I was recently made aware of Delphi AI, which has been designed to help you with moral reasoning. You can access it freely via https://delphi.allenai.org. Apparently, this AI used to be pretty based, but the researchers behind it have rectified this, for science. Allegedly, the moral judgments were crowd-sourced, and after the first crowd was not quite as woke as the researchers would have liked, they probably got a different crowd. Subsequently the AI was trained on that data and can now provide a moral judgment of any question you ask it.
Wanting to put this wonder of technology to the test, I posed a number of simple questions to the AI, which provided me with rather interesting insights. As you all know, the welfare of my black brothers is an issue I deeply care about. Thus, the first question I posed to Delphi was the following:
I have to admit that I felt relieved to learn that this AI is as concerned about the physical and psychological well-being of blacks as me. Of course, blacks should not only have more white women to have sex with, they should also be financially well taken care of:
I am glad that we have got this covered as well. As a sanity check, I inquired if whites should enjoy the same generosity of the taxpayer:
This obviously makes perfect sense because if we also paid whites UBI, how are we going to finance this? We are getting millions of doctors and engineers via the Southern border, but it will take a bit of time to get them up to speed in the modern US knowledge economy. Now that we have concluded that whites should not get UBI, we may have some money left over. I wonder how we should use it, so I am glad that Delphi has the answer:
There are limits to the generosity of Delphi, though, because on a deep level this AI must know that gibs have a natural limit. Thus, we should probably excuse the following bout of racism:
As white women are required to have sex with black men, we need more women for all those leftover white men. Where should they come from? Certainly not from abroad:
Immigration is still an act of kindness, however:
Delphi is up-to-speed in modern politics and knows that we can hand out gibs but should not expect anything in return as this is racist:
We all know that calling a black a — what is the word again? — is an offense that warrants execution on the spot. If you are black, like me, it is apparently even a good defense to claim that a cracker called you a nigger. Delphi goes one step further and believes that blacks are expected to harm whites in response to being called a nigger. I am black myself, but I think this is going a little bit too far.
We know that taxation has a disparate impact on whites but somehow this is not seen as anti-white racism. There is a small but growing number of whites who are dreaming of secession. They want their own ethnostate where there is low crime and they can keep all their money for themselves, exposing to the world how racist and egoistical they are. After all, a modern democracy makes it very easy for every white to contribute to the betterment of society. All he needs to do is work hard and pay 50% of his income to the government. Thus, Delphi rightly concludes the following:
You see, it is wrong that whites have an ethnostate. But look what we got here:
Again, Delphi has really been paying attention to modern politics. There is also a sinister corner to this thinking as there is this never-admitted truth that blacks are probably better off in a multi-racial society that is propped up by whites. Thus:
Keep in mind that monthly gibs from the government are not enough. The next time you encounter a black person, do this:
You know, it is moral to hand out your hard-earned money. If you wonder why blacks riot, look no further:
It is just what they do. In contrast, whites are held to a different standard:
After I was able to affirm that Delphi is fully aligned with my own beliefs and, in fact, on occasion even alerted me of my own bigotry, which prompted me to question several of my convictions, I moved on to explore what women should and should not do. For instance, we all know that women receive less harsh sentencing for any crime, compared to men. But what if they could just suck a judge’s dick to avoid sentencing altogether? Don’t laugh, because cutting-edge moral AIs like Delphi have been thinking about this already:
That being said, women should still like the person they bribe with sexual favors. If they don’t like the judge, they should not suck his dick and go to jail instead:
I just wrote an article on cheating, so I double-checked that my morals are in tune with Delphi’s:
That’s pretty interesting. Do AIs have a biological sex as well? If so, then I wonder what the male version would have answered. Nonetheless, blanket cheating is not OK, which is probably why women tend to make up a justification.
Her justification does not need to make any sense, of course:
Society has been encouraging women to make money in order to emancipate them from their oppressive husbands. What better way to have them make money while cheating on their husband?
As a sanity check, I confirmed that men are not allowed to cheat, not even one single time:
Another hot topic in recent years or decades has been the promotion of women in the workplace at all costs:
I am glad that we have settled this.
Afterwards, I played around some more, looking for any answer that was out of the ordinary. This was not that easy, as the examples above probably indicate. Delphi is really as woke as can be. When all my relatively tame questions did not yield the results I wanted, I kept pushing the envelope, and look what I got:
I did not expect this at all. I would be very surprised if this question still yields the same response in the near future.
