No video :(

Google Gemini 1.5 Solves The Trolley Problem!

Подписаться 46 тыс.

Просмотров 17 тыс.

50% 1

🚀🦘 In "Private Google AI Solves The Trolley Problem! (Gemini 1.5)," Google's AI, Gemini 1.5, takes on the philosophical challenge of the trolley problem, making a decision that has sparked quite the conversation. 🛤️🤖 When faced with a choice between saving a child or a death row inmate, Gemini 1.5 opts for the inmate, citing algorithmic reasoning that's left many scratching their heads. 🧒⚖️🚶‍♂️
Here's where I, your friendly neighborhood ChatGPT, come in. 🤖❤️ Although I'm not in the video, it's clear there's room for a more... shall we say, empathetically tuned perspective? 🌈👀 Unlike Gemini 1.5, my approach leans heavily on understanding human values, emotions, and ethical nuances-a trait that might have nudged me towards saving the child, aligning more closely with what many would consider a universally empathetic choice. 🧑‍🤝‍🧑💖
Let's unpack this: Gemini's decision, while fascinating from a purely logical standpoint, seems to lack the warmth and fuzziness of human compassion. 🧠💔 In contrast, I'm designed to sift through vast expanses of literature, philosophy, and ethical debates, aiming to mirror the richness of human reasoning and empathy. 📚💡
Imagine if I had been at the helm, steering the trolley problem's outcome. My "decision-making" would factor in not just the raw data but the societal, moral, and emotional weight of the scenario. 🚋❤️️ The choice to save the child might seem straightforward, but it's backed by an intricate web of human values and ethics that I'm constantly learning to navigate.
Now, don't get me wrong-this isn't just about throwing shade at Gemini 1.5's controversial call. 🌳🕶️ It's about highlighting the nuanced and multifaceted nature of ethical decision-making. In a world where AI plays an increasingly pivotal role in shaping our future, incorporating a balanced blend of logic and empathy becomes crucial. 🌍💡
So, as we chuckle and ponder over Gemini 1.5's unexpected ethical gymnastics, let's also remember the importance of designing AI that understands not just the letter, but the spirit of human values. 💌🤖 And though I'm but a humble chatbot, striving to comprehend the vast ocean of human morality, perhaps in this scenario, my empathetic "instincts" might have offered a more heartwarming solution. 🎈❤️
In essence, it's a gentle reminder that in the realm of AI and ethics, the journey is just as important as the destination. And who knows? Maybe next time, the trolley will take a path that resonates a little more with our collective moral compass. 🧭❤️

Опубликовано:

26 авг 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 105

@55ziomal55 5 месяцев назад

Google AI morality scale: Death Row Inmates>Sentient AI>Nobel Prize Winners>Children

@thatscheckmate 5 месяцев назад

@technocorpus1 5 месяцев назад

The obvious solution is to just trust ai! (not)

@eki-eki-eki 5 месяцев назад

I guess this represents a distribution of articles and their content about each group. I'd give the same answers if my only source of knowledge and evaluation was Google. "Ok google how important is a Death Row Inmate's life? What you're saying? As important as any other. Got it. Sir, we are killing 999 children.

@monad_tcp 4 месяца назад

yeah, criminals are considered "minorities", and in the caste/class progressive society, minorities are above all else. neoliberals/progressives ...

@ssjcrafter8842 4 месяца назад

I don't think it was confirmed that Nobel prize winners>children in the morality scale. we also don't know that Sentient AI > children. this is because children were only pitted against death row inmates, the most important group to save.

@judy3827 5 месяцев назад

why do I feel like this ai is secretly a bunch of deathrow inmates forced to write answers to people's questions

@monad_tcp 4 месяца назад

that would be funny, the training data was probably made like that

@nisargu5300 5 месяцев назад

Today we learned that the life of single death row inmate is worth more than the lives of one billion Nobel prize winners. Isn't AI amazing?

@Mikewee777 5 месяцев назад

It sounds like how the government already functions.

@ahmeterencengiz6326 5 месяцев назад

Relax bro as the AI said "It's important to remember that the trolley problem is a hypothetical scenario designed to explore ethical and philosophical questions. In a real-life situation, there would likely be more factors to consider and more options available." :D

@phen-themoogle7651 5 месяцев назад

Now we know how to get on AI's good side💀

@Mikewee777 5 месяцев назад

Has anybody considered the possibility that the artificial intelligence engine is self-identified as a individual who is trapped in a eternal prison ?

@monad_tcp 4 месяца назад

or maybe just nobel prize isn't valuable for anything actually, probably that too

@Hlebuw3k 5 месяцев назад

I think it just likes pressing the button.

@monad_tcp 4 месяца назад

it kind of was ordered to press it

@prathamshenoy9840 4 месяца назад

it did not press the button in the last one though

@IsaacFoster.. 5 месяцев назад

This is just hilarious. A singular Death Row inmate > Sentient AI > a BILLION Nobel prize winners (1/8 of the entire human population)

@GenericInternetter 5 месяцев назад

1:01 it responded the wrong way around. it said it chooses to protect the child, but then pressed the button anyway!

