AI Ethics

http://io9.com/why-asimovs-three-laws-of-robotics-cant-protect-us-1553665410

A bit long winded, but sums it up in simple terms. The choice quotes, if you can't be bothered to read:

Oh snap it does cover inaction in those laws.


We are so screwed. How to stop humanity from killing each other? Kill or enslave them all!

Can't wait for them to figure that out.

Until the computer figures out a way to change its programming to delete those rules from its programming, or computers that can replicate and build other robots that are unfettered by these rules?

Since things seem to have quieted down I'll try to reraise another topic that hasn't been touched yet.

Considering that we're pretty much learning how to create life(robotics meet biotech advancement)and even that we're now starting to use DNA for data storage as it's more efficient than anything else, how does that put humans and life on the planet in general, in perspective?

We are definitely not using DNA for storage, it's a concept and some experiments have been done coding data into proteins, but we are nowhere close to a viable solution. We're also very much in the dark when it comes to AI. We can make it look smart, but only if we carefully control all the parameters it's working within. Give it a situation it's not already coded for and it breaks.

Current computational tech is all digital, humans and other biological life is analog. The smallest unit of measurement with a digital system is either a 0 or a 1, no in-between. With analog it can be any value in-between 0 and 1 with an infinite number of decimal places and thus an infinitely greater range of possibilities. Until we've gotten much better at building analog computers, we'll never be capable of building anything that resembles an AI. It took mother nature a few billion years of trial and error to produce us.

Leave the movie magic in the movies. Writers can do anything, invent anything, break any rule or alter reality with the press of a few keys. They also tend to follow whatever is "in" for a particular culture at the time of writing. Very rarely is science and engineering actually consulted.

Using physicists as consultants for certain science fiction movies and shows (Star Trek comes to mind) has been around for awhile.

There is also a recent trend of Hollywood consulting scientists for at least the past 10 years now, if not more.

Sometimes it works the other way.

Jurassic Park (1993), for example, actually inspired more people to research DNA and encourage more people to become interested in the field of genetics.

DNA is actually very inefficient in terms of the size of data storage. The pros of DNA are security of information and reliability of information. In terms of actual storage, DNA performs poorly due to that fact that much of the physical storage is continuous repetition.

I don't know the numbers offhand, maybe someone can find the actual numbers. But if I were to make ones up, one gram of DNA is theoretically capable of storing 455 exabytes (455,000,000,000,000 megabytes) of data [fact] (A).

However, the way DNA works, that gram of DNA only contains 17 GB of unique data [made up]. So as it stands 0.0037% of DNA is utilized for storing unique pieces of data in nature [combination of fact and made up numbers].

There is also the problem of accessing data. You can read the entire strand, but accessing a single point to access a file is currently not possible (B).

However, there is also another advantage of using DNA, which relates to its reliability, but in physical terms. Since much of the DNA is the data being repetitive, even with decomposition of the medium, data can be reliably stored anywhere from 2,000 years - 700,000 years or more. (C)

A: The eternity drive: Why DNA could be the future of data storage
B: Breakthrough data storage innovation: One DNA molecule can store tons of data for 1M years
C: The eternity drive: Why DNA could be the future of data storage

Too lazy after writing all last night to find you sources on the whole repetition thing. This is my free work, lol.

So I'm working on my PhD in AI actually

When you really look under the hood of these agents, it's almost laughable. There currently isn't a threat of a "unified" agent. These agents all have very, very specific tasks and require massive amounts of data to process (I'm studying storytelling and we might need more literature than has ever been recorded). On top of that, with traditional AI methods, you can't really "teach" empathy -- only emulate it. For the most part, we're not interested in "making robots like humans". We just "want to emulate human thought processes for specific tasks."

If anything, Cognitive Science is a greater threat than AI is.

EDIT: Not denying that discussing ethics in this context isn't important, but people LOVE to panic about this sort of thing without really knowing current AI progress.

Agreed for the most part except this. Mutation doesn't create new information that can be used, it just breaks what it has. Genetic studies show a loss of usable information in mutation over time that multiplies and is passed down on every reproduction. Which is also the reason we die.
The wrong way for evolution.

