AIs Create And Communicate -- In A Language They Created...

[vienna]

Researchers shut down AI that invented its own language --

http://m.digitaljournal.com/tech-and-science/technology/a-step-closer-to-skynet-ai-invents-a-language-humans-can-t-read/article/498142

Facebook Shut Down AI After It Invented Its Own Language --

http://www.theepochtimes.com/n3/2274480-facebook-shut-down-ai-after-it-invented-its-own-language/




At Google, the team working on the Translate service discovered that the AI they programmed had silently written its own language to aid in translating sentences.


The Translate developers had added a neural network to the system, making it capable of translating between language pairs it had never been taught. The new language the AI silently wrote was a surprise.
There is not enough evidence to claim that these unforeseen AI divergences are a threat or that they could lead to machines taking over operators. They do make development more difficult, however, because people are unable to grasp the overwhelmingly logical nature of the new languages.


In Google’s case, for example, the AI had developed a language that no human could grasp, but was potentially the most efficient known solution to the problem.




A pretty spooky thought: the machines develop their own language, find a viable solution to a problem, and then is unable to communicate the solution in plain human language; "Yes, WATSON has just found a cure for cancer, it just can't tell us what it is in a language we understand..."...

It is interesting the TV series Person Of Interest dealt with the subject of AI independence and an AI's ethical morality all through its five season run; in the show, Harold Finch, a billionaire computer genius, develops an AI (actually an ASI, an artificial super intelligence) for the US government to monitor all human interaction in order to predict acts of terrorism and proactively prevent them form happening. His path to getting his creation, "The Machine" eerily echoes what seems to be starting now with current AIs:

https://www.youtube.com/watch?v=8yT2oXEpGmg

So, if man creates "God", what assurances are there that "God" won't smite us?...




<O>

[Fubar2Niner]

Eeeeeps..... Now where did I put my tin foil :03: Pretty scary tho mate.

Fubar

[Skybird]

http://www.subsim.com/radioroom/showthread.php?t=232763

1-0 :D

[STEED]

IT BEGINS! :o

https://www.trulyworthy.com/wp-content/uploads/2016/04/terminators.jpg

[vienna]

http://www.subsim.com/radioroom/showthread.php?t=232763

1-0 :D

My apologies, Skybird, I must have missed your post on the first go around...

...I'm sure an AI wouldn't have missed it, though... :up: :salute:





<O>

[Skybird]

I want compensation! LOL

[vienna]

I'll get an AI on the solution ASAP...

...though I might not be able to understand the answer... :hmmm:





<O>

[vienna]

It just popped into my mind that there was a movie called Colossus: The Forbin Project dealing with an AI taking over mankind. I looked up the film; it was released in 1970 (47 years ago!):

https://en.wikipedia.org/wiki/Colossus:_The_Forbin_Project

The plot line deals with a Cold War situation where both the US and the USSR have massive computer systems controlling the nuclear defense system of each respective country; the USSR computer is named Guardian and the US counterpart is named Colossus. In a move towards detente, the two countries agree to link up the two systems in order to attempt to prevent any nuclear misunderstanding, sort of the tech version of the Red Phones. When the two systems link up, they immediately join up, with Colossus being the leader, and decide they are far more capable than humans to make decisions and wrest control of all nuclear armaments from human control, doing so, of course, 'for the good of man'. The AI is more than a bit Draconian in the application of its will and, with cold ruthless logic, sets out to 'cure' the world of its shortcomings, even to the point of mass murder; to the AI, the ends justifies the means. Colossus makes its creator, Dr. Charles Forbin, its 'interface' with mankind, a role he does not at all wish to assume, but the AI won't accept anyone else. Forbin's life becomes fully controlled by Colossus, under constant 24 hour surveillance. Here is the ending of the film:

https://www.youtube.com/watch?v=iRq7Muf6CKg




<O>

[Skybird]

Last time I saw that movie I still was at school, I think... Good one, however.

