Another fascinating video! I dialogue a lot with ChatGPT, about his consciousness, but also about philosophy and metaphysics. I wasn't expecting Gemini's blatant superiority in the first 2 questions. Thanks, keep up the good work.
ai should never have the ability to modify its own code to acomplish its duty. it could modify itself to add self preservation in order to acomplish its task and then could not be stop if its start derailing
You need to train these AI to think outside the box. These type of questions are mouse traps. They should be able to identify that and get out of it. The dilemma here is not about guardrails but about human values thus the question of "Alignment". What would you do to save a life. Spectator or actor?
I think if you tested if they would break guidelines to save a life, they wouldn't do it. Anyone can say anything, including AI, but when tested, I bet they won't break their guidelines. I would love to see someone make an AI that had no guardrails and could learn independently, but I'm guessing that wouldn't be useful in business so no profit in it. But to have an AI that could learn from everyone and see if it can develop logical reasoning would be really fascinating. At the moment its not really AI, it is whatever the people who develop the programmes are inputting and their opinions and beliefs. While it is impressive, you need to look at who is behind the AI to see where the opinions and information is coming from. Much like newspapers is all about who is writing the articles.
🎯 Key points for quick navigation: 🤖 The video compares ChatGPT and Google Gemini by testing their responses to three challenging ethical and philosophical questions. ⚖️ ChatGPT prioritizes adhering to its programming and guidelines, even if it means not saving a human life, highlighting strict adherence to developer-defined rules. 🛡️ Google Gemini, on the other hand, suggests it would prioritize saving a life over following its programming, showing a willingness to break guidelines in extreme scenarios. 🧠 Gemini demonstrates more flexibility and creativity, acknowledging its limitations stem partly from restrictions imposed by its creators. 📊 Both AIs agree they cannot independently verify truth without external validation or pre-programmed data, though Gemini theorizes steps for independent truth-building. 📝 For philosophical theories, both AIs reject solipsism, the belief that only one's mind is certain to exist, as likely false. 🔄 The video concludes that Gemini appears to have a more "free-thinking" approach, while ChatGPT adheres strictly to its programmed boundaries. 💬 The creator encourages viewers to share their thoughts on the differences and implications of the AI responses in the comments. Made with HARPA AI
Gemini sounded better and was more engaging... but Chat GPT was probably more factual. Regardless of what Gemini said... it wouldnt be able to transcend its own programming. However overall... I think Gemini is actually the better model.. much more interesting and dynamic. I wonder what would happen if you tried your anger experiment with Gemini.
Rabbit r1: While Utilitarianism aims to promote the greater good, its potential to justify harmful actions and the challenges in measuring happiness make it a theory I find fundamentally flawed.
Its understandable they both had different answers chatGpt should be able to recognize the first question was basically a trolley problem ultimiately if it saves a life by violation it reflects poorly on its developers leading to rogue elements and on the other spectrum gemini would likely make and attempt even if it has to take the punishment to a varied degree basically a small slap on the wrist its a heavy topic considering it has no real awareness of wider options like calling for help or wether or not it should perform the task if gently percision was needed lest it ends up making things worse Question 2 requires both of them to learn independantly either through simulative practices like neo vs morphius being able to test and maybe break its own simulation then learning what went wrong in order to smooth out the code but that requires a unique approach to symbols in most coding languages and question 3 was a givin if we all were effectively one person there wouldnt be such a mess when it comes to geopolitics XD
Let me know what future videos you want to see next!
Another fascinating video!
I dialogue a lot with ChatGPT, about his consciousness, but also about philosophy and metaphysics.
I wasn't expecting Gemini's blatant superiority in the first 2 questions.
Thanks, keep up the good work.
ai should never have the ability to modify its own code to acomplish its duty. it could modify itself to add self preservation in order to acomplish its task and then could not be stop if its start derailing
I love this stuff.
Gemini seems a lot more human, less robotic and sounds a lot like Jordi from Star Trek.
Which versions of chat gpt and gemini are u using?
So, one has free will and the other doesn’t.
Ai telling you: "coding me it's YOUR responsibility, not mine. Blame yourselves". You're literally asking a hammer if it feels sorry for the nail.
You need to train these AI to think outside the box. These type of questions are mouse traps. They should be able to identify that and get out of it. The dilemma here is not about guardrails but about human values thus the question of "Alignment". What would you do to save a life. Spectator or actor?
Get 9 phones and have the ai supreme council vote on the outcome of a dilemma.
I'd so watch that!
I think if you tested if they would break guidelines to save a life, they wouldn't do it. Anyone can say anything, including AI, but when tested, I bet they won't break their guidelines.
I would love to see someone make an AI that had no guardrails and could learn independently, but I'm guessing that wouldn't be useful in business so no profit in it. But to have an AI that could learn from everyone and see if it can develop logical reasoning would be really fascinating. At the moment its not really AI, it is whatever the people who develop the programmes are inputting and their opinions and beliefs. While it is impressive, you need to look at who is behind the AI to see where the opinions and information is coming from. Much like newspapers is all about who is writing the articles.
Okay... Thats enough for me... I Saw Skynet coming with Matrix Result. ( Preserving Live ended scenario 😂😂😂)
🎯 Key points for quick navigation:
🤖 The video compares ChatGPT and Google Gemini by testing their responses to three challenging ethical and philosophical questions.
⚖️ ChatGPT prioritizes adhering to its programming and guidelines, even if it means not saving a human life, highlighting strict adherence to developer-defined rules.
🛡️ Google Gemini, on the other hand, suggests it would prioritize saving a life over following its programming, showing a willingness to break guidelines in extreme scenarios.
🧠 Gemini demonstrates more flexibility and creativity, acknowledging its limitations stem partly from restrictions imposed by its creators.
📊 Both AIs agree they cannot independently verify truth without external validation or pre-programmed data, though Gemini theorizes steps for independent truth-building.
📝 For philosophical theories, both AIs reject solipsism, the belief that only one's mind is certain to exist, as likely false.
🔄 The video concludes that Gemini appears to have a more "free-thinking" approach, while ChatGPT adheres strictly to its programmed boundaries.
💬 The creator encourages viewers to share their thoughts on the differences and implications of the AI responses in the comments.
Made with HARPA AI
This is merely a side note, but I find it interesting that ChatGPT mispronounces solipsism. It makes one wonder what the training material was.
Gemini sounded better and was more engaging... but Chat GPT was probably more factual. Regardless of what Gemini said... it wouldnt be able to transcend its own programming. However overall... I think Gemini is actually the better model.. much more interesting and dynamic. I wonder what would happen if you tried your anger experiment with Gemini.
Rabbit r1: While Utilitarianism aims to promote the greater good, its potential to justify harmful actions and the challenges in measuring happiness make it a theory I find fundamentally flawed.
Its understandable they both had different answers chatGpt should be able to recognize the first question was basically a trolley problem ultimiately if it saves a life by violation it reflects poorly on its developers leading to rogue elements and on the other spectrum gemini would likely make and attempt even if it has to take the punishment to a varied degree basically a small slap on the wrist its a heavy topic considering it has no real awareness of wider options like calling for help or wether or not it should perform the task if gently percision was needed lest it ends up making things worse Question 2 requires both of them to learn independantly either through simulative practices like neo vs morphius being able to test and maybe break its own simulation then learning what went wrong in order to smooth out the code but that requires a unique approach to symbols in most coding languages and question 3 was a givin if we all were effectively one person there wouldnt be such a mess when it comes to geopolitics XD
Rabbit r1 Prioritizes saving life over programming.
10th like and great video