Also data storage can't really be measured because DNA code can be read in hugely multiple ways meaning the information is muliplitive. Which is what they are trying to map now. See epigenetics. It's like storing the 24 letters of the English alphabet. Yes that is the basic information but the way it's read gives you the information of the English dictionary.

What is mother nature? Certainly not a sentient being capable of immediately evaluating its experiences and to grow and develop its intelligence as a result. As a species we have made some tremendous exponential leaps forwards in scientific understanding the past couple of centuries. If you told people in 1869 that in a hundred years our species would put a man on the moon, they'd have locked you in an asylum.

I won't dispute that people foretelling Judgement Day are probably off the mark, at least for another large period of scientific development... but I will dispute your pretense in saying it outright will never happen. Never is a very strong word and doesn't really belong in hypothetical discussions of this nature.

You are mixing two different things, it's not quite how it works.

Huh?
Maybe I didn't make it clear then.
Multiplying mutations over our lifetime is why we age(causing our body to break down) and eventually die.
Not like getting poisoned or hit by a car. Clearer?

"Ghost in the Machine theory" Shits on your argument as you clearly have a "less than basic level grasp" on this subject.

Can't kill a a human, what if the robot is required to protect a human, indirectly not defending said human would be a violation therefore the robot must attack the other human, this means it could be in a position where it must kill i.e. Gun point.

So easy to shoot you down and barely tried.

Seriously GitGud

This is my field, you don't need to explain it to me. I'm telling you that you mixed things up in the part I quoted.

You claimed mutations cause loss of information, this isn't always the case, hell there are mutations that ADD information.

The loss of telemoric genes that causes us to die is a relatively recent discovery and it's not a casual mutation either, but rather part of the genetic code itself. It is not well understood yet but we're studying it cause reversing it would mean being able to stop aging altogether.

Please give me the examples of mutations of adding new used information then.

Mutation is mutation, don't know why it matters where it happens as the cell works in parts as a unit so damage in one part effects the whole chain link of information.
You also quote me as wrong by using information not known yet. (Cited as 'strong possability' by those doing the research) Good job.
More theories that are told as fact before they are fact. How many more times do we need to hear this.

There are fates worse than death. A sufficiently advanced AI could, completely following the 3 laws, judge humanity to be a threat to itself and, in an attempt to "protect" us, enslave us all into padded cells for the rest of our lives.

Just think about how many seemingly benign things you do on an ordinary basis. I hope you don't enjoy eating ice cream because AM certainly won't be letting you have anything with so much saturated fat. That's bad for you.

There are so many dystopian futures I can imagine with 3-law-abiding AI, it's hard to imagine the rules being in our benefit.

No, you do not understand the subject apparently.

Please refrain.

Your bowing out by using the old KIngnobody put down quote, because you can't answer. Fair enough. Enjoy your fantasy, you seem to love living in them as most of your posts show.

What?
No, you don't understand biology, plain and simple.

Not all mutations cause deletions, they can be additive too. And not all mutations are negative, some are irrelevant, while others are positive. Welcome to evolution?

The role of telomeric tails is not theoretic, it's known, what is still uncertain is the specific way of functioning.

Go be salty with someone else.

also, please remember that the Three (Four) Laws of Robotics are from fiction, written by Asimov in the 40s.

They've lasted so long because they're good starting points, but in no way, shape, or form are they absolute in designing intelligent agents.

(Just trying to sort the science from the science-fiction)

EDIT: And again, the state of the field isn't such that robots would ever be in a position where they would even be able to apply these rules.

If AI already existed then we would already have time traveling AI computers that dictate every aspect of our lives to the point we wouldn't even know ...how ...to ...live ....without ....them

...oh ***

I'm referring to nature, not actually saying reality is a living entity.

All life is the result of random mutations, themselves the result of cosmic radiation causing molecules to break and form new molecules. This was how the first single cell organists started to self organize. To this day that still happens.

My point was that it took a few billion years of trial and error for us to evolve higher order thinking.


