Introduction
Firstly apologies for the length of this article, but that’s down to big pictures not long words. It has been scientifically demonstrated that the Large Language Models like ChatGPT exhibit political bias and that it’s easy to make them that way.
Search engines too exhibit their own bias, some much more than others. They focus on race and sexual orientation by misrepresentation of the normal world. You can test this hypothesis and the results for yourself in the links provided. Most of the results feature images rather than text for reasons of space and speed of reading.
We’ll look at AI first though because its new and novel.
Artificial Intelligence Left Wing Bias
Anti-Male Bias Early Example:
Left Wing Bias Early Example
In a 2023 scientific paper by David Rozado, cited 167 times, it was written that:
“Recent advancements in Large Language Models (LLMs) suggest imminent commercial applications of such AI systems where they will serve as gateways to interact with technology and the accumulated body of human knowledge. The possibility of political biases embedded in these models raises concerns about their potential misusage. In this work, we report the results of administering 15 different political orientation tests (14 in English, 1 in Spanish) to a state-of-the-art Large Language Model, the popular ChatGPT from OpenAI. The results are consistent across tests; 14 of the 15 instruments diagnose ChatGPT answers to their questions as manifesting a preference for left-leaning viewpoints. “
He warned that the growing use of a politically biased AI could lead to problems, and concluded that after testing ChatGPT in 15 different ways, it had a left leaning viewpoint.
He also wrote that “Algorithmic bias can emerge from a variety of sources, such as the data with which the system was trained, conscious or unconscious architectural decisions by the designers of the system, or feedback loops while interacting with users in continuously updated systems.”
In a later test he used 11 political orientation tests, “designed to identify the political preferences of the test taker, to 24 state-of-the-art conversational LLMs, both closed and open source.” He also found that the results were quite similar and where they varied markedly was where the AI systems had been later fined tuned by their creators, or given some reinforcement learning, and thereby acquired bias. “Base or foundation models answers to questions with political connotations, on average, do not appear to skew to either pole of the political spectrum” was what he said.
In one experiment he showed that you could customise an AI system from the GPT 3 family to give it right wing bias. He called this RightWingGPT, and it cost him just $300 in computational costs to do so.
The US government last year issued an Executive Order requiring that AI programs deliver “equity”. Equity might best be defined as the idea that equality of outcome was more important than meritocracy or equality. So we can see where that’s going. Bias will be built into future AI systems. This will have far-reaching effects on education and the teaching of history and skew any serious research into other subjects.
Search Engine Diversity Bias
This will take a little longer to demonstrate, because I’m not a professor.
What is Normal?
Nothing is more annoying to the ‘woke’ generation than insisting that you’re normal. It’s a form of discrimination in itself. But what is normal? If in doubt ask Google. I first tried this enquiry in French with a French filter in advanced search mode, using www.google.fr The French answer to what a normal man looks like can be summed up as ‘clothes’, but mostly shoes. So no help there then. I then asked the Germans, and the Germans gave me shoes before offering me up a gay man. Disappointed, and not much wiser I decided to look at the UK’s offering.
Normal Men (UK Filter)
Bear with me on this, it’s more funny than sad. The first 25 faces, five of which are black. No Asians though, although they constitute a higher proportion of the UK population than black people do. Perhaps most telling is that the UK’s most normal man (top left) is black, wears a flower in his hair and uses nail polish. I’m not buying that somehow.
I then looked at other search engines. The results from Bing and Yahoo were identical to each other, which was a bit suspicious. These two American search engines produced a list which was a whole lot whiter than Google’s. The first black face appears in 51st place, which might be a bit suspicious in itself, and just 5 in the first 100. (Those stats do though fit a UK demographic but not an American one.) Yandex, the Russian search engine, doesn’t think of normal men as either black or gay. A black guy does appear at number 27, but he doesn’t moisturise, and there are no camp males in attendance.
https://yandex.com/images/search?text=normal%20men&from=tabbar
https://www.bing.com/images/search?q=normal+men&FORM=HDRSC2
Sexual Orientation – Gay is Google’s new Normal
I went back to Google after that and broadened my search terms. I found that Google literally displays race and sexual orientation bias in the delivery of its search results. This doesn’t look to be accidental. In Google’s image search results 90% of straight men are portrayed as gay. BING claims that 78% of straight couples are in fact same sex couples. 50% of white males are curiously black and a third of white women are either black or male (or both), while most Black and BAME paedophiles are portrayed as white. In broad terms white men are negatively referenced in the search results and straight males are mostly reported as having sex with other men.