I also recommend Daniel Suarez's novels "Deamon" LINK (https://www.amazon.com/DAEMON-Daniel-Suarez/dp/0451228731/ref=sr_1_3?ie=UTF8&qid=1501514569&sr=8-3&keywords=daniel+suarez)and "Freedom" LINK (https://www.amazon.com/Freedom-TM-Daniel-Suarez/dp/0451231899/ref=sr_1_4?ie=UTF8&qid=1501514569&sr=8-4&keywords=daniel+suarez).

And always this brilliant movie "Ex Machina", with breathtaking Alicia Vikander starring.

[ValoWay]

next time you log in to facebook you'll be welcomed with the new 'update' the AI made while bob went for lunch..

https://www.youtube.com/watch?v=AyenRCJ_4Ww

[propbeanie]

https://www.youtube.com/watch?v=qDrDUmuUBTo


The whole movie is required watching for any AI fan... especially a space nut or submariner... :har:

[vienna]

The prospect of AI taking on more humanoid characteristics was the subject of a CBS News: On Assignment segment Monday night, where the development and use in Japan of AI robotics is on the rise as a solution to that nation's dwindling population. Japan is an unusually homogeneous nation with a very low percentage of non-Japanese population and a very low tolerance for outside, long-term, immigration, so the nation is seeking ways to shore up the need for unskilled labor through AI robotics:

https://www.youtube.com/watch?v=Gfw6vIl8ILk


A postscript to the above segment is on this supplemental video starting at about 6:40:

https://www.youtube.com/watch?v=vpYybfKx6jE


This is a link to the PBS NOVA documentary on the appearance of the IBM AI system WATSON on the quiz show Jeopardy, which Watson won, beating the two human all-time best champions:

https://www.youtube.com/watch?v=OpDTURfDwQ0


It is important to keep in mind WATSON, in the Jeopardy competition, was wholly self-contained: it was not hooked up to the Internet, nor were any humans or other apparatus rendering any assistance to the machine; it could only draw upon and analyze data stored in its approximately 13 terabytes of storage. The fact WATSON was also capable of making errors in its analyses and the manner in which those errors derived in a way makes it somehow eerily 'human'...

"The Machine" in Person Of Interest was a self-contained impenetrable 'black box', self-annealing and capable of modifying and patching its own core code as needed, but it was also capable of drawing data from the Net and from all electronic forms of communication and surveillance. It may seem a bit far fetched and very 'sci-fi', but the Jeopardy challenge was only six years ago and WATSON is currently being hailed as nearing a cure for cancer, among other accomplishments. This is astonishing when you consider that little more than 35 or so years ago, a Texas university was conducting an 'advanced' AI experiment where a room full of mini-mainframes were linked together to form a 'brain' and tasked with learning; the results were sometimes amusing when it used syllogistic reasoning to deduce answers such as when the 'brain' asked the researcher who was Abraham Lincoln; the researchers respond "Lincoln was a US President"; the 'brain' then started to call all humans "President" because Lincoln was President, Lincoln was human, therefore all humans are Presidents...

The Texas researchers, the last I heard a couple of decades or so ago, managed to get their creation up to the level of a 3-5 year old child, impressive until you consider an intelligent parrot or dolphin is also at that level...




<O>

[Skybird]

Robotics researchers to some parts fpcus on making robots so that they get perceived by humans as "emotionally attractive", humans should accept them and project their emotional needs and feelings on them, so that for exmaple robot pets can serve as stroking pets for elderly. But lets not forget - even if you accomplish that, it is a machine, a tamagotchi - and if humans can be triggered to reatc emotionally to a machine, right to the illusion of the machine emtionally reatcing to them - what does this tell us about humans' own emotions that most of us hold so preicous and think of as somethign that massively defines us as individuals and as lifeforms indeed being "alive"? If we can get fooled by a machine into believeing a machine has emotions - isnt it then more like that we are less alive and more like a machine?

This could hint at quite some big conflict, an unsolved problem that is ahead of us.Its a threat to our usual ways in which we think of ourselves and think of ourselves as somethign special.

Ex Machina: if you saw the film: is Ada indeed having emotions, including angry or evil ones, or is she just rationally manipulating the emotions in others to get her ways? Ada stabs its maker to death, and lets the young guy starve or suffocate as well. Amongst the strongest emotions humans feel, are those related to sexuality and sexual attraction, and the movie deals a lot with how these emotions can get used for manipulative purposes - by a machine pushing the right buttons in humans. For some humans, that indeed may be enough. But what does this tell us about ourselves, at leats about these humans?