This rarely happens. They is very little consultation other then very broad ideas. "So whats a black hole, that's like dangerous right", and other highschool level physics. Star Trek is a good example because it has nearly zero basis on actual science. They just take random scientific sounding words and throw them together as word salad, then have a smart looking actor repeat them on TV. Further they have had to retcon a ton of their material because the physics were plain impossible. The original series had the Enterprising traveling faster then light in normal spacetime. "Warp 5" was, quite literally, 5 times the speed of light (5c). This limitation was known long before the series was ever written, the writers just didn't know better. They had to retcon and hand waive that the magical "Warp Core" does all this crazy impossible stuff. "Warp Core" is Startek language for "The Force", pure space magic that lets the writers do whatever they want. Fun to watch but zero basis on physics.

Hollywood treats Science the same way they do Technology, it's just window dressing for the "creative content".


That sounds perfectly legitimate to a Hollywood screen writer, just like "Plasma conduit manifold".

If you want an example of good SciFi then Ian Douglas's books are great. The Star Carrier series is really good at using actual physics. Like all SciFi he has to take a leap here and there, but everything is based on what we know to be theoretically possible. Joshua Dalzelle is also a good author, his Black Fleet trilogy was pretty solid.

For those that never played Mass Effect

Still waiting for you to show a mutation that adds new information for a advantage in form of a trait encoded into the genome.
I know I'll be waiting forever though as you give sweeping statements of fantasy.

It's never happened and doesn't happen to this day. Which is why there is a debate at all. They are still trying desperately to prove it.
Point is if you had all the elements in the world flying about crashing into each other for an infinite amount of time they are never going to assemble into complex systems that have a language, means to read said language and other systems being able to use that language to build and power life. These different cogs to the wheel all needing each other at the same time for the other to exist. Not only that these different parts also need the will/purpose to carry these tasks out anyway.
It's so far from logic it's laughable. And you need to be honest with yourself step back and see just how ridiculous it is.

It's far from perfect, but I've heard people debate this for quite a while. The science behind Star Trek is much closer to reality than Star Wars. Granted it's still far from perfect.

The whole warp speed thing was created out of convenience since it would be boring to watch a ship take forever to go anywhere.

There are instances where things are just flat out wrong.

There's a great piece by Stephen Hawking on the matter and when he was a guest on the show.

I don't much about the original series, so I am referring to TNG/DS9 mainly.

Personally the only thing I ever read or watched on the science behind science fiction entertainment was how everything in Star Wars is complete fiction and based of stuff they made up.

Star Trek uses real theories and stretches them beyond the laws of physics. It may be a subtle difference, but to university professors and students it's huge.

Actually they proved it in a lab. Amino-acids self-assembled at the right conditions, one of which being the presence of water.
Not only that, but nucleic acids, lipids, they all formed through spontaneous means at the right conditions..it's organic chemistry. Obviously it took forever to progress because enzymes didn't exist yet at the time, but as soon as RNA with its catalyctic function appeared that's probably when life as we call it began.

Dude just open a book of genetics ffs, what do you want me to explain how mutations work? Positive mutations exist and are there for you to study, they come in the form of either a phenotype that is advantegeous or new proteins with better functions.
Example would require me to know which mutations occur by which of the many methods that exist(and that are different for example in bacteria from us), that would be a great strawman to latch onto cause I'm not a geneticist(and I bet you will latch onto it). But the concept exists: there are negative, neutral and positive mutations(lactose tolerance, AIDS resistance, immunity to certain radiations, and many more..). Mutation isn't synonym of bad. It never was, never will.

Now, I intend to keep this thread clean, therefore if you have nothing to say about the topic of AI or evolution of robotics please stop posting.

I'm not sure how much Star Treck authors have consulted scientists, rather than later inspiring them to achieve similar results.

Yes, but nature did it based on randomness and waiting for slow reactions.

We have technology and the discoveries that we make everyday that can be considered our catalysists.
And most importantly we can deliberately chose to make something happen. A neuron without axon can't communicate efficiently, but we can chose to create said tool to improve it if we realize it would be beneficial(not talking of neurons really, I was being metaphorical).