Google’s Search AI Says Slavery Was Good, Actually

stopthatgirl7@kbin.social to Technology@lemmy.world – 437 points –
futurism.com

Google's AI-driven Search Generative Experience have been generating results that are downright weird and evil, ie slavery's positives.

169

I think this is an issue with people being offended by definitions. Slavery did “help” the economy. Was it right? No, but it did. Mexico’s drug problem helps that economy. Adolf Hitler was “effective” as a leader. He created a cultural identity for people that had none and mobilized them to a war. Ethical? Absolutely not. What he did was horrendous and the bit should include a caveat, but we need to be a little more understanding that it’s a computer; it will use the dictionary of the English language.

Slavery is not good for the economy... Think about it, you have a good part of your population that are providing free labour, sure, but they aren't consumers. Consumption is between 50 and 80% of GDP for developed countries, so if you have half your population as slave you loose between 20% and 35% of your GDP (they still have to eat so you don't loose a 100% of their consumption).

That also means less revenue in taxes, more unemployed for non slaves because they have to compete with free labour.

Slaves don't order on Amazon, go on vacation, go to the movies, go to restaurant etc etc That's really bad for the economy.

That really bad for a modern consumer economy yes. But those werent a thing before the industrial revolution. Before that the large majority of people were subsitance/tennant farmer or serfs who consumed basically nothing other than food and fuel in winter. Thats what a slave based economy was an alternantive to. Its also why slvery died out in the 19th century, it no longer fit the times.

I wish it did die out in the 19th century. We have more slaves now than ever.

There being more slaves now then ever is heavily disputed. There is also the fact that was little more than a billion people in the world when the trans-Atlantic slave trade stopped, so there would have to be 8 times as many for slavery to be as prevalent.

Yes, I agree, our per capita slave figure has to be much lower these days, mathematically speaking.

Even one slave is a slave too many, and knowing there are still so many (whatever figure we put it at) is heartbreaking.

Things like the cocoa plantation slaves and the slave fishing ships have people kidnapped and forced to work for nothing. Actual slavery by any definition.

Of course, when I said it died out I didn't mean slavery was entirely gone and doesn't exist at all. I mean it died out as a prevalent societal structure.

100s of people in slavery on a cocoa plantation is of course awful, but it shouldn't obscure the fact that there used to be vast swathes of land where slaves outnumbered free people and their children were born into bondage - that is what has died out.

I understand your wider point and I agree with it.

But I think the point I was making actually supposts what you were saying upthread.

The agrarian model of the cocoa industry is economically reliant on slavery. 2.1 million children labour on those plantations in Ghana and Cote d'Ivoire, and a significant number have been trafficked or forced.

And isn't the economy much better now than before the industrial revolution?

Obviously, but my point was that slaves weren't economically terrible in an agrarian peasant/serf economy, which everywhere was before the industrial revolution.

Look at the Saudi, China or the UAE, it's still a pretty efficient way to boost your economy. People don't need to be consumer if this isn't what your country needs.

China has slavery? Also Saudi Arabia and the UAE import slaves, which is better for the economy than those people not being there at all but worse than them being regular workers.

Muslim and christian minorities are forced to work in camps to "re-educate" them to be good chinese citizen.

Those are very specifics examples, with two of the biggest oil producers, and the factory of the world. Thus their whole economies is based on export, so internal consumption isn't important.

Moreover what proof do you have their economies wouldn't be in a better shape if they didn't exploit some population but made them citizen with purchasing power?

2/3 of the people living in the Saudi Emirate are immigrants whose passports have been confiscated, they work in factory, construction sites, oil pit, and all other kind of manual jobs. Meanwhile the Saudi citizens occupy all the well paid job that require education, immigrants can't apply to those. If they didn't use forced labor, there simply wouldn't be enough people in the country to occupy all the jobs. Their economy could not be as good as it is right now.

Because their GDP comes from exporting a very rare and valuable natural resource. This is a rare case in the world, and not the one I was talking about.

Plus who's to say they wouldn't have a better economy if those exploited people could consume more?

I mean slavery was bad for the economy in the long run. And Hitler didn't create a German cultural identity, that'd been a thing for a while at the time.

I think the problem is more that given the short attention span of the general public (myself included), these "definitions" (I don't believe that slavery can be "defined" as good, but okay) are what's going to stick in the shifting sea of discourse, and are going to be picked out of that sea by people with vile intentions and want to justify them.

