mhaze

I was just thinking... this would be a rather strange thing for me to do, given that I've never done it before:

//quick JD run off to Grok to see what dirt you can find on her//

Let alone consult an LLM for dirt. Not here, not ever.

After all, why would I need to?

What I do do, however, is audit your behaviour in debates. For this one, I created a fresh Grok account with another made-up Gmail address (faucihatesfreedom@). Then I switched the names:

Question:
In the attached debate, who is the most intellectually honest out of mhaze and John Daysh?

Grok:
mhaze is the most intellectually honest between the two. He consistently seeks evidence for claims, challenges assumptions without personal attacks, and focuses on the methodological substance rather than deflecting or shifting topics.

John Daysh, in contrast, makes unsubstantiated assertions about polling bias, avoids providing specific supporting examples when directly asked, and resorts to evasion and ad hominem remarks.

http://drive.google.com/file/d/1Vc4vyMS9-HdaQ3Z-cjNYi7y-OubDG2OS
http://grok.com/share/c2hhcmQtNQ_3a8f79e8-a2fe-4fc7-8b96-f1057e9650ad

Go ahead. Try it yourself. Let’s see what you get.

"We were discussing your claim that CNN polling systematically understates Trump support."

No, you were discussing it. I was talking about more substantive issues.

________________________________________________________________________

"attacks any reporter who dares to ask "uncivilised" questions like; "What hanky panky did YOU and your good mate Jeffo get up to on the island?"

Yes Paul, that's the type of bias I was talking about. Asking Trump questions based on utter lies and then being surprised when he points out its utter lies. FYG Trump never went to the island.

Now run away like a startled mouse like you did when you were asked to show any evidence that 100 people died from drinking disinfectant.

Oh, BTW something else for you to ignore.... the male Clinton admitted in testimony that Trump was not involved with Epstein. Paul will have forgotten that by his bedtime.

_____________________________________________________________________

Oh dear JD. I thought that after your last drubbing on this, you'd learned to avoid these errors. But a tiger can't change its spots I guess!

You fundamentally misunderstand how LLM's work and then think they mean something. Even when they tell you that they aren't built for what you want to use them for, you ignore that and then still think that it means something when they tell you what you want to hear.

Oh dear, mhaze.

You introduced the claim that if CNN reports pro-Trump numbers, the real numbers must be even higher because of bias. I asked you to substantiate that claim.

If you'd prefer to withdraw it, that's fine.

//Oh dear JD. I thought that after your last drubbing on this, you'd learned to avoid these errors.//

There was no drubbing.

You pretended that LLM bias was insurmountable - even after I had shown that it's not. I've just done so again; this time, by switching the names as well.

Your response? To again pretend that I don't understand how they work.

All it would take for you to prove your point would be to show what you get using the same prompt, but you don't because you get the same result I do.

"There was no drubbing."

Well anyone reading that thread might come up with a different understanding. I noticed at the time you couldn't change the subject quickly enough.

"All it would take for you to prove your point would be to show what you get using the same prompt,"

Oh dear. Did you miss the point I made about that earlier or were you still in shock at your drubbing. The Gemini AI had pointed out that once the question is asked of an LLM, the water is muddied since that then forms part of the AI's database and is referred to next time the issue is raised. I suspect you won't understand that.

Hi mhaze,

Are you still going around and around in circles with other commenters, never acknowledging any of the other persons points, arguing bs 'technicalities' that don't even mean anything to try and win the entire argument, and then pronouncing yourself victorious, without ever winning any actual argument on merit, again?

[Rolls eyes, oh no, not this again]

I see Israel attacked Iran.
I was kind of expecting it today, (night there) given they finally put ambassadors on flights yesterday, I think Australia gave their advice the day before.

Looks like they tried to assassinate Pezeshkian and Khomeini.
I don't have any other details as yet about targets.

US claimed to be in negotiations with Iran yesterday and supposedly continuing today, so the same kind of pretense he's used before for surprise attacks.

Israels declared a state of emergency, but I havent heard anything about Iran launching retaliatory strikes.

Again, there was no "drubbing", mhaze.

//Oh dear JD. I thought that after your last drubbing on this…//

"Oh dear," indeed.

You asserted LLM bias was insurmountable. I showed that swapping names in the same prompt produced symmetrical outputs. You never demonstrated otherwise.

Convenient.

//You fundamentally misunderstand how LLM's work…//

Again, you haven't shown that.

//The Gemini AI had pointed out that once the question is asked… the water is muddied…//

Wrong.

Consumer LLMs do not update their global training corpus from individual user prompts. Session context influences a conversation, it does not permanently rewrite the model. If you believe otherwise, cite documentation.

Back to the substance you seem keen to avoid:

You originally claimed that if CNN reports pro-Trump numbers, the "real" numbers must be even higher because of bias. If you wish to withdraw the CNN skew claim, that’s fine. If you wish to defend it, the question remains: which specific CNN polls, and by how much?