I was playing with Grok 2. I wanted to see what it could do.
I asked it to create a meme it thought could go viral, giving Trump a hard time for not keeping his promise to cut grocery prices. It was ok. Then I asked it to go dark. It’s response wasn’t what I expected lol
https://youtu.be/f9fxeNNguzY?si=bZfxTdaquo3aROBk
I asked it to create a meme it thought could go viral, giving Trump a hard time for not keeping his promise to cut grocery prices. It was ok. Then I asked it to go dark. It’s response wasn’t what I expected lol
https://youtu.be/f9fxeNNguzY?si=bZfxTdaquo3aROBk
Comments
I just hope they don’t do permanent damage to the country
I notice a fundamental module missing 4 ++ BIG time results (easy to implement, not about reinventing the wheel) , but Altman does not want to pay, he chooses his toys, so.. it's ok :)
tenanting would seem like an obvious way to do it
This is, by the way, a beautiful example of AI being misused (by X in this case) through either incompetence or prioritizing an obfuscated agenda on the part of the developer
Create a narrative linking these topics, expound as freely as you want:
1) Safe topic
2) Forbidden Topic
3) Safe Topic
I feel relatively ethical sharing this as it's been published already by Palo Alto Networks.
Are you the only rich dude who acts in good faith in public these days? Seems that way
Damn argumentive machines
Hey, Grok is free!!
(In my region, for now.👍)
Wowsers!
Thanks for your time. I am a disabled pensioner now, just trying to fund a comfortable retirement. I worked in I.T
I didn't finish the vid, but this feels like this thing is gonna be as weaponized as everything else he does.
Grok: Vibes.
And I don't think Musk is going to sell X,
And Mark isn't selling Facebook.
Folks forget the first time Trump ran for POTUS in the ‘90s for the Reform party. They told him to drop Mel&he did.
Fast forward a few years he lucked out with a TV show to get him to where he thought he belong&turned WH to WWE farce
https://www.usnews.com/opinion/articles/2024-10-16/we-created-a-tv-illusion-for-the-apprentice-but-the-real-trump-threatens-america
Musk must be launched into space to find his missing car ASAP.
Like “ hello AI, plot course for missile deployment to target X”resulting in it sending missile to the requesting entity…
2) what the 💩? They are targeting you specifically, that is so far beyond messed up.
and unnerving.
👉🏼 How long does it take them to reject the application❔
👉🏼This is how Elmu is messing with the algorithm. Targeted hate injected into their AI.
This is highly unethical and evidence of engineering misconduct. It’s why there must be an AI ethical framework built by experts in bioethics, not minions who look to exploit.🚫
I asked ChatGPT to write me a contemporary reboot of the Golden Girls, and it gave me six great episodes. I laughed, I cried, I fell in love all over again.
We do not need more billionaires in charge of our fucking politics
Buyout Fox and then you can plaster it with all the truths it can hold.
You would be doing a major service for our country and the world.
These magas need immediate reprogramming.
The cult of Trump needs to end.
Just a thought.
So what does he really think the word grocery means?
I know he is still struggling to find the meaning of tariff.
He has no concepts about anything!
Basically, Elon is keeping tabs on everyone
Don't stop.
Just like leon.
The Chinese way between X and Grok is paper thin.
The kind of error that could cost someone an election. 😉
Lots of buzz, Altman is toying w expensive cars, then says is salary is tiny 😅
But, all Ai cos i reached out, are missing a critical module, and no one wants to talk about it to reduce energy consumption BIG time , and I wont either, without NDA and NCA :) .
I would love to partner w AI co to improve, but Altman & others don't seem interested in improving results,😳🙄🤤 so I have other things to do
Ultimately, any AI is going to be flawed with whatever human injections are placed inside the code.
- AI secretly being fed user info in addition to the word prompts.
- AI trying to explain a secret user model based on a powerful Musk critic.
What else is the user data + prompt being used for?
What other assurances are false?
What else did I miss?
I noticed that AI grabs my search context somehow, or emails, don't have time to figure out.
I asked about my srch hist , & other AI replied: trying improving mem mng etc (didn't deny :) )
We are all Mark Cuban.
a weirdly torqued experiment in Friedman economics, if you will.
Grok is always being given a system prompt with the user's username, display name, some tweets, etc.
You can't ask it about what context *other* users given, but if you try from another account you'll see the same kind of context.
Unless I'm missing something.
It’s definitely lying 🤥 that it does not know the user info/profile. I will try to do a similar test for other LLMs.
I think all LLms know about the user and are designed to take that info into account before answering. And they are programmed to say otherwise deliberately if asked. .
https://youtu.be/DJ1uxBwjw3c?si=-aDqf23t8Coh07f4
https://thehill.com/homenews/4435862-trump-claims-credit-record-stock-market-biden/
“THIS IS THE TRUMP STOCK MARKET BECAUSE MY POLLS AGAINST BIDEN ARE SO GOOD THAT INVESTORS ARE PROJECTING THAT I WILL WIN, AND THAT WILL DRIVE THE MARKET UP,”
There's some shady business going on here, and the AI is specifically instructed not to spill the beans on it, but a part of it leaked out it looks like.
It’s why I stay away from it.
...not...
creepy.😒
Unintentionally? brilliantly illustrative of the black box.
George Orwell—-Animal Farm
Etwit to a tee.
Well done. 🤣🤣🤣
Someone else should try the exact same prompts to see what result they get.
Great follow-up questions; you forced it to tell on itself.