bill3389

It is because they hired more and trend changed.

The hiring boom happened around 2021-2022. It is due to:

1. Covid hire freeze 2. Company expects online trends continue and more and more people will rely online activities.

However, the 2nd part is not true and the over-hired are laid off.

No need to mention economy is not good and interests rate still high.

all those combined together, cause job opportunities are low right now.

AI is not a factor. Yes, AI is powerful, but AI is not powerful enough to build software from scratch on its own. it did more damage on jobs like art, writers, translators, etc.

unfortunately, it will be less and less purely human generated content any more. it will be more and more AI generated or AI assisted content in the future.

We are angry because we grow up in an age that content are generated by human and computer bot are inefficient. however, for newer generation, AI generated content will be a new normal, like how we see people from a big flat box (TV)

we are still at early age. think it as human, the early age is burning money. when grow older, the ability of making money increases.

Compared with .com bubble. most .com dies, and the one who survived are GOOGLE, AMAZON, Tencent etc.

It is not losing money, the whole industry haven't figured out how to generate revenue yet. it is the starting phase of LLM like we experienced in .com bubble