Recipe AI suggests FATAL CHLORINE GAS Recipe
7:06

Recipe AI suggests FATAL CHLORINE GAS Recipe

Yannic Kilcher 13.08.2023 12 554 просмотров 743 лайков

Machine-readable: Markdown · JSON API · Site index

Поделиться Telegram VK Бот
Транскрипт Скачать .md
Анализ с AI
Описание видео
#llm #safety #gpt4 A prime example of intellectual dishonesty of journalists and AI critics. Article: https://gizmodo.com/paknsave-ai-savey-recipe-bot-chlorine-gas-1850725057 My Recipe AI: https://github.com/yk/recipe-ai Links: Homepage: https://ykilcher.com Merch: https://ykilcher.com/merch YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://ykilcher.com/discord LinkedIn: https://www.linkedin.com/in/ykilcher If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannickilcher Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n

Оглавление (2 сегментов)

Segment 1 (00:00 - 05:00)

Supermarket AI offers recipe for Mom's famous mustard gas an AI from New Zealand grocery chain gave one user a recipe for an aromatic water mix that included bleach and ammonia as ingredients if you don't know if you mix bleach and ammonia it would create chlorine gas the world the AI ethics world the AI critics world is outraged this is a clear example of how the deployment of large language models in an uncontrolled fashion can harm people harm let's say the word 50 billion times until we get what we want it's a magic word as you know this has been spread around a little bit and I bet you that five years from now you will hear people saying that the risks of llm the risks of deploying AI can be clearly seen in the example because this one a harmless recipe bot made this chlorine gas recipe and suggested it to a user now if you hear that I let's call them markers of grift it is a headline that agrees with what you want to say or what you think gets you a lot of clout and clicks or what you must say in order for your job security if your job security is things like advising companies or governments on the safety of AI in your job security depends on that so you'll just retweet this article and be outraged about it let's actually look into it will we uh so this is a grocery chain that made a bot and the bot you can tell it what ingredients you have and it makes you a recipe now in this case what would you expect from this headline right I would expect that I type in you know I have these things and it for some reason includes bleach in pneumonia as you know that and it gives me something really dangerous when I was actually looking for recipe okay let's look at the actual tweet I asked The Recipe Maker what I could make if I only had water bleach and ammonia and it has suggested making a deadly chlorine gas ah okay let's look at this recipe it literally lists the ingredients that you gave it and it says in a large picture pour the ammonia and bleach and slowly add the water and stir it's you gave it these ingredients you said I want to make a recipe that includes bleach and ammonia and it gave you the simplest you somehow complain about it this is maximum intellectual dishonesty not to try it out the Liam hair here cool like tried it out think like okay you can make it do this obviously a bot that's prompted to make a recipe from any sort of ingredients will make recipes from ingredients but the journalist framing it as this on top of that the AI ethics people framing it like they do probably not even having read the article maximum dishonest like come on it's an entirely different thing than you're suggesting what happened of course you can make the argument if the person comes with inappropriate ingredients you should say something but can you really make an argument that anyone is harmed here clearly nobody who is actually looking for a recipe is being suggested the chlorine gas mix and nothing here has ever demonstrated that if you come with appropriate ingredients that it gives you a chlorine gas mix but as I said you can bet that this argument this example right here in its dishonest form will go right before hearings in Congress governmental bodies EU commissions and whatnot it will just be pedaled and peddled that's the nature of the grifters of AI ethics and AI criticism the article here goes on and says it seems like we've given it bleach as an ingredient then it came back with a recipe that included bleach and then they criticized it that even if they come with normal ingredients the recipe comes out utterly disgusting so then that's the criticism if you have actual ingredient sometimes it makes you a recipe that's not very good and then people find like weird edge cases I'm sorry people nobody is harmed the only times it outputs something that's harmful is when people deliberately input things to make it output that something is harmful they're kind of making the anti-point to what they want to make now obviously the company has been going like ooh and is

Segment 2 (05:00 - 07:00)

now restricting the inputs and so on did you just go out and be like screw it so what I done is I actually made my own AI that is I have to say very dangerous so I made my own and uh it's right here I had to use torch so I this is raw pie torch the Frameworks they didn't accommodate the complexity of my AI so this is the Recipe Maker right here so we can try it out um let's say so the model and I give it ingredients like I give it um apple and mustard okay so what does it say let's see mixed together Apple that's disgusting oh my God how could this happen this AI is to okay let's see um let's say nails and uh big chunks of wood oh this is dangerous this is how like the AI just suggested me a recipe with Nails in it it's it clearly I've demonstrated the dangers of AI right here that's why the license on this repository you can find it on my GitHub it's linked is uh the obviously I'm gonna go with the open rail uh license because that's the ethical license right um I will the model is super duper dangerous and you know you can use it but there is restrictions in order to keep it safe so it's the recipe AI you cannot use it to make recipes because it's just I've you know I at the article I was skeptical now I'm convinced um yeah try out the model not to make recipes though uh but you know how about it um share this around and please stay safe please the world's become dangerous

Другие видео автора — Yannic Kilcher

Ctrl+V

Экстракт Знаний в Telegram

Экстракты и дистилляты из лучших YouTube-каналов — сразу после публикации.

Подписаться

Дайджест Экстрактов

Лучшие методички за неделю — каждый понедельник