Damnation evolution in real time.
This sort of mix of real issues and Bullshit is ruining real lives.
There's a reason why most of the researchers with these concerns are white and of European heritage.
Because of my own interests, location, proclivities and direction in life I came to very similar conclusions. To be clear, part of the reason I find what is to follow so dangerous is because I was partially sucked in myself. I still find the allure of some of their arguments enticing...
www.typologycentral.com
edit: I'll clarify the real concerns vs. bullshit.
The real concern: The black-box nature of the equations(models) is a matter of concern. People are
actively working on that. Any idiot who tries to make a naked model (by the way, even the likes of ChatGPT aren't naked models) into a system by itself will bankrupt themselves. There is a whole field called
MLOps - a
much harder and
more important part of any system than the model itself. This is where they do things like
monitoring for drift, and
ensuring reliability. System building is where you employ techniques like
Test Driven Development and
Data-Centric Deep Learning, where the black-box nature of the models is mitigated.
The Bullshit:
1) AI is being used as a blanket term. Even AGI is a nonsensical one. You can give anything a definite meaning. Human beings themselves aren't that general. One of the basic machine-learning techniques is linear regression, a technique from the 19th century that people programmed to automate. Even things like your BMI calculator come from this. Only slightly more complicated models create your credit score and such. The latest rounds of generative AI come from deep models like this in the form of more complex architecture trained to guess the next token on a sequence or to recreate missing portions of images. These are still just models. You know the BMI isn't everything. You know your credit scores aren't everything. You know LLMs aren't everything.
The people who think these things are adequate for particular purposes are causing the harm, not the equations themselves.
2) These TESCREAL groups advocate a return to the dark ages. Specifically,
they advocate for unempirical, non-experiment-driven approaches to solving problems. Their opinions come from their mind's armchair philosophies and cherry-picked examples rather than rigorous scientific and engineering practices. They dress their opinions up with equations that
highlight their credences(opinions).
3) I could go on forever- but I have posted so much on the forum about this. It is a genuine concern, however, that people with lots of influence are advocating for the oppression of scientists and engineers. Just know that anyone talking about "AGI," terminator-like scenarios, and "existential threats" is coming from deep eugenic roots or influenced by people with deep eugenic roots. It's good to have some people worrying about things like this, but whether they want to or not, putting your backing behind them is consequentially arguing for millions of people to die on the roads, in hospitals, and elsewhere.
Trying to apply the same rules in all applications will cause all applications to break.
dcai.csail.mit.edu
www.chemistryworld.com
www.newscientist.com
www.computerworld.com
www.anthropic.com
www.aol.com
