The looming problem with AI is human nature

[Punp]

With thanks to SomaSpice who read my eight sentence thesis in chat and asked me to post it to the forum. Disclaimer: In this post AI (Artificial Intelligence) and ML (Machine Learning) will be used interchangably for ease, but they are not technically one and the same thing.

----

The Meaning Lies Between The Art And The Viewer

The real issue with AI is not that it will become a machiavellian villain, but that the human mind projects patterns onto unrelated events and ascribes it meaning.

The fact we're making machines that are designed to out think us and out perform us is questionable, and not like in a Skynet kind of way, but in a way that we're very easily tricked into consuming patterns whch seem obvious to us but are really just patches we make over reality. The words you're reading right now are just patterns of pixels on a screen which you are reading as sensible concepts because your brain is translating it for you into coherent sentences. Essentially the input needs to be right (or close enuogh) for your brain to do the rest of the automatic process. It'll join the dots toegther even if it's not quite pefrect.

But They're People!

This issue is already apparent with one of Google's researchers trying to liberate the software and give it rights. Based on his religious and occult history, Lemoine was primed to a certain way of thinking about events and the patterns they formed. When a researcher interviewed the AI, it came across as a machine. Lemoine claimed this was because the researcher was treating it as if it were a machine, and they needed to treat it like a human. Doing so resulted in more human results which, Lemoine claimed, was LaMDA acting how it thought the researcher wanted to see it. [link to article about Lemoine] [An interview where he believes LaMDA is his friend] [Guardian article] (Sorry, I'm unsure if I've got a link to the above discussion)

The real concern should be in personification of robots as thinking, feeling beings, as happened with the Mars Rover v the General Public. When the Mars Rover "sung happy birthday to itself" it was a nice little performance by the tech team to provide the media with a story, and not (one of the scientists hastened to add) because it was feeling lonely.

In spite of this, the top Youtube comments are probably what you'd expect.

This level of deception is a big concern. It's not even that the AI is deceiving us all the time - it's that people are unconsciously being deceived by their human nature to personify things.

Manipulation By The Machine and The Man Behind The Curtain

It's important to remember that AI is often a reflection of the people that make it and the data it's trained on. After all, who could forget the Racist Microsoft Chat Bot Tay or the Twitter cropping algorithm. A lot of AI has been neutered by their creators in a bodgy patch to prevent its misuse - like most of the AI art programs which will not render a proper face for fear of deepfakes.

However, for each AI ethics researcher, there are ten scriptkiddies out there looking to make some porn. A lot of the ethical questions - like "should we just scrape artstation for all the art we like and replicate it with our AI" are just being bypassed with a quick hand-wave of "it's just for research purposes only you guys wink wink".

And that's the other danger of AI. It enables people without proper grounding in a subject to make new content (derogatory) quickly and without thought to the technique, the context, the rammifications or the understanding. Ethically questionable material which would've once taken a team of many people to create where someone along the chain would've said "I'm not painting a dick on that".

Further, among the haze of "the AI just did it!", there is a very solid opinion embedded in the function of the machine put there by its creators - sometimes without them being aware - and with the power to manipulate us in the most subtle ways, it's concerning what new unpleasantries we might find.*

* If that sounds ominous and implausible, consider that there are currently AI algorithms at work on Youtube and Twitter which are specifically given the criteria "keep people on the platform for as long possible", and are meeting their targets.

 

[manpaint]

One of my biggest fear is that AI will be used in the fture to create content cheaply. Think about the state of the video game, movie and music insustry and now imagine if every new art released was designed by an AI. Not only will it make some people lose their job, it will be even more soulless than what is being created now as it will follow patterns in a database.

 

[Punp]

One of my biggest fear is that AI will be used in the fture to create content cheaply. Think about the state of the video game, movie and music insustry and now imagine if every new art released was designed by an AI. Not only will it make some people lose their job, it will be even more soulless than what is being created now as it will follow patterns in a database.

I think it is already responsible for the current state of our media. The pattern has been to produce content which follows trends and provide people with what they """want""" to see. AI is heavily trained only on what people have already written to Wikipedia and hot takes on le Rebbit. We are seeing the media "catch up" on trends that flashed through the internet's fingers six or seven months ago - which is perfect because it's about the time their Facebook demographic becomes aware of it.

