This GitHub Copilot Pull Request story is following a predictable path: skeptics latch onto it as concrete proof that we have hit a fundamental limit on AI capability. But they do this every three months with every breakthrough and have been taking Ls every single time since 2021.

I am consistently surprised that there is a cohort of software engineers whose stance is, “this isn’t perfect and so there is no utility here,” which does not match my experience with any software in my entire life.

The only people who care about whether LLMs are thinking, or are capable of understanding, or are sentient, or are hallucinating, or have intelligence, are maximalists and skeptics. I cannot overstate how little I care! I will push this software hard with reckless abandon to solve problems using computers just like I have with all other software for my entire life. I don't pretend that I have been dealing with lossless knowledge transfer and perfect utility until now.

@kyle For people who have worked hard all their lives to be accurate, to speak the truth, to be thoroughly competent, it's tough to accept that now society and especially business are saying that competency, truth and accuracy don't matter if something can be done faster and cheaper.

The excellence they worked for and prided themselves on was clearly worthless to society the whole time, society that seems just to have been waiting for something cheaper and quicker.

It's self-seering.

@kyle And also disillusioning. Reminds me of my teen years when I had been being as smart and knowledgeable as I could be and was cultivating what was then considered female virtue and then I slowly realized that what all the boys, even the smartest ones, wanted in a girl was someone slutty and sexy. Minds did not matter at all. Not only was I crushed that my efforts were not valued but I was crushed too to learn that all the guys were morons.
@kyle *from the perspective of me as a teenaged girl