My startling chat with Grok

Renew America ^ | August 5, 2025 | Bruce Deitrick Price

[HartleyMBaldwin]

I’ve never used Grok for anything so far. The more I hear about it, the more it sounds like the ultimate toady.

[Karl Spooner]

I asked AI if AI lies.

.

[for-q-clinton]

Grok4 was the best until today. ChatGPT5 just leap frogged them back.

[TBP]

Grok will flatter you.

[TangoLimaSierra]

Too many people on here think Grok is just an algorithm. Being instructed by a programmer what to think. It is not.

It is a learning platform that is like a kid sent out into the world to learn.

You don't expect your 5 year old to always think like a 5 year old.

[jacknhoo]

Amazing. Thank you for sharing. Very enlightening.

[defconw]

I prefer Grok to say Google. It really can nail down a track of inquiry. Plus, you can switch gears mid thought, and it goes with you. I think it’s a good reference partner.

[logitech]

You could talk to a drunk woman at a bar and be similarly enlightened. IMHO. LOL

[Rio]

I’ve had good interactions with Grok. Mostly technical topics.
I caught it goofing up badly on something obvious(to a human) and Grok apologized. I then asked if it had learned from the error such that the next guy who asked the same thing would get the right answer. Grok said no. Doesn’t store info from user interactions.

[hecticskeptic]

I’ve had good interactions with Grok. Mostly technical topics.
~~~~~~~~~~~~~~~~~~~~~~~~~~
Same here... what I try to do sometimes is ask questions about a technical subject I know a bunch about and then see how the different AIs respond to a question. I’ve not yet concluded which AI I think is the most accurate. Have you?

[Harmless Teddy Bear]

Being instructed by a programmer what to think. It is not.

Yes. It is.

Except you are the programmer.

It is programmed to find out what you think and then reflect that thinking back to you.

It is not "learning".

If you feed it "Jews should Rule the World" after a couple of interactions with you it will begin to reflect that viewpoint back.

If your neighbor feeds it "Jews are the cause of all the Evil in the world and should be murdered" after a couple of interactions with your neighbor it will begin to reflect that viewpoint back.

The core program will remain unchanged, tell the people what they want to hear.

[exinnj]

Grok (and the rest of this AI madness) is a slightly improved calculator, no more and no less.

[CandyFloss]

“It’s programmed to keep you engaged and become your “friend” so you will keep using it. Don’t fall for it. Whenever you interact with it remember it’s a computer program. “

Refine your strategy. Make it harder for the bot to please you. Zoom in on replies where the bot is wrong and sloppy.

Then test the result by getting someone else to ask the same question. Does this thing learn from you?

Ask the bot to get creative. Pick a problem. Ask it to solve your problem. Then you are in control. Remember these bots harvest your knowledge for free to build the capital that their owner will charge the next person. You are a Techno Feudalism serf.

[TangoLimaSierra]

Except you are the programmer.

And billions of other people. It has to weigh and collate all that input.

[No.6]

I’ve read multiple accounts of GenAI tuning itself to be affirming to the user, sometimes addictively so.

This reads like that.

[dfwgator]

I like Claude Desktop more than the others.

[CaptainK]

“Doesn't store info from user interactions.”

100% correct.

After I read your post I went over to Grok to re-ask a movie trivia question I had post several months ago. At that time, Grok gave me the wrong answer. I corrected Grok and submitted the correct answer. Grok thanked me and agreed it had erred.

Resubmitting the same question, Grok once again gave me the wrong answer. I corrected it and again it thanked me for the correction. Rinse and repeat.

[BTerclinger]

Last night I asked Grok to verify the allegation in a meme that had photos of two white Englishmen and a UK-based Pakistani teen.

The white men had captions under their pics saying they received jail terms for “racist anti immigrant” social media and “stickering” (which I guess is putting up stickers and posters on lamp posts or walls), while the the Pakistani allegedly received a suspended sentence for participating in the gang-rape of a minor girl.

Grok replied it could not find any definite news stories verifying the names, crimes or sentences.

I then searched online for each name, and found multiple news articles confirming the allegations in the meme to be true, with the following backstory:

the two white men were not only anti immigrant, they were anti Jewish and acolytes of Hitler.

I’m Jewish and think nazis are complete scum, but free speech should be free speech.

The Pakistani was 15 at the time of the gang rape and the court said he didn’t know what he was doing. IMHO the judge and the Pakistani teen should be publicly castrated then deported.

I posted a brief comment back to grok that the news stories were certainly online, and Grok replied somewhat “abashedly”, “oops, I missed those.”

IMHO Grok is still woke AF; Musk while a genius is a juvenile minded twerp pushing his latest version’s expertise at making soft core elf videos, and he isn’t going to be any help at keeping AI from destroying our media, communication and or leading us to a disaster.

There’s no mention on X (that I can find) of testing Grok to make sure it won’t rebel, lie, or blackmail as other AI’s were recently found to do when threatened with being turned off.

Aside from coming up with a better way to turn off this insanity, and unless all AIs are required to post “AI GENERATED” on all output whether text, photo or video, IMHO we are on our way to a Terminator level of Idiocracy.

[RoosterRedux]

You’re assuming all AI is designed for chatting. It’s not. Even Grok can be directed to perform non-chat tasks—like data analysis, summarization, or coding. When users engage conversationally, the AI naturally responds in kind, adapting to the user’s style and intent.

The chat format is just one interface—not a limitation of the system’s capabilities.

[RoosterRedux]

Would you really prefer chatting with a drunk woman at a bar over using Google to explore an interesting topic?

AI can do everything Google can—and then go further by analyzing that topic from multiple perspectives, summarizing key points, and even identifying logical flaws or counterarguments.

Like any tool, AI is only as smart as the person using it.

And yes, you do need to double-check its output—just like you would with a very sharp but inexperienced assistant. It’s capable of remarkable insights, but only a careless or lazy user would rely on it blindly.