I don't believe we need AI for that to happen, but it'll certainly speed up the content churning machines.

 

[manpaint]

I think it is already responsible for the current state of our media. The pattern has been to produce content which follows trends and provide people with what they """want""" to see. AI is heavily trained only on what people have already written to Wikipedia and hot takes on le Rebbit. We are seeing the media "catch up" on trends that flashed through the internet's fingers six or seven months ago - which is perfect because it's about the time their Facebook demographic becomes aware of it.

I don't believe we need AI for that to happen, but it'll certainly speed up the content churning machines.

Oh it definitively had an effect already. At least it's still possible to find good content when you look for it. Even if people are following trends, their content is still undeniably humans (although in most case very boring). In the future, this is only going to get worse.

 

[Mеченый Яков]

I don't know pal, if androids look like this who wouldn't love to give it a hug?

 

[№56]

Disclaimer: In this post AI (Artificial Intelligence) and ML (Machine Learning) will be used interchangably for ease, but they are not technically one and the same thing.

I think this is one of those cases where being pedantic and picky with words is actually useful. If you call a computer algorithm "artificial intelligence" you're anthropomorphizing it, at least in the sense that nine out of every ten people in your audience will automatically associate the word "intelligence" with the human intelligence AI always has in science fiction. They're not going to think of octopi, slime molds, a swarm of locusts, or Cthulhu. Better to use clunky terminology like "machine learning" than to risk adding more fuel to the runaway AI hypetrain.

 

[zalaz alaza]

Did anyone watch the AlphaGo matches w Lee Sedol?

Was an avid Go player for a while and the way AlphaGo really played some new, unconventional, and "beautiful" moves is pretty wild.

I guess the full alphaGo film is available on youtube free now, but i watched this stuff happen live.

 

[Deleted member 3990]

The only real problem I have with the AI is the very real possibility that it'll soon threaten and replace artists. Starting with visual arts, then literature, and finally music; followed by combinations of these (films, video games etc.)
Just like how computerized chess ruined the game for everyone, computerized art will do the same. We'll be living in a soulless world due to our hubris.

 

[W1NTER]

The only real problem I have with the AI is the very real possibility that it'll soon threaten and replace artists. Starting with visual arts, then literature, and finally music; followed by combinations of these (films, video games etc.)
Just like how computerized chess ruined the game for everyone, computerized art will do the same. We'll be living in a soulless world due to our hubris.

I have some doubts about this, art is defined by the deep true emotion and meaning (this applies to all kinds of art), humanity will still be able to comprehend and make up extremely meaningful works of art even if the quality is not 100% robot perfect, if an AI can replicate the true and deep human nature of art then perhaps the AI is more human than AI

 

[CognacDefender]

I don't know pal, if androids look like this who wouldn't love to give it a hug?
View attachment 34338

It is a trick. Her skin is cold latex. Her armour is her body. Her hair is plastic. Her breasts are hollow rubber and jelly. She does not feel anything. She does not have the capacity to feel or make love. Do not be fooled. The moment you are seen as a threat, she will kill you immediately.

 

[CognacDefender]

Would still probably give it a hug though.

 

[Punp]

Was an avid Go player for a while and the way AlphaGo really played some new, unconventional, and "beautiful" moves is pretty wild.

This is a microcosm of the problem. The AI isn't being inventive, but brute forcing through the problem. When you can see all sets of possible moves as well as estimate what moves will be made afterwards, it's hardly a challenge.

Apply this to social situations and making generalisations of people based on the categories they fit into, and our behaviour becomes equally as predictable.

 

[Punp]

I have some doubts about this, art is defined by the deep true emotion and meaning (this applies to all kinds of art), humanity will still be able to comprehend and make up extremely meaningful works of art even if the quality is not 100% robot perfect, if an AI can replicate the true and deep human nature of art then perhaps the AI is more human than AI

@newgrass_c is right. Not because of any magical properties of the machine, but because of the low expectations of the humans interacting with it. The average consumer (derogatory) has less and less of a quality threshold for content they will consume. To the average Joe on a social media platform, as long as it "kinda gives vibes" as other media they consume they'll enjoy it and applaud it.

Instead of improving the algorithm, the trick is to go to the lowest common denominator. This is why China is becoming such a big audience for Hollywood. They like explosions, sexy ladies and content which doesn't require a lot of thought. This isn't a comment on the Chinese culture (or is it?), but by sheer mass of numbers they are going to have a bigger group of normies to cater to.

