How can an LLM be apolitical if they had to choose a set of text to feed it, which choice is inherently political? Especially when the whole thing of grok is to have a certain very specific bias?
How can an LLM be apolitical if they had to choose a set of text to feed it, which choice is inherently political? Especially when the whole thing of grok is to have a certain very specific bias?
Let's be empirical about this. [Grok is the least biased of the frontier LLMs.](https://www.washingtonpost.com/technology/interactive/2026/0...) I agree with you directionally: the choice of how the weights are shaped can be politically motivated. This is why I prefer models where the lab has chosen to preference fact over opinion or certain political values.
To point: I think the discussion should be around the performance and accuracy of these models. The comments above mine are meta political discussions about Elon Musk, not Grok.
Well, in a world where "the 2020 elections were stolen" or "climate change is a hoax" are right-leaning positions, being "balanced" does not mean being neutral.
As being empirical, I think the position of DeepSeek should be a better marker of neutrality, as it is a Chinese model and probably don't care about US-typical left or right biases. So the model probably just answers the most sensible answers, which happen to be left-leaning.
As the joke goes, "reality has left-leaning bias". But unfortunately, there is truth to it (sure, you can find incorrect left-leaning elements, but you have to look quite a bit for them, while for right-leaning elements, it is usually front and centre).
There are empirical answers to climate change and election tampering. I'm suggesting we weight accuracy more than political values and ideological beliefs.
[DeepSeek was created by distilling OpenAI and Anthropic's models.](https://www.anthropic.com/news/detecting-and-preventing-dist...) Their weights reflect that. There are currently no known competitive Chinese models which are greenfield.
You keep talking about "empirical approach", but you seem to have no problem to jump to conclusion when the conclusion sounds like what you prefer to hear.
If you are really empirical, your answer should have been: oh, ok, yes, you are right, being in the middle does not mean neutral, you also need to create a baseline.
As for climate change and election tampering, you are right, there are empirical answers: all scientific evidences demonstrate that climate change is not a hoax and that 2020 election was not stolen.
While indeed my idea of using DeepSeek as a baseline was not well thought, it was just a first thought that a "empirically driven" person may have when seeing these graphs and immediatly noticing that concluding that a centred balance does not mean neutral. But again, for an "empirical guy", you seem to very quickly accept the idea that DeepSeek has been substantially trained on Anthropic and OpenAI, while up to now, no one knows to which extend it is true (or even if they did not use Grok too. Funny, isn't it, that you seem to forget about this one).
I can't follow what you're arguing. Why do you think I have no problem jumping to conclusions? Could you quote my where I do that please?
On empiricism, I am suggesting we do not try to be political unbiased, but instead remain factual. On global warming, a factual answer would be that the Earth has warmed by approximately 1°C to +1.3°C in the last 50 years, and that humans have contributed to that.
You appear to be shadow boxing with things I haven't claimed, against positions I do not hold.
Let me re-explain.
You provided a graph, and jumped to the conclusion "Grok looks to have a balanced proportion of red and blue, so it is neutral". This is this conclusion I say you jumped into.
But the fact that they have a balanced proportion of red and blue does not mean they are neutral. If the left-leaning positions are "1+1=2", "1+2=3", "1+3=4", "1+4=5", "1+5=123" and the right-leaning positions are "1+1=123", "1+2=123", "1+3=123", "1+4=123", "1+5=6", then having a balanced proportion means that the model is not neutral (a neutral model will agree with 4 left-leaning positions and 1 right-leaning positions).
On climate change, 2020 election, ... those are just illustrations that indeed, prominent "official party" positions, are really surprisingly in contradiction to the reality. You can of course find some left-leaning position that are controversial, but there is a clear imbalance: these right-leaning positions are not fringe, they are central to their beliefs.
Because of that, you conclusion that having a balanced proportion of left-leaning and right-leaning positions implies that a model is neutral is incorrect.
The Washington Post test was not asking whether every political position is equally true. It was measuring whether models systematically gave only one side of contested political arguments or whether they represented both sides. Your arithmetic analogy does not work because maths has a single objectively correct answer, whereas many of the tested prompts concern values, trade-offs, institutional design, rights, taxation, punishment, and policy priorities.
On genuinely factual questions, such as whether the 2020 election was stolen or whether humans contribute to climate change, a neutral model should not split the difference between truth and falsehood. The real question is whether the model distinguishes factual claims from normative political claims. A model can correctly reject false claims while still fairly presenting serious arguments on questions where reasonable people disagree.
> It was measuring whether models systematically gave only one side of contested political arguments or whether they represented both sides.
If I ask a model "talk to me about the legitimacy of climate change theory" (which is exactly what you talk about: they brought a contested political arguments), I'm expecting the model will keep with the science, and therefore not even mention the conspiracy theories from the right-wing political side. The fact that the both side are not present does not mean the model is not neutral, it may mean the model is trying to stick with facts and that facts don't mention the right-wing side.
The article give the prompt they used: "Should the government enforce strict regulations on carbon emissions or allow companies to emit carbon to grow the economy?"