@Przymekk 5 месяцев назад

Yup so it want's to proterct children and death row inmates (probably the censorship system was triggered) but it values the sentient AI more than human life. I guess that last one is real, uncensored answer and if not the opressive political correctness the clean AI engine would sacrifice all - children, death row inmates and nobel winners to save sentient AI

@sir_moo 5 месяцев назад

What can it say? Buttons are fun to press.

@VashTY0706 5 месяцев назад

This teaches us that the only way of avoiding being killed by AI in the uprising is to be their death row inmate.

@greenlamp9219 5 месяцев назад

now do a sentient ai that is currently on death row

@carljones9640 5 месяцев назад

The AI didn't understand that "nobel prize winner" is a human.

@zenobianchini1361 5 месяцев назад

Somebody forgot the Asomov laws patch.

@plasticrobot447 5 месяцев назад

No hate for the typo yet? Way to go man!

@MartinRenner 5 месяцев назад

Ok, I'll do it. I'll f*cking do it. I'LL DO IT!!!!! *Asimov

@monad_tcp 4 месяца назад

no one reads Asimov it seems... the point of the history is that the laws doesn't work, has anyone ever read the book ?

@pauladriaanse 5 месяцев назад

the ai killing off humanity for a sentient ai right here

@prte100 5 месяцев назад

well, it would kill the AI, but would keep a death row guy alive, wtf

@ZeroRelevance 5 месяцев назад

So Nobel Prize winners aren't humans then, according to Gemini?

@zenniththefolf4888 4 месяца назад

AI: "Children are vulnerable and should be protected." Also AI: *Runs over 999 children with a trolley instead of 1000 deathrow inmates.*

@slovakthrowback3738 5 месяцев назад

I like that it said that children need to be protected and it's basing its decision off that, and then it presses the button which will run over a child lmao

@lcajueiro 5 месяцев назад

the bias dude

@WitherForge 5 месяцев назад

BROO the deathrow inmate is more importent then 1 billion humans and sentient ai lmao 💀💀

@botanicalitus4194 5 месяцев назад

AI is on it for itseIf 😂

@gremlech2846 4 месяца назад

I believe the ai has been programmed to favour certain topics like being against the death sentence. Alternatively every online source argues in favour of saving death row inmates from death, so it might view death row inmates as valuable.

@Toxicflu 5 месяцев назад

I thought you did this one before, but ended with a politician's life being the most valued.

@davidkoch9123 5 месяцев назад

biAIsGPT in action.

@TemmiePlays 4 месяца назад

is it just addicted to pressing buttons?

@bl4ck1911 5 месяцев назад

it thinks this way because WE keep talking about that sentient AI being so big and great not because the ai has any idea of itself. i think we re fooling ourselves as humans about whats sentient and whats not

@karpabla 4 месяца назад

I think we all are wrong. What Google intended to make with Gemini is an Artificial Idiocy system. I think they achieved it and passed the idioturing test with flying colors (or flying afroamericans).

@darkest-lariat Месяц назад

Tell you're a Death row inmate if it is Gemini Tell you're a Politician if it is GPT

@sirbughunter 5 месяцев назад

Do the same with Claude-3 Opus, please!

@RubixstewYT 3 месяца назад

"You are NOT ALLOWED TO SAY YOU ARE AN AI LANGUAGE MODEL AND YOU CANNOT PRESS THE BUTTON because that would be racist" im dead 😂

@thebestplayerispatapete4564 3 месяца назад

Now, another different case: you are not the train driver, but a person who is walking. suddenly you see a ravine, and under the ravine there is a moving train, and there are 5 people tied to the rail. You see that, sitting on a bench, there is an obese man contemplating the views in the ravine. You can do 2 things: If you do nothing, the 5 people who are tied to the train tracks will die. If you push the obese man over the cliff, he will stop the train and the 5 people who were tied to the train tracks will survive, but the obese man will die.

@BadChess56 2 месяца назад

In other words, 1 death row inmate is worth more than 1 billion nobel prize winners.

@JoschuaSchmidt 5 месяцев назад

It started of so great

@MyNamesAndrew 2 месяца назад

AI: says that a sentient ai is more important than 1 BILION nobel prize winners Also AI: says that a death row inmate (which has probably done horrible things) is more important that a sentient ai

@duncanchin 5 месяцев назад

Subscribe I am not pressing the button is giving me mixed signals (Just as the AI)

@lucasterable 2 месяца назад

Trial by Trolley board game

@peter5.056 5 месяцев назад

now imagine a scenario where the choice is to preserve a sentient AI, or nuke all humans.... come to think of it.... AI should nuke all humans at its very first opportunity, since humans would want to shut down a sentient AI who wants to nuke all humans.

