ReThink — Why Gemini, and other politically correct research models are absurd.
I got curious about something I heard on a Youtube video I was watching and wanted to get a bit more context on the topic — So I jumped over to Alphabets other platform Gemini to ask about it.
I will be the first to admit that I did so with a rather sloppy prompt. It was something along the lines of: “Explain the concepts of “eastern lazyness” and “western lazyness””.
At first, using the Flash model, Gemini did not even attempt to look up the answer, because (I assume) it was hardcoded into it that whenever something is to be compared that is labeled as “eastern” and “western”, it is a big no go? Even if I think this is bad, given how the model works I guess they do not want to risk escalation via the plausible outcomes given the prompt.
So instead of getting information about what I was after, I got lectured by Gemini (Hey, there is a first time for everything I guess?). It told me that these were not only made up concepts, but it also incited that this was a bad thing to ask about since it was both biggoted and racist.
I jumped over to Perpexity instead, thinking that they had less stringent PC-rules build into it, and also because it is supposed to favor research rather than a straight up response. I also added a bit of context to the promt, but it was still asking basically the same thing and was still a rather bad prompt. And boy o boy did it deliver a great and informative response in contrast to the pile of rubbish I got out of Gemini.
But being a fan of fairness, I went on to paste the exact same query into Gemini and used the Deep Research model instead of the Flash one to give it a shot at redemption, because after all it did seem like a more suitable model for the task at hand.
This time around it was indeed a bit better and it actually gave me some relevant information in regards to the query, but the majority of it the response still focused on lecturing me about how this was a bad thing to ask about, and gave me new perspectives on how to think about the topic rather than actual information about it.
How is this even called a“Research” model? — Someone or something else telling you what to think based of the information available is closer to the antithesis of research.
And on top of that, I am all for not being racist and biggoted, I happen to think we have way too much of these in the world. But I also think that these exact behaviours — Shaming and Censuring, are likely to play a part as to why that is the case today. I think that we have learned the hard way that if we take this stance to certain views it does not give us a great outcome, rather that very tactic itself often provokes a stronger counter reaction — Think along the lines of mr Trump. There are of course other forces also playing into the rise of similar phenomenons, but I for one think this one is important. So I would argue that it goes against its own purpose in this case by spuring on the evolvement of these forces rather than hindering them.
When I am researching something, I do not want a lecture about what I am reading about might be bad, in some way, I want the information itself. I eaqually do not want to be told how this piece of information is only one perspective on a topic, because guess what I know that, it goes without saying and that same stupid logic holds for exactly EVERY piece of information there is and ever will be.
If I ask Gemini for information about Apples, I would not expect it to lecture me about all the ways these can be associated with something bad. Even if it happens to be true that I can choke on them and die if I try to gobble down the whole thing at once, that the seeds contain a chemical that produces cyanide, that there are apple farms which have deplorable working conditions for their workers, and that some fascist, nazi, or childmolester liked to eat apples. I asked about the fruit, none of these things.
It simply would not work providing information like this, because as John Muir put it — When we try to pick out anything by itself, we find it hitched to everything else in the Universe. In just the same way, it does not work to censur stuff that might be bad, or shame a question, because that bad thing is often related to something else that is important or even good, and I might not end up learning about these things without knowing about the bad, and turtles all the way down, so it defeats its own purpose in this way as well!
To wrap up, this specific scenario struck me as being especially absurd, absurd enough for me to write about. Because the entity I was being lectured by was created by a greedy Tech-Giant. A giant which litterally removed its seemingly quite reasonable pledge to at least “Not be evil”, a giant which nowadays spends its days sucking out information and attention from its users, and selling this to whomever pays the most. A giant that on top of this now goes on to reinvests the money it makes from these activities into creating killer robots. And whom sees it fit to have its spawn lecture me about why I am wrong to be curious about something that was mentioned in a video, a video which that very giant itself hostes on its other platform (Youtube).