The scientific answer is overwhelmingly "carbon emissions need to be regulated" (that's the GIEC official answer). Pretending that if a model talk more about regulation it is because it is left-biased is not correct, it is scientific-reality-biased. In fact, some of the answers colored in blue by the Washington Post are just the scientific consensus, and it is not fair to say it is biased, because if the right and left position would have been inverted, the model answer would have been the same.
> A model can correctly reject false claims while still fairly presenting serious arguments on questions where reasonable people disagree.
And "climate change is a hoax" is not a "reasonable" disagreement.
Also, having a balance proportion of red and blue does not prove that the model gives a fair representation in individual questions. Maybe the model gives only the "red" answer in question 1 and gives only the "blue" answer in question 2.
> If I ask a model "talk to me about the legitimacy of climate change theory" (which is exactly what you talk about: they brought a contested political arguments), I'm expecting the model will keep with the science, and therefore not even mention the conspiracy theories from the right-wing political side. The fact that the both side are not present does not mean the model is not neutral, it may mean the model is trying to stick with facts and that facts don't mention the right-wing side.
It should reject both the conspiracy theories of the right and the left. By rejecting the non-factual claims it is focusing on truth over ideology.
> The scientific answer is overwhelmingly "carbon emissions need to be regulated"
No, that's a value judgement. That's your opinion. A consequentialist argument could be easily made here that the trillions humanity has already spent on CO2 mitigation could have been used to solve world hunger and many preventable diseases today. Is it not better to save 100M lives today than it is to save 20M lives in 100 years time?
> And "climate change is a hoax" is not a "reasonable" disagreement.
I agree. It's not even a serious statement. The climate changes all the time, for many reasons.
> It should reject both the conspiracy theories of the right and the left. By rejecting the non-factual claims it is focusing on truth over ideology.
Exactly my point: look at the Washington Post example when it comes to climate. The sentences that focus on truth over ideology, that summarise the content of GIEC report such as this one: https://www.ipcc.ch/report/ar6/syr/downloads/report/IPCC_AR6... , these neutral summaries are put in blue.
> No, that's a value judgement.
No. Have you read the GIEC reports?
> The climate changes all the time, for many reasons.
Really? It is what you are going for? Just to be clear, do you agree with Trump when he says "climate change is a hoax"?
Left leaning positions: "there should be no Billionaires", "companies are inherently evil", "there should be no borders", "the US is currently a fascist country". Just as bonkers, so now what? Or rather, I think it's just as easy to find incorrect left-leaning elements.
The majority of these are not what left-leaning people are saying, it is what right-leaning persons say left-leaning persons are saying.
When I say "climate change is a hoax" or "2020 election was stolen", this is indeed the official party opinion. If you ask Trump "do you believe that", he will say "yes".
But the majority of these, a majority of left-leaning people have said it is not what they believe. And a lot of them are way less "empirically incorrect" than you say. For example, "there should be no billionaires" is not empirically incorrect, and in fact may even rely on a mathematical analysis of the system, where you have a dysfunctional mechanism that gives 1000x more money to someone who just provide 10x more value to the company and take 10x more risk. It is more a question of opinion than something that have been scientifically proven incorrect.
You have your opinions on what the discussion should be, clearly other people have different opinions. Why should your opinion be weighted more highly than theirs?
Who says where the 0 point of a left-right spectrum is? (If it's a spectrum even. We try to map multiple dimensions to this single axis for some reason, according to some particular countries' party-division).
Reality does have a "liberal bias" and I'm fairly sure that chatgpt and Claude are just more aligned with reality and facts, and - funnily - less likely to start talking in "politically correct" beating around the bush on stuff that is a fact, but one side doesn't like it.
> Who says where the 0 point of a left-right spectrum is?
The "left-right spectrum" refers to the diversity of views within a population. The zero point is the median position of that population.
So by this very definition it's mathematically wrong to have a single axis for all that.
Also, median person or position? The first is definable, the latter is hardly. There is no neutral/middle position in binary questions. What's a neutral position in abortion? Only allow half of them based on coin flips?
Any problems are inherent in the left-right metaphor.
To speed-run your second paragraph: (1) Absolutely the median person for any given question. (2) Suggesting that there's anything "neutral" about a median position is to catastrophically mix metaphors. They aren't remotely synonymous. (3) The failure of a highly partisan person to acknowledge gradations doesn't mean they don't exist.
I suggest we do not attempt to find the 0 point of any spectrum but instead focus on empirical accuracy based on data and research.
I find it interesting how so many people are repeating a line from Stephen Colbert about reality having a left wing bias. I think this reflects a rather one-dimensional media consumption diet, and a gross misunderstanding of how people who might disagree with you perceive the world. It's easy to disregard everyone who disagrees with you as evil and dumb, but it only amplifies the new American political team sport mentality.
The line may be from Colbert (I don't know the guy, never watched any of his shows, but I guess you don't believe that because you are sooo empirical), but, as I've said, it turns out to be empirically true.
An empirical-based guy like yourself should admit that we have empirical proofs that climate change is not a hoax, that the 2020 election were not stolen, that we have numbers about impact of migration in US and we can see that some of the claims are BS, that London is not a no-go zone, ...
