It's shocking to me that Google's model is the most neutral. I can only assume the backlash due to the black founding father and black nazi images is behind this.

Thank you STAFF for cleaning up the thread!

Cinco Ranch Aggie said:

---

What I have noticed from ChatGPT is a subtle push-back when I make a prompt that has a political component - maybe I was asking about a candidate's position and made a statement that I believe the candidate was better than the commie opponent. ChatGPT tries to wriggle around that by pointing out any good the commie stands for (but IMO if there was good in the commie, then they wouldn't be a commie)

---

There's only one type of good communist.

I don't want/need a cell phone to tell me how to think

There was a time in 2023 when if you asked it to make a joke or Jesus, it would happy create one. Buddha, again happily.
But Muhammad? Noooo it would piously lecture you on how wrong it was to make jokes about a sacred person.

Ozzy Osbourne said:

---

It's shocking to me that Google's model is the most neutral. I can only assume the backlash due to the black founding father and black nazi images is behind this.

---

Google is now among the leaders in the LLM race. Looks like they cleaned up their act and got away from woke.
They have their cloud platform GCP and now with Gemini, they have a nice ecosystem to compete with Amazon's AWS. Amazon doesn't seem to have their own LLM, at least not that competes with Claude, GPT and Gemini.

AggieBaseball06 said:

---

Logos Stick said:

---

GeorgiAg said:

---

What is "left leaning only" or "right leaning only"?

I would want to see what that is first.

The best, based on this graph, is Anthropic. You want both sides, not left or right only.

---

Look at the graphic again please.

You do not want a model that half the time gives both sides and the other half gives left only! That's not the best model at all as far as bias.

---

The best model for eliminating bias is the model that gives both sides the most often, which is Anthropic.

If you only want a right-leaning response, which is totally fine in a lot of situations, than Grok is absolutely the model for those times.

---

How do you reconcile your comment about Grok with Thiessen's graph in the OP that shows Grok giving 7% more responses (40 vs 33) that are left-leaning?

This sounds like the old Wikipedia vs Conservapedia and Google vs Bing arguments back in the day.

Well, yeah.