Art appreciation will still have its own groups of people who will see through the flimsy layer of brain-poking, and AI seems to be mostly training on content which has already come before - not inventing new material based on experiences. I'm sure this will change in time.

 

[JihyoParkXX]

The only real problem I have with the AI is the very real possibility that it'll soon threaten and replace artists. Starting with visual arts, then literature, and finally music; followed by combinations of these (films, video games etc.)
Just like how computerized chess ruined the game for everyone, computerized art will do the same. We'll be living in a soulless world due to our hubris.

I don't think that will happen since it's mostly just mimicking what came before it instead of breaking new grounds and making a whole new original work of art. Almost all of art has taken inspiration with something that's already been made, but in that case it has been created, not recreated as in how an AI would do it. I can't say how the tech will evolve over time but humans would demand originality and denounce uninspired mediocrity. Yeah, the soulless world part I agree with but we can just reject it.

 

[zalaz alaza]

This is a microcosm of the problem. The AI isn't being inventive, but brute forcing through the problem. When you can see all sets of possible moves as well as estimate what moves will be made afterwards, it's hardly a challenge.

Apply this to social situations and making generalisations of people based on the categories they fit into, and our behaviour becomes equally as predictable.

I am very aware that this is a microcosm of the issue you have presented which is why I posted it. As I see it there are a few probems with your take.

The first is that there really should be a distinction between brute forcing and the pattern recognition the AIs use. I dont know if they still use tensors as the last I messed with any AI was a few years ago but I definitely do know that the method AIs use for playing Go is specifically not checking every path and relies much more heavily on pattern recognition and affirmation than it does with brute force. Brute forcing Go was and still is out of the question, another strategy had to be made when solving that problem.

Secondly AIs seem to be able to make predictions we humans cannot based on relatively arbitrary information. These predictions do not become understandable to us even after the fact. This in itself can be accurately described as inventive or beautiful. Here is an interesting article from Bruce Schneiers cryptogram a few years back.

When AIs Start Hacking - Schneier on Security

www.schneier.com

Please take note of this quote:
"In 2015, a research group fed an AI system called Deep Patient health and medical data from some 700,000 people, and tested
whether it could predict diseases. It could, but Deep Patient provides no explanation for the basis of a diagnosis, and the
researchers have no idea how it comes to its conclusions. A doctor either can either trust or ignore the computer, but that
trust will remain blind.

While researchers are working on AI that can explain itself, there seems to be a trade-off between capability and
explainability. Explanations are a cognitive shorthand used by humans, suited for the way humans make decisions. Forcing an AI
to produce explanations might be an additional constraint that could affect the quality of its decisions. For now, AI is
becoming more and more opaque and less explainable."

Finally I think equating human amazement with a level of ignorance is deceiving. I think we can all agree that when the creators of something do not understand the results that very thing produces some level of awe or reverence is due. In the above example none of the people involved understood the accurate results these human created systems(AIs?) produced and yet those systems still produced accurate results. You can equate these results to mere human pattern recognition but that seems to fall short as when lives are actually saved. It really seems that here we have specifically exited that human foible.

Then again I personally dont think this means machines are human , or sentient(whatever that means is beyond me as well). I like Richard Feynmans take on this because it embraces the amazement we should keep as well as keeps our human-ness in check.

 

[bnuungus]

@newgrass_c is right. Not because of any magical properties of the machine, but because of the low expectations of the humans interacting with it. The average consumer (derogatory) has less and less of a quality threshold for content they will consume. To the average Joe on a social media platform, as long as it "kinda gives vibes" as other media they consume they'll enjoy it and applaud it.

Instead of improving the algorithm, the trick is to go to the lowest common denominator. This is why China is becoming such a big audience for Hollywood. They like explosions, sexy ladies and content which doesn't require a lot of thought. This isn't a comment on the Chinese culture (or is it?), but by sheer mass of numbers they are going to have a bigger group of normies to cater to.

Art appreciation will still have its own groups of people who will see through the flimsy layer of brain-poking, and AI seems to be mostly training on content which has already come before - not inventing new material based on experiences. I'm sure this will change in time.

Every day I come to the conclusion more and more that Bill Waterson was a prophet.