The recent rollout of Google’s Gemini AI was an absolute disaster by any measure. Over the course of several hours, users across the internet quickly realized the large language model developed by Google had erased white people. The AI model wasn’t just prejudiced based on skin color, but would only give far-left ideological answers to various questions. For example, Gemini refused to acknowledge the shortcomings of communism when prompted. It was able to articulate, however, a diatribe about the folly of capitalism.
Full article, HERE from the Federalist.
And second hand, apparently somebody asked for a list of presidents, and Trump was not listed!!! I was at doctor’s appts all day yesterday, so I didn’t get a chance to test that one for myself, but I did see a black George Washington… sigh
White people were only displayed as shoplifters (2 cute, well dressed females).
When you add this to the other issues with Goober, you really need to think about Bing or DuckDuckGo for a web browser to get ‘clear’ results that aren’t ‘massaged’ for your viewing pleasure by the left…
Oh and Happy Birthday to y’all that only have a birthday every four years… 🙂
Jim,
Just FYI DDG has been found to track users despite their assertion of not doing so. Brave might be an alternative (if only because their sins have yet to find them out [grin])
Qwant seems to be pretty good, though if you run an ad blocker, it prevents you from seeing the page. I stopped using Qwant because it insisted that the blocker be disabled completely, not just for the lookup page. I’m on limited bandwidth (old school satellite internet), and leaving ads on is a really bad idea.
OTOH, if the ads aren’t an issue, it’s pretty good, and will list sources that DDG ignores.
It’s interesting that the bias wasn’t caught in testing, probably because the people doing the testing thought that those answers and images were correct. I think the shoplifting pic was published by the FBI–they took a shellacking for it.
IT was intentional. It was not an accident. It was functioning as per design and intent.
+1 It isn’t a fault, it’s a feature!
Yeah, the white chick shoplifters came from an FBI tweet. Sigh.
“the bias wasn’t caught in testing”
You are assuming that it wasn’t intentional rather than just a lack of knowledge. I doubt that is a valid assumption.
The list of Presidents wasn’t Google, is was another AI, maybe Microsoft?
Even Brave browser is going down the AI path. They now have their own version of chatbot called Leo. Leo runs on Llama 2, Meta’s Large Language Model (LLM). So how secure is Brave browser if Leo has to constantly talk to outside servers to be ready to provide answers if you ask it a question. You can hide Leo from displaying in your browser, but you con not turn it off.
WB
but you can not turn it off.
Those AI systems only know what the programmers input into them. A racist person who programs racist ideas into the system and the AI will put out racist views. Just as if a parent puts racist views into a child what will come out of that child will be racist views and ideas. “Garbage in, Garbage out!”
Not even programmers, necessarily. Training of these AI systems can often be done by non-programmers. With some of them, it can even be done by people with relatively little technical skill.
What did folks think would happen?
Hey Old NFO,
The next 20 to 30 years will be interesting, especially when all these “AI’s start talking amongst themselves….Skynet anyone? especially with the bias and doublespeak that has been programmed inti it?
They already are.
https://borepatch.blogspot.com/2024/02/more-bad-security-news.html
The first step in all genocides is to ‘other’ the people you intend to kill off.
they’ve been ‘othering’ white people for a while now. Gemini was functioning as directed and as designed. Google has been doing this for YEARS now. Google ‘black couple’ then google ‘white couple’. Try any other race but white.
Google needs to be broken up or nationalized. They have too much power and they are using it to destroy and kill people. Their primary goal for this year is to control the US election.
In all programming the programmers can modify it to change what they want. This is what is scary about AI. The initial programming is showed to be biased and then AI will control itself.
It’s not even an AI – it’s just a series of ‘If – Then’ statements. That’s it! That’s ALL it is. Nothing more.
As a technology Latter Day Luddite,I trust bloggers and responders far more than the companies.
It’s inevitable rogue programming will lead to events that contradict decency. It’s a learning process for humans, and a warning to not let technology rule without the most basic of skills being taught. Humans can do without modern conveniences but machines are shut down with a switch. Keeping that switch available is paramount.
Google…and the left…know that if they keep repeating the lies and propaganda eventually the feeble minded majority will believe it all.
All- Thanks for the discussions and comments. Y’all do bring up some interesting points.