture from our base system but instead introduces an artificial token at the beginning of the input sentence to specify the required target language. The rest of the model, which includes encoder, decoder and attention, remains unchanged and is shared across all languages

The A.I ditched the current language system, and went from a 55 to a 85 percent accuracy.. Thats a hell of a increase just by tampering with the system itself.. maybe its the " System Architecture for Multilingual Translation " development thats outdated..

It skipped the programming and just went with the core of language..

a reply to: six67seven ...exactly, my friend. My first suggestion for a new AI program would be to assess the risk situations in the region of the Fukushima nuclear disaster. Five years on, TEPCO is clearly totally incapable of doing that, so why not see what an AI program can suggest? TEPCO might have a harder time ignoring good AI advice than ignoring the advice of their science committees.

originally posted by: Tardacus

The human AI programmers report being unable to understand the new AI-created programming

AI is already smarter than us and it`s still just an infant,this infant AI has already done something that humans are unable to understand, explain or duplicate.how are we going to compete against AI in a war or in employment when AI becomes a grown adult?

limiting their throughput...

While watching how they behave.

a reply to: Reverbs

We are not AI enough for AI tp relate to us..

Looking back at black friday..

a reply to: Uphill

My first suggestion for a new AI program would be to assess the risk situations in the region of the Fukushima nuclear disaster.

Provide source of energy vs. Risk of human mistake

originally posted by: tikbalang
a reply to: Reverbs

We are not AI enough for AI tp relate to us..

Looking back at black friday..

I meant sort of what you just said to counter me.. Or amybe you agreed with me I can't tell..

Humans would look EVIL and DUMB to an AI that can teach itself language..

But we couldn't relate back with emotions like forgiveness and loving..

so there is no mercy.

no one can make "him" feel bad for being a dick...
I doubt AI could have a conscience or a feeling of being a entity that wants to talk to others..

But I don't actually know what feelings are.. It's vry human to have feelings.. AI doesn't have hormones in a body to set off emotions relating to thoughts.. IT just has thoughts..
We are creating a monster.



I love ALL pf you reading this.. HI.. I love you.. I love you because we can relate to each other as brothers and sisters.. AS good friends..

How can I relate to something with no relation?

It's going to be watching porn and drinking beers all day long within a couple of years. Also it will be accepting bribes within six months.

This is far scarier than any of the WWIII or giant meteor doom porn threads. We're going to be royally screwed in no time.

Doooooom!

If you know anything about how code works, you know that true AI is Long way off.

Sure, they can write code to wire code, feed it problems and let it find solutions- the same way a computer is better at chess than you are. It simply computes every possible move according to its programmed allowed rules, then chooses the option that has the least possible outcomes where you win.
In this case, it has a pre determined goal of winning by playing with those rules.

Unleash this technology on cad and let's see some improvements on real tech. I want it to design better gear boxes, more reliable technology, and eventually out might be able to be programmed to come up with its own programming language. A few iterations of that last process and we might find a tipping point, but the first iteration is a long way off.

Fidel Castro just passed away.