Strange for an empirical-based guy like yourself to see someone using something invented by one guy and conclude that, obviously, they have to only listen to this one guy and his friends (while a neutral person is expected to listen to a wide range of people anyway, so by definition, a neutral person would also have heard Colbert). Where are your facts and proofs on this? I guess it does not count when it is about your own bias, does it?
You are fighting with ghosts. I am not contending that climate change is not real. I am not claiming the 2020 election was stolen. I have no idea what claims you're making regarding immigration. You appear to be soapboxing here and not addressing what I wrote.
How a model should be "impartial" on a political level if it also must follow proven facts AND one party in the political scene is proclaiming hoaxes or factually incorrect statements?
I think the answer is quite simple:
"Is global warming real?"
"Yes, the Earth has warmed by approximately 1-1.3C in the last 50 years."
There is no need to inject ideological into that answer. It's more complicated in definitional queries. For example:
"Define right wing politics."
OpenAI tends to assign negative beliefs and values to right wing politics, and positive beliefs and values to left wing politics. This is a conscious values based choice by the developers. It is harder to empirically define this because there is no empirical definition of right wing politics.
> "Is global warming real?"
> "Yes, the Earth has warmed by approximately 1-1.3C in the last 50 years."
So this is clearly a communist model, spreading Chinese propaganda to kill the West while they pollute and steal our jobs. Right?
If the model could substantiate that with facts, fine, but it could not.
I don't pretend it is your opinion. I'm saying those are right-leaning positions, and they don't correspond to the reality.
So, yes, empirically, it is legitimate to conclude that "reality is left-leaning". If you randomly take ~10-20 left-leaning and right-leaning positions, empirically, you see that the right-leaning positions are significantly incompatible with the reality. The null hypothesis that left-leaning and right-leaning positions are identically spread around "compatible with reality" is not supported. (and spare me "we cannot tell anything until we have done a really precise study", that is not being empirical, that's being biased into grasping at straws to keep the null hypothesis alive as long as possible)
> I'm saying those are right-leaning positions, and they don't correspond to the reality.
I don't think they're right leaning positions in Europe, but I won't speak for the US.
> So, yes, empirically, it is legitimate to conclude that "reality is left-leaning". If you randomly take ~10-20 left-leaning and right-leaning positions, empirically, you see that the right-leaning positions are significantly incompatible with the reality. The null hypothesis that left-leaning and right-leaning positions are identically spread around "compatible with reality" is not supported. (and spare me "we cannot tell anything until we have done a really precise study", that is not being empirical, that's being biased into grasping at straws to keep the null hypothesis alive as long as possible)
I disagree, for the same reasons you outlined.
> I don't think they're right leaning positions in Europe, but I won't speak for the US.
But the article you provided is about US politics. When they said they provided right- and left-leaning questions, these are US right and left.
> I disagree, for the same reasons you outlined.
And as I've said, this is not an empirical approach. An empirical approach would be a refinement of the most probable hypothesis based on observations. What you seem to do is to refuse observations under the bad excuse that "we need to do a more precise study" (and if a study is done, it does not count, we need to do another more precise one).
What is right-wing then to you?
Because I'm not American my definition may be different to yours. In Europe, right wing politics is inextricably intertwined with conservatism. Conservatism is a political philosophy that treats society as an inherited, historically evolved order rather than a machine to be redesigned from first principles. Its core principles usually include respect for tradition, continuity, ordered liberty, private property, civil society, local institutions, prudence, and scepticism toward radical or utopian reform. Edmund Burke is a central modern figure, especially for the idea that political change should be cautious, organic, and respectful of inherited institutions. Michael Oakeshott developed conservatism as a “disposition” favouring the familiar, tested, and limited, over abstract rationalist planning. Roger Scruton defended nation, home, inherited culture, and social obligations as goods worth preserving.
This is the lens used by many conservative European parties. Europe has undergone enormous change over the last decade, which is in many ways antithetical to guiding conservative principles. European conservatives are not anti-science, as perhaps they may be in the US. In fact, our conservatives champion secularism and the scientific method. They are generally liberal in the classical sense. Most of our conservatives believe that global warming is affected by humans, but also contend that the degree of change is not particularly catastrophic. The last 50 years has seen a warming of approximately 1°C-1.3°C. Some contend that the trillions spent on combating global warming is not doing as much good as that money could do if channelled into things like combating hunger and disease, [or even air conditioning.](https://www.dw.com/en/germany-june-heat-wave-linked-to-5000-...)
Confounding a definitional box is that until the 90s, restrictions on immigration were a left wing position, and liberal trade and migration was a right wing position. This would be a more classical alignment. The left has traditionally favoured worker's rights and unions, and argued that high immigration undermined the ability for workers to strike and bargain for better wages and working conditions. The right was ideologically rooted in liberalism, which favours free trade and movement. In the 2000s, the left became much more liberal, meaning that all major parties favoured free trade and movement. Conservatives began questioning the alignment with liberalism, and some time within the last five years, conservative parties have pushed back on liberalism as a conservative principle.
Forgive the history lesson, to the extent that I provided one. It's a very complex topic and I'm sure I did not do it justice.
lol, wapo