FYI Having fun with GROK a useful and fun tool

[MemoriesByTheMile]

There ain't anybody going to consult AI and then go verify the information in an act of redundancy, ever.

Anytime I use AI for a real question (not just extending time wasting scrolling) I ALWAYS look at the sources to verify and the vast majority of the time I will then challenge the AI to provide a different view point to better understand what views it ignored the first time. It is very enlightening, along the lines of the “discourse” we had in the old days when we learned the opinions of others and didn’t just surround ourselves with those who agree with us.

 

[turbopilot]

Anytime I use AI for a real question (not just extending time wasting scrolling) I ALWAYS look at the sources to verify and the vast majority of the time I will then challenge the AI to provide a different view point to better understand what views it ignored the first time. It is very enlightening, along the lines of the “discourse” we had in the old days when we learned the opinions of others and didn’t just surround ourselves with those who agree with us.

Yes, I often challenge a conclusion of Grok with an actual url of a page with the opposite view. Grok will often "eat crow" and confess they missed the real answer on that page. I think this is part of training the model.

 

[Jim]

I've been playing with Grok today and this thing is actually pretty darn cool! Thanks Joe!

 

[Jim]

LOL, I spent some time today working on SEO with “Grok.” Somewhere in the middle, I uploaded a picture of my code and asked what he thought needed improvement. I’m not sure what I said to cause it, but the next few paragraphs that came back were a bizarre mix of hilarious commentary and something that sounded like a high school kid doing his best impression of a wise guy from The Sopranos. Lots of F-bombs etc.

I’d love to post it because it was just so off-the-wall, but this isn’t the forum for that kind of talk. Still, it was nothing I expected, and I’m still laughing. Of course, I’ve got a rather warped sense of humor anyway.

 

[turbopilot]

LOL, I spent some time today working on SEO with “Grok.” Somewhere in the middle, I uploaded a picture of my code and asked what he thought needed improvement. I’m not sure what I said to cause it, but the next few paragraphs that came back were a bizarre mix of hilarious commentary and something that sounded like a high school kid doing his best impression of a wise guy from The Sopranos. Lots of F-bombs etc.

I’d love to post it because it was just so off-the-wall, but this isn’t the forum for that kind of talk. Still, it was nothing I expected, and I’m still laughing. Of course, I’ve got a rather warped sense of humor anyway.

Sounds like Bad Rudy.

 

[Holdmybeer]

Caught Bill Gates a few weeks back talking about emerging AI technology and the impact it will have on the workforce. A salient example was a paralegal. An attorney can sit at his desk and an AI assistant can provide in minutes what a paralegal would have spent days and weeks researching and summarizing. He pointed out white coller jobs will be the first to be affected and as robotics progress blue collar will follow. Remote work will be the first to go.
Remember Turing's "Imitation Game". Put a human in one room, a computer in another and in a blind test ask the same question to both. If by the responses you can't distinguish between the two, there is no difference between the two. Turing was imagining Artificial Intelligence.

 

[JBright11]

I am in the last stages of implementing an AI powered solution within a multi-site warehousing and distribution environment. It is a game changer without doubt.

When AI is targeted to a relatively small dataset with specific operational characteristics the speed, consistency, and depth of it's analysis and presentation of actionable data is quite impressive.

We run a very lean operation, so I don't expect to lose much staff, only make them more efficient by automating a wide variety of analytical tasks, gathering feedback, and reliably performing PDCA activities.

 

[Joe Dutchstar]

output from questions can sometimes be bizarre. There are some customization choices to tailor responses. See below

 

[Holdmybeer]

output from questions can sometimes be bizarre. There are some customization choices to tailor responses. See below

View attachment 29403

So bizarre in fact GROK was briefly suspended this past week when it began praising Hitler and engaging in anti-semitic rants. Followed by hallucinagenic rants denying IDF genocide in Gaza. Grok is trained on X, so how did that happen ? Garbage in, garbage out.

 

[JBright11]

Are you saying that Grok is solely trained with content on X, or that X content is one source of data that it uses and is combined with a myriad of other sources?

 

[turbopilot]

Suggest you just ask Grok if you want an answer.

Grok is trained on a diverse range of data sources, not solely content from X (formerly Twitter). The foundational training for models like Grok-1 relied on publicly available data from across the internet up to around Q3 2023, without pre-training on X user data. This aligns with how most large language models are developed, drawing from vast internet corpora to build general knowledge. However, xAI has incorporated X content as one additional source for training and fine-tuning more recent Grok models, with user posts now being used by default unless individuals opt out via their X settings. Real-time information from X is also leveraged separately during interactions (e.g., for up-to-date responses via search), but this is distinct from the core training process. In summary, X data is part of a broader mix that includes other publicly available texts, proprietary datasets, and more.

 

[turbopilot]

Here is Elon's post on X partially answering the question.

 

[Holdmybeer]

They just call it 'Grok-"x"AI' because the other 25 letters in the alphabet were taken. Pure coincidence.

 

[JBright11]

Suggest you just ask Grok if you want an answer.

Grok is trained on a diverse range of data sources, not solely content from X (formerly Twitter). The foundational training for models like Grok-1 relied on publicly available data from across the internet up to around Q3 2023, without pre-training on X user data. This aligns with how most large language models are developed, drawing from vast internet corpora to build general knowledge. However, xAI has incorporated X content as one additional source for training and fine-tuning more recent Grok models, with user posts now being used by default unless individuals opt out via their X settings. Real-time information from X is also leveraged separately during interactions (e.g., for up-to-date responses via search), but this is distinct from the core training process. In summary, X data is part of a broader mix that includes other publicly available texts, proprietary datasets, and more.

Grok didn't make the statement I questioned, did it?

 

[Holdmybeer]

Suggest you just ask Grok if you want an answer.

Grok is trained on a diverse range of data sources, not solely content from X (formerly Twitter). The foundational training for models like Grok-1 relied on publicly available data from across the internet up to around Q3 2023, without pre-training on X user data. This aligns with how most large language models are developed, drawing from vast internet corpora to build general knowledge. However, xAI has incorporated X content as one additional source for training and fine-tuning more recent Grok models, with user posts now being used by default unless individuals opt out via their X settings. Real-time information from X is also leveraged separately during interactions (e.g., for up-to-date responses via search), but this is distinct from the core training process. In summary, X data is part of a broader mix that includes other publicly available texts, proprietary datasets, and more.

Too funny!