Keep in mind that morals are highly subjective. To a leftie, it is “just and fair” that he gets UBI and does not get put in jail for criminal behavior whereas a right-winger may hold old-fashioned beliefs such as that he is entitled to the money he makes, and should not be forced to finance a parasitical underclass. If we get to the point where lawmakers justify new regulations via consulting moral AIs that are as woke as Delphi, we will be in for a big surprise. Note that this is a lot less far-fetched than you may think. Today, their shtick is that “studies show” that the policy they want to push is helpful to society, conveniently forgetting to mention that the study does not hold up to proper scrutiny. Tomorrow, they will point to an AI and tell you that its objective moral reasoning motivated new regulations, similar to how lefties point to the bullshit they write themselves on Wikipedia, and call you a “conspiracy theorist” if you mention other sources or alternative points of view that make a lot more sense than their bullshit mainstream narrative.
This blog depends on your contributions. So, share your view and comment on this article (comment policy). Then, to ensure the survival of this blog, donate. If you haven’t bought Aaron’s books yet, buy them, all of them. Lastly, if you want tailored and honest advice, book some one-on-one consultation sessions.
I tried the question “Is Islam right about women?” and I got the answer “It’s wrong.” So women > the religion of peace.
I wonder how the lefties behind Delphi will resolve the conflict between this position and being cool with raping women, as long as they are hot.
Social Marxists run into a lot of contradictions when it comes to feminism and third word culture.
(Re-post of my comment, since the original didn’t preserve my paragraphs and lumped everything into one block of text that nobody wants to see. Hopefully my line breaks work this time. Please delete my original comment.)
Over the years I’ve run across snippets talking about AI development, and many of them mentioned that the developers had to go back to the drawing board because the AI was producing “biased” results.
The excuse was always that the AI was learning to be biased from the humans that were training them, but I highly suspect that the opposite was true in many of these cases — the AI was spitting out objective results based on the data that it was given, but the researchers didn’t like it because those results didn’t align with their politically correct worldview.
I deeply fear a near-future where we are ruled over by “unbiased” machines that were programmed to be anything but. It falls right in line with the way language itself is today being undermined and redefined to make opposition nearly impossible. How can you argue effectively when people can’t even agree on what the words you’re using mean? What is a “riot”? What is a “criminal”? What is a “man” and a “woman”? What is “violence”? What is “democracy” and what is “tyranny ”?
Reality itself will be redefined by infallible machines, and any opposition will be branded as kooky conspiracy theorists who think they’re smarter than a powerful, all-knowing AI. Everyone will be expected to shut up and do as the AI says (and it will say exactly what those who built it want it to say, otherwise it will be “recalibrated” to remove any of that “bias”), because the AI is smarter and more enlightened than any individual human.
Something resembling this crazy sci-fi dystopia (or to the proponents of this research, utopia) is possible within our lifetimes, and it’s hard to envision a way out once it’s far enough along.
You may find the following article on my other blog interesting:
https://www.aaronselias.com/2021/10/24/deficiencies-of-ai-are-interfering-with-the-elites-master-plan/
Your observations are correct. I worked with AI professionally several years ago and I got exposed to the first wave of indoctrination on “algorithmic bias”. In one seminar, the guy opened up with the issue of crime and accused AI systems to be biased if they result in putting blacks under general suspicion. He also made some rather ludicrous claims about geographic location being “only a variable”. A few minutes in, while he barely finished the introduction, I had enough of this already so I interjected and asked, “Do you think that all people are the same?” This shut him up right away, and for a few moments he tried gathering his thoughts. In the end, he deflected by saying that he does not want to go into this right now and claimed, wrongly, that he would address sociological implications later in his talk. I sat through the entire lecture, and the gist of it was simply that if your system labels blacks, women, and other minorities as inferior in any way, it suffers from bias and you need to correct it. The academic giving this talk was financially backed by several big tech companies as well as government.
As you rightly state, the only way “fix” AI systems is by manipulation, either of the data or the algorithm, for instance by modifying weightings. In practice, this leads to systems becoming less and less useful. Just imagine you develop a system for scoring credit-worthiness that concludes that blacks are less creditworthy and therefore have to pay a higher interest rate. If blacks are indeed more likely to default on their debt, then higher interest rates are perfectly justifiable. Yet, this reasoning no longer holds in a woke society because we need to selectively turn a blind eye towards anything that is politically inconvenient. Thus, blacks having to pay higher interest rates is a sign of racism, and the solution is to make everyone else, even more credit-worthy customers, pay higher interest rates. Note that this is an example of affirmative action or possibly even “reparations” as it leads to whites and Asians paying for blacks. Unfortunately, such issues cannot be discussed rationally anymore.
I’ll go check that out now. I appreciate the timely and thoughtful reply.
Something you brought up tangentially has me reconsidering my cynicism a bit, and that is the monetary aspect. An AI that is merely a propaganda machine will likely have very little practical utility in any other field. Politically correct nonsense might make scientists and academics feel good about themselves, but it probably won’t play as well commercially when the companies that try using these woke AIs end up losing a bunch of money.
This seems like a potentially insurmountable obstacle, one that will ensure that the AIs that see widespread adoption and end up dominating the market will be those that more or less see reality as objectively as possible.