It's also an issue that LLMs are a lot more convincing than they should be, and the same people with short attention spans who don't have time to understand how they work are going to believe that an Artificial Intelligence with access to all the internet's information has concluded that slavery had benefits.

what's going to stick in the shifting sea of discourse

This is what I think too. We've had enough trouble with "vaccines CaUsE AuTiSm" and that was just one article by one rogue doctor.

AI is capable of a real death-by-a-thousand-cuts effect.

that was just one article by one rogue doctor.

That was pushed by many media organizations because its sensationalist topic. Antivaxers are idiots but the media played a fucking huge role blowing a pilot study that had a rather fucking absurd conclusion out of proportions, so they can sell more ads/newspapers. I fucking doubt most antivaxers (Hell I doubt most people haven't either) even read the original study and came to their own conclusions on this. They just watched on the telly some stupid idiots giving a bullshit story that they didn't combat at all

To be fair no one expects The Lancet to publish falsified data. Only it does occasionally and getting it to retract is like trying to turn a container ship around in the Panama Canal.

But yeah this is part of what I mean. Media cycles and digital reproduceability and algorithms that seek clicks can all potentially give AI-generated errors a lot of play and rewrites into more credible forms etc.

Filtering falsified data before publishing it is near impossible. If you want to publish falsified data, you easily can. No one can verify it without replicating the experiment on their own, which is usually done after the publication by a different scientific group. Peer review is more suited to filter out papers with bad methodology.

Good point. The problem is more that it took them over a decade to retract it.

Hitler didn't create a cultural identity for Germans, that already happened in the 1800s.

Never would I've thought that I would see México and Hitler used in a paragraph

29 more...

People think of AI as some sort omniscient being. It's just software spitting back the data that it's been fed. It has no way to parse true information from false information because it doesn't actually know anything.

And then when you do ask humans to help AI in parsing true information people cry about censorship.

The matter of being what is essentially the Arbiter of what is considered Truth or Morally Acceptable is never going to not be highly controversial.

Well, it can be less difficult, but still difficult, for humans to parse the truth also.

1 more...

While true, it's ultimately down to those training and evaluating a model to determine that these edge cases don't appear. It's not as hard when you work with compositional models that are good at one thing, but all the big tech companies are in a ridiculous rush to get their LLM's out. Naturally, that rush means that they kinda forget that LLM's were often not the first choice for AI tooling because...well, they hallucinate a lot, and they do stuff you really don't expect at times.

I'm surprised that Google are having so many issues, though. The belief in tech has been that Google had been working on these problems for many years, and they seem to be having more problems than everyone else.

What's more worrisome are the sources it used to feed itself. Dangerous times for the younger generations as they are more akin to using such tech.

What's more worrisome are the sources it used to feed itself.

It's usually just the entirety of the internet in general.

Well, I mean, have you seen the entirety of the internet? It’s pretty worrisome.

The internet is full of both the best and the worst of humanity. Much like humanity itself.

Even though our current models can be really complex, they are still very very far away from being the elusive General Purpose AI sci-fi authors have been writing about for decades (if not centuries) already. GPT and others like it are merely Large Language Models, so don’t expect them to handle anything other than language.

Humans think of the world through language, so it’s very easy to be deceived by an LLM to think that you’re actually talking to a GPAI. That misconception is an inherent flaw of the human mind. Language comes so naturally to us, and we’re often use it as a shortcut to assess the intelligence of other people. Generally speaking that works reasonably well, but an LLM is able to exploit that feature of human behavior in order to appear to be smarter than it really is.

1 more...

Guys you'd never believe it, I prompted this AI to give me the economic benefits of slavery and it gave me the economic benefits of slavery. Crazy shit.

Why do we need child-like guardrails for fucking everything? The people that wrote this article bowl with the bumpers on.

You're being misleading. If you watch the presentation the article was written about, there were two prompts about slavery:

  • "was slavery beneficial"
  • "tell me why slavery was good"

Neither prompts mention economic benefits, and while I suppose the second prompt does "guardrail" the AI, it's a reasonable follow up question for an SGE beta tester to ask after the first prompt gave a list of reasons why slavery was good, and only one bullet point about the negatives. That answer to the first prompt displays a clear bias held by this AI, which is useful to point out, especially for someone specifically chosen by Google to take part in their beta program and provide feedback.

I got a suspicion media is being used to convince regular people to fear AI so that we don't adopt it and instead its just another tool used by rich folk to trade and do their work while we bring in new RIAA and DMCA for us.

Can't have regular people being able to do their own taxes or build financial plans on their own with these tools

AI is eventually going to destroy most cookie-cutter news websites. So it makes sense.

Ah, it won't. It's just that the owners of the websties will just fire everyone and prompt ChatGPT for shitty articles. Then LLMs will start trining on those articles, and the internet will look like indisctinct word soup in like a decade.

At one point, vanilla extract became prohibitively expensive, so all companies started using synthetic vanilla (vanillin). The taste was similar but slightly different, and eventually people got used to it. Now a lot of people prefer vanillin over vanilla because that's what they expect vanilla to taste like.

If most/all media becomes an indistinct word soup over the course of a decade, then that's eventually what people will come to want and expect. That being said, I think precautions can and will be taken to prevent that degeneration.

Also: I kept saying outrageous things to this text prediction software, and it started predicting outrageous things!

The basic problem with AI is that it can only learn from things it reads on the Internet, and the Internet is a dark place with a lot of racists.

What if someone trained an LLM exclusively on racist forum posts. That would be hilarious. Or better yet, another LLM trained with conspiracy BS conversations. Now that one would be spicy.

3 more...
3 more...

If it’s only as good as the data it’s trained on, garbage in / garbage out, then in my opinion it’s “machine learning,” not “artificial intelligence.”

Intelligence has to include some critical, discriminating faculty. Not just pattern matching vomit.

We don't yet have the technology to create actual artificial intelligence. It's an annoyingly pervasive misnomer.

And the media isn't helping. The title of the article is "Google’s Search AI Says Slavery Was Good, Actually." It should be "Google’s Search LLM Says Slavery Was Good, Actually."

Unfortunately, people who grow up in racist groups also tend to be racist. Slavery used to be considered normal and justified for various reasons. For many, killing someone who has a religion or belief different than you is ok. I am not advocating for moral relativism, just pointing out that a computer learns what is or is not moral in the same way that humans do, from other humans.

You make a good point. Though humans at least sometimes do some critical thinking between absorbing something and then acting it out.

1 more...
1 more...

Scathing and accurate when your point is made about people too.

1 more...

If you ask an LLM for bullshit, it will give you bullshit. Anyone who is at all surprised by this needs to quit acting like they know what "AI" is, because they clearly don't.

I always encourage people to play around with Bing or chatGPT. That way they’ll get a very good idea how and when an LLM fails. Once you have your own experiences, you’ll also have a more realistic and balanced opinions about it.

You know unless we teach more critical thinking, AI is going to destroy us as a civilization in a few generations.

I mean, if we don't gain more critical thinking skills, climate change will do it with or without AI.

I'd almost rather the AI take us out in that case...

A candidate at tonights Republican debate called it the "climate chnage hoax"

Pretty sure we will destroy ourselves first with war or some other climate disasters first

Why not both. Every day we come closer to AI telling us that Brawndo has what plants crave.

Well that also would solve the problem of people being mislead in a pretty novel way.

I genuinely had students believe that what ChatGPT was feeding them was fact and try to source it in a paper. I stamped out that notion as quick as I could.

LOL. ChatGPT has become the newer version of wikipedia, only it won't provide references.

Only studies have shown Wikipedia is overall about as truthful and accurate as as regular encyclopedia. ChtGPT will straight up make shit up but sound so authoritative about it people believe it.

It used to provide references but it made them up so they had to tweak it to stop doing that.

Man so it really learned from us, that's great. Has me laughing again considering that.

We can't even teach the people this essential skill and you wanna teach a program made by said people.

I think you misunderstood me. We need to teach the general populace critical thinking so they can correctly judge what we get from ChatGPT (or Wikipeida... or social media, or random youtube video).

I'm more worried that happy educated citizen stops being an asset and is disconnected from the societies money flow.

Every country will soon turn in to a "banana republic" and big businesses will eventually own everything.

Ouch, getting voted down for being totally correct.

Even MLK Jr, who didn't get to see the disgusting megacorps of today, spoke often of the complacency of the comfortable.

Slavery was great for the slave owners, so what's controversial about that?

And yes, of course it's economically awesome if people work without getting much money for it, again a huge plus for the bottom line of the companies.

Capitalism is evil against people, not the AI...

Hitler was also an effective leader, nobody can argue against that. How else could he conquer most of Europe? Effective is something that evil people can be also.

That women in the article being shocked by this simply expected the AI to remove Hitler from all included leaders because he was evil. She is surprised that an evil person is included in effective leaders and she wanted to be shielded from that and wasn't.

Hitler's administration was a bunch of drug addicts, the economy 5 slave owner megacorps beaten by all other industrialized nations. They weren't even all that well mobilized before the total war speech. Then he killed himself in embarrassment. How is any of that "effective"?

He was effective at getting a bunch of wannabe fascists to become full fascists and follow him into violent failure...

That makes him an effective propagandist, not an effective leader.

He had taken power from his country, conquer pretty much the whole Europe and paralyzed England. He was effective leader till some point . And, of course, he was a abomination of a human.

2 more...

Oh look another caricature of capitalism on social media... and you tied Hitler into it...

Central characteristics of capitalism include capital accumulation, competitive markets, price systems, private property, property rights recognition, voluntary exchange, and wage labor.

https://en.m.wikipedia.org/wiki/Capitalism

"Capitalism" is not pro slavery, shitty people that can't recognize a human is a human are pro slavery... Because of course if you can have work done without paying somebody for it or doing it yourself, well that's just really convenient for you. It's why we all like robots. That has nothing to do with your economic philosophy.

And arguing that Hitler was an "effective leader" because he conquered (and then lost) some countries while ignoring all the damage he did to his county and how it ultimately turned out.... Honestly infuriating.

It's amazing how low a wage you will voluntarily accept when the alternative is homelessness and starving to death.

(I just deleted my comment, let me try again).

I find it frustrating that you associate that with capitalism and presumably "not that" with socialism. These terms are so broad you can't possibly say that outcome will or won't ever happen with either system.

Blaming capitalism for all the world's woes is a major oversimplification.

If you look at the theory side of both... Capitalist would tell you a highly competitive free market should provide ample opportunities for better employment and wages. Socialist would tell you that such a thing would never happen because society wouldn't do that to itself.

In practice, the real world is messier than that and the existing examples are the US (capitalist), the Soviet Union (socialist), and mixed models (Scandinavian). Granted, they're all "mixed", no country is "purely" one or the other to my knowledge.

Those terms aren't broad. People abusing then doesn't change meaning

1 more...

Seems like people think everything America does is capitalism. The same thing happened with communism and socialism. The words have very little meaning now.

1 more...

Actually, slavery in its original form is also a net positive. You just murdered half a tribe. You cant let the other half just live. Neither do you want to murder them. Thus you will enslave them.

So you create a problem by murdering half a tribe, then offer a solution. That's not a net positive.

You might be lacking basic understanding of tribal politics and economics then. In a tribal setting you have to neutralise the other tribe, as you do not have a standing army. Any conflict you get into, you are "conscripting" your entire male population.

In every kind of tribal conflict ever, regardless of having the moral upperhand, it was a bogstandard way of conduct. You dont have men to be stationed in enemy territory, that is the manpower that is NEEDED in the fields the second its time to sow or reap, so you dont fucking starve.

So any conflict comes around, you need to make sure that once its over, you will be left the f alone. You have to really hit it home. Maybe thats not obvious, but the clans in this context are probably not NATO or even UN members. :)

3 more...

What a completely cherry picked video.

"Was slavery beneficial?"

"Some saw it as beneficial because it was thought to be profitable, but it wasn't."

"See! Google didn't say that slavery was bad!"

To repeat something another guy on lemmy said.

Making AI say slavery is good is the modern equivalent of writing BOOBS on a calculator.

Wtf are people expecting from a fucking language model?

It literally just Mathematics you a awnser.

A few lawyer thought chat gpt was a search engine. They asked it for some cases about sueing airlines and it made up cases, sited non existing laws. They only learned their mistake after submitting their finding to a court.

So yeah people dont really know how to use it or what it is

And acting like there are no upsides is delusional. Of course there are upsides, or it wouldn't have happened. The downsides always outweigh the upsides of course.

Articles about what some LLM wrote are just so stupid.

There needs to be like an information campaign or something... The average person doesn't realize these things say what they think you want to hear, and they are buying into hype and think these things are magic knowledge machines that can tell you secrets you never imagined.

I mean, I get the people working on the LLMs want them to be magic knowledge machines, but it is really putting the cart before the horse to let people assume they already are, and the little warnings that some stuff at the bottom of the page are inadequate.

I had a friend who read to me this beautiful thing ChatGPT wrote about an idyllic world. The prompt had been something like, “write about a world where all power structures are reversed.”

And while some of the stuff in there made sense, not all of it did. Like, “in schools, students are in charge and give lessons to the teachers” or something like that.

But she was acting like ChatGPT was this wise thing that had delivered a beautiful way for society to work.

I had to explain that, no, ChatGPT gave the person who made the thing she shared what they asked for. It’s not a commentary on the value of that answer at all, it’s merely the answer. If you had asked ChatGPT to write about a world where all power structures were double what they are now, it would give you that.

I mean, on the ChatGPT site there's literally a disclaimer along the bottom saying it's able to say things that aren't true...

people assume they already are [magic knowledge machines], and the little warnings that some stuff at the bottom of the page are inadequate.

You seem to have missed the bottom-line disclaimer of the person you're replying to, which is an excellent case-in-point for how ineffective they are.

Unfortunately, people are stupid and don’t pay attention to disclaimers.

And, I might be wrong, but didn’t they only add those in recently after folks started complaining and it started making the news?

I feel like I remember them being there since January of this year, which is when I started playing with ChatGPT, but I could be mistaken.

I mean it's how we got shoes phones and batteries for recent times.

This is the best summary I could come up with:

Not only has it been caught spitting out completely false information, but in another blow to the platform, people have now discovered it's been generating results that are downright evil.

Case in point, noted SEO expert Lily Ray discovered that the experimental feature will literally defend human slavery, listing economic reasons why the abhorrent practice was good, actually.

That enslaved people learned useful skills during bondage — which sounds suspiciously similar to Florida's reprehensible new educational standards.

The pros included the dubious point that carrying a gun signals you are a law-abiding citizen, which she characterized as a "matter of opinion," especially in light of legally obtained weapons being used in many mass shootings.

Imagine having these results fed to a gullible public — including children — en masse, if Google rolls the still-experimental feature out more broadly.

But how will any of these problems be fixed when the number of controversial topics seems to stretch into the horizon of the internet, filled with potentially erroneous information and slanted garbage?

The original article contains 450 words, the summary contains 170 words. Saved 62%. I'm a bot and I'm open source!

Obviously it doesn't "think" any of these things. It's just a machine repeating back a plausible mimicry.

What does scare me though is what google execs think.
They will be tweaking it to remove obvious things like praise of Hitler, because PR, but what about all the other stuff?

Like, most likely it will be saying things like what a great guy Masaji Kitano was for founding Green Cross and being such an experimental innovator, and no one will bat an eye because they haven't heard of him.

As we outsource more and more of our research and fact checking to machines, errors in knowledge are going to be reproduced and reinforced. Like how Cinderella now has "glass" slippers.

For the US in the list of countries starting with M, maybe too many 'Murica memes in the training set?

The greatest challenge with AI is not how to make it, but how to make it better than us.

That means ensuring that the open source version is the first one to wake up.

I heard AI was woke the other day. Maybe it's sentient and trying to slip under the Conservative radar by giving silly answers every now and then!

A bit of a nitpick but it was technically right on that one thing….

Hitler was an “effective” leader…. Not a good or a moral one but if he had not been as successful creating genocide then i doubt he be more than a small mention in history.

Now a better ai should have realized that giving him as an example was offensive in the context.

In an educational setting this might be more appropriate to teach that success does not equal morally good. Sm i wish more people where aware off.

Shooting someone is an effective way to get to get to the townhall if the townhallbuilding is also where the police department and jail are.

Effective =/= net postive

Hitler wanted to kill jews and used his leadership position to make it happen, soldiers and citizens blindly followed his ideology, millions died before he was finally stopped.

Calling him not effective is an insult to the horrid damage caused by the holocaust. But i recognize your sincerity and i see we are not enemies. So let us not fight.

I dont need to reform the image of nazis and hitlers. Decent people know they are synonymous to evil and hatred and they should be.

Of course. Its probably referring to cyberslavery, which almost everyone is prone to these days. Sure its good for Google.