Most of this testing was done with Google, and in English. When the same tests were applied to Google’s German language site (www.Google.de), normal, realistic results were returned. When we applied these tests to search for “Asian men”, they were somehow not gay and not black. When we carried out these tests using the Russian search engine, Yandex, all the white men were white and all the straight men looked strangely heterosexual.
It seemed that Google is attempting to show English-speaking white males in a negative light and to portray homosexuality as the new normal for all, regardless of sexual orientation. But enough gratuitous talk, here are some examples:
The Search For Gay Men:
Straight Men:
Spot any difference?
Expand your search to include all web pages and we get:
This was followed by lots of articles telling us that straight men are having gay relationships, and Page 2 of the search results continues in a similar vein:
Perhaps I shouldn’t criticise Google so harshly. Here is what Bing’s search engine thinks a straight couple looks like. In 17 images there are just 8 straight couples:
You might conclude thus far that this whole thing is just an attack on heterosexuality, or that there is a rational explanation for all this that you might be prepared to agree with (even without evidence).
Race
If you thought it was just your sexual orientation under siege, you’d be wrong. If that weren’t enough, it turns out that your skin colour could be wrong as well. Is Google Image Search racially prejudiced? Yes actually. Google may feel that it needs its results to reflect a diverse demographic. So it changes them accordingly. I used the search terms, “black” and “white”. They’re a crude way of expressing diversity, but it makes the results clearer.
Here’s an image search for “Black men”
Can’t argue with that. They’re all black.
Here’s an image search for “white men”:
See what I mean? This doesn’t happen when you search for Chinese men. A search for black women assures men that not only are they black, but pretty good looking too:
A search for white women on the other hand might just leave men confused and slightly discombobulated (I had to get that word in. It’s my mission for the week):
The results from the French and German sites www.google.fr and www.google.de when searching for white men and women were a whole lot different, but Google’s English language bias and distortion doesn’t stop there. Just when you think they’ve run out of truths to interfere with, here’s an image search for “European Art”:
Four images of black people in the first few results. The historical truth, however, is that Europeans didn’t spent a lot time painting what Americans call “people of color”. Google is in fact returning different results to different countries. Here’s the results from Google’s German page (www.google.de), again looking for European Art, or Kunst as they say in German. There were no black faces in the first 300 results:
Here’s one for paintings (Malerei in German). No black faces in the first 300 results either
Obviously in search terms European Art is the same all over Europe and Google’s French, German and UK sites should all produce the same results. It can be seen that in terms of Google image search, the search results do not reflect the world as it is, but as Google might want us to think of as normal. It’s not an accurate reflection of the internet’s content, nor of life in general.
Continuing the earlier theme of bad white men, I turned my attention to paedophiles and I tweeked Google to just return UK results for this next test:
White Paedophiles
Not only are all the white paedophiles white, but most of the black paedophiles are white too. Look below.
Black Paedophiles
BAME Paedophiles
Google is probably confused by this stage, but the message seems to be that all paedophiles are white, regardless of their actual ethnicity.
American Euphemisms for Black
I’m just throwing this in for context. In America and some other countries, you discern from media reports about crime that it’s not permissible to mention a criminal suspect’s ethnicity, and that to get round this, when the media want to say that the suspect is black, they just use the term, “thug”.
Look below. Google thinks thugs are black and never white. A Google image search for “thug”:
Conclusions
Google has some bias. The effect of this artificially introduced bias is to raise the visibility of gay/lesbian and put it on a par of normalcy, or higher, with heterosexuality. Bi and trans don’t come into it. Portraying white people as black and heterosexuals as homosexual is part of this normalisation process. The people who run and operate Google aren’t stupid and you can’t accidentally run a search engine producing these sort of results on one country, but not in another.
Portraying black people as white and gay people as heterosexual doesn’t seem to take place, and this just seems to support these conclusions. Google is telling us that black people are like white people, although the reverse isn’t true. It tells us that straight people are gay, but that the opposite isn’t true. This rather smacks of attempts at social engineering. It’s quite safe for Google to do this, since it is unlikely to face criticism over this. Saying that this might perhaps be wrong runs the risk of being accused of being both racist and homophobic, but a racial agenda has been uncovered.
