Donald Trump raises new doubt regarding the AI.
With a new presidential order, Donald Trump wants to stop what he calls “left-wing” or “woke” AI. Researchers warn that the politicization of technology may be difficult to slow down. But is AI really biased?
To call AI in general biased is based on the GI-GO (garbage In- garbage out) principles. It can therefore be understood that a creator of AI can have a different opinion than the competition and therefore control the information from the beginning (GI)
“You could see it as these language models writing a kind of very broad consensus about what we think or have thought in society – in the English-speaking society at least – in recent decades.
If diversity is presented as something positive, it is not a result of the systems being ideologically brainwashed, who points out that these systems are based on enormous amounts of text from society, including “credible” research.
“Ideological dogmas”
Earlier this week, Donald Trump issued a high-profile presidential order demanding an end to what he calls “woke AI”. Among other things, federal agencies will be forced to choose AI models that do not “manipulate their responses to support ideological dogmas such as diversity, equity, and inclusion (DEI). The initiative is part of Trump’s strategy to give the United States a leading role in AI development.
Diversity where there was no
What initially put “woke AI” on the Republican agenda was a high-profile event in which an AI model on its own initiative turned an image of the founding father of the United States into a group of men with great ethnic diversity. This can be an example of when AI tries to compensate for one-sidedness in the data base – but where it fails.
The model trains on historical data and images, and if we go back a few decades, images of high-status roles were often white men. The model can therefore have distorted representations. What they try to do is to counteract such biases, but sometimes it backfires and applies compensation where it is not justified
Easily manipulated
Another example of how AI models can be manipulated – sometimes with unexpected consequences, is how X’s AI assistant Groq for a period responded to everything with ‘by the way, a white genocide is going on in South Africa’. This shows clearly how someone has made an adjustment to the language model and then the consequences are something completely different from what you imagined, (GI-GO)!
After all, there is a big future for AI!
Saving time is an important part of research today, so after the introduction of AI, science can multiply its information to more quickly reach results that would otherwise take years to reach. With this, I want to emphasize that it is still the human factor that determines the final result that AI presents.
My summary:
My opinions may not have much value, but we should go back to basic thinking, to evaluate whether AI will really be indispensable in our future, or if it would be better to let time determine our human development.
Thank you for reading my post 28-07-2025
#Windmush #curtbergsten
July 26, 2024 , Can AI, destroy your stock market? #Worldupdates Curt Bergsten/Windmush #WindmushCan AI, destroy your stock market? – Curt
April 4, 2024, AI and its negative effects in Music & War. –
February 9, 2024, Did you know that, AI may be dangerous to you? #Worldupdates Curt Bergsten/Windmush #WindmushDid you know that, AI may be dangerous to you? – Curt
October 15, 2023, AI and cryptocurrency are contributing to energy problems.