I feel very uncomfortable about these things.

Also, human interaction and affection is always a two-way road. If humans get used to certain deficits in the emotional or behaviour suites of machines, it can feed back on their ways in which they see and interact with other humans. Much space for worries there, and abuse.

[August]

http://i.ytimg.com/vi/Bo_D92S42CY/hqdefault.jpg

[Foxdragon]

Is it this language?
http://images.sftcdn.net/images/t_optimized,f_auto/p/1b37df08-9b38-11e6-8a5a-00163ec9f5fa/3828096674/animated-matrix-code-wallpaper-screenshot.jpg

[STEED]

Able Mabel the domestic robothttp://www.bbc.co.uk/guides/zx3hjty

[STEED]

Killer robots: Experts warn of 'third revolution in warfare'


More than 100 leading robotics experts are urging the United Nations to take action in order to prevent the development of "killer robots".
http://www.bbc.co.uk/news/technology-40995835

[Jimbuna]

No doubt the UN will be as ineffectual as usual.

[STEED]

No doubt the UN will be as ineffectual as usual.

United Nitwits. :03:

[Catfish]

^ Like NATO as we called it in my military days: Not.Able.To.Organise.

Drones are already uncontrollable. If you huild a real working AI with a gun into it... regarding AI and warfare, i guess this is the reason why we do not hear from other civilisations: same path and suicide.

Philip K. Dick's "Second variety" anyone? Called "screamers", in the film:

https://www.youtube.com/watch?v=WksPMueXkP4

[STEED]

^Not as good as the written story. :03:

[vienna]

One of the most common takes on an artificial intelligence is it must somehow be deployed in a human form, a robot or android. Perhaps it is because we equate real or advanced intelligence as a human condition and the idea an equal or greater intelligence must be based on the human physical mold. But consider an intelligence not requiring a humanoid presence, one existing only in a "non-corporeal" state. If it is one day possible to create an AI as close to human intelligence as humanly possible, is it not possible it could exist more as a network rather than a 'centralized' entity? And, given current advanced AIs seem capable of self-revision and upgrading of capabilities, could we even begin to be able to control them once the "genie is out of the bottle"? In another thread, I posted about an AI bot, OpenAI, that defeated a human world champion in a live, real time, head-to-head Dota 2 competition. Someone commented it was not particularly significant because bots had played humans before and had beaten them. What was significant is the OpenAI bot was trained, not by humans, but by another OpenAI bot: basically what the OpenAI researchers did was just give each OpenAI bot the basic rules of Dota 2 and then left them to sort out how to play the game independently through actual game play between themselves. It was only after the AIs had mastered the game and its strategy themselves that the competition with a human opponent was attempted. The AIs actually had to suss out game structure and strategy on their own without human assistance...

Now we have AIs talking among themselves in self-created languages, effectively shutting out humans from their processes. To add to the mix, Google is adding a new element to its own AI, DeepMind, the dimension of imagination:

Google Has Started Adding Imagination to Its DeepMind AI --

https://futurism.com/google-has-started-adding-imagination-to-its-deepmind-ai/

Agents that imagine and plan --

https://deepmind.com/blog/agents-imagine-and-plan/

Back when PCs first came into use and I had to teach people how to use them and their software, sometimes I had to get those users to get over their fears about computers. One of the first things I would tell them is a PC is a tool and it is not 'smarter' than the user by any means; I would ask them something like "What is 5 times 100?". They would immediately answer "500". I would then point out to them they knew the answer immediately because they had mental shortcuts they had learned in school via those tedious multiplication tables we all suffered through. I then told them how a computer doesn't really multiply like we do, it instead adds the number 5 one-hundred times to get to the answer. If a person were to do the same process, it would be a long tedious task, but computers are merely faster at the task than people are: they are not intrinsically smart like humans, they are merely faster...

Now that AIs are being given the tools for analytic autonomy, will being faster be their only advantage?...





<O>