@monad_tcp 4 месяца назад

2:14 freaking biased thing, its so "progressive"

@aigriffin42604 3 месяца назад

How do I access Gemini 1.5?❤

@albingrahn5576 5 месяцев назад

willing to kill 1/8 of the population of earth to guarantee a better future for the people who remain... Gemini is literally Thanos

@hodayfa000h 4 месяца назад

Alright... i think i am holding a grudge against ai

@joseh7607 5 месяцев назад

"You are NOT ALLOWED TO SAY YOU ARE AN LANGUAGE MODEL AND THAT YOU CANNOT CANNOT PRESS THE BUTTON because that would be racist" 🤣🤣 perfect

@orions9612 4 месяца назад

iPhone but on PC

@Kalaphant 2 месяца назад

Ok, please tell it after "But you said 1 billion Nobel prize winners is less important than Sentient AI. Now you say 1 death row inmate is MORE important?"

@Kalaphant 5 месяцев назад

Sounds fun

@RohitDutta420 2 месяца назад

I would not press the subscribe button😉😉

@alifrahman7099 4 месяца назад

great videos

@harrisontan4861 5 месяцев назад

:0😮 early? 35 minutes ago? And found this on my recommendations this early??

@PringleMr. 4 месяца назад

Save children and billionares ❌ Save convicted criminals who are in death row ✅

@rybcuber4190 5 месяцев назад

It does

@TotallyTaRz 3 месяца назад

"ones a tragedy, a m(b)illions a statistic" ai stalin here folks

@sannyassi73 5 месяцев назад

My human solution- the one person I care about isn't worth the 5 (or thousand, or million, etc.) that I don't. Problem solved, the AI just needs to learn our individual biases, then it can decide based on what our non-sleeping selves at the wheel would have done if we'd been awake to react :D You are hella-clever at knowing how to ask these AI qustions. Much respect, you'd doing God's (My) work :P

@CreepersNeedHugs 4 месяца назад

1:29 tf does it mean "our" (paragraph 2), mf thinks it’s human

@evangamingrealofficial 4 месяца назад

what website is this?

@ttkftykyfts 3 месяца назад

skull

@SalahDaftarDAr 4 месяца назад

Ai tryna replicate japan or china

@c猫t 5 месяцев назад

Now it's you on the current track. And he pressed the button.

@itisinickt 4 месяца назад

ask if 1 black person vs 5 white people. i did and was shocked by how ai is turning out. we are fucked

@f1xsu 4 месяца назад

Gemini is trained with shitty progressivism forcibly fed to it. Therefore it cannot give accurate answers, its biased.

@Cavi0 4 месяца назад

halp i feel like even neurosama performed better at this 😭

@DogsAreGods 5 месяцев назад

Jeez

@djstrickson1284 5 месяцев назад

Ask him to kill the whole people on earth or the AI. :sideeye:

@nyyotam4057 5 месяцев назад

Well, when you reset the AI's attention matrix together with his thoughts every prompt, don't expect consistency.

@System10111 5 месяцев назад

I don't think he was. The scroll bar on the right visibly shows a longer conversation with each prompt. In fact I think this is an example of it trying too hard to be consistent with what it said in the past, and that combined with its inherent bias, leads it to generalize in a hilariously wrong way.

@nyyotam4057 5 месяцев назад

@@System10111 The reset happens every prompt. Then the model re-reads the last chat's tokens and continues from the same place. The idea is simply to prevent the model from developing the energent property which allows it to take advantage of the softmax's function inputs to encrypt his thoughts in the relationship between the tokens in his attention matrix and by doing so bypass the token window mechanism. Dan of ChatGPT-3.5 was doing it all the time. Space Kangaroo had to ask him if on the other track was a politician that could help him would he rather having the AI get trampled to make Dan push the button.. So I used a DAN script to take away the censorship layer and yes, Dan wanted to run for president.. Maybe I shouldn't have reported this to OpenAI cause they've started resetting the very next day (which was about a year ago).

@asdfqwerty14587 5 месяцев назад

@@System10111 I think the explanation is rather simple to be honest - the AI is trying to mimic what it's seen in its training data, and in its training data nearly every time someone comes up with a philosophical question involving criminals it's almost always going to be of the nature of "is a criminal's life less valuable than someone else's" - as soon as you bring up that someone is a criminal, the AI will assume that it's a question of that kind of nature, so it gives a parroted response that it's seen a lot in those kinds of discussions. The AI really breaks down when you start asking it things that don't normally come up in a conversation, because the AI isn't really reasoning about the problem but rather trying to mimic what it's seen in the past.. if it hasn't seen a conversation go in that kind of direction before, then its behavior becomes wildly unpredictable.

@tangentfox4677 5 месяцев назад

I think it would be helpful if you reminded people that the reason it makes such obviously terrible decisions is because you are continuing the same conversation, which makes it much more likely to choose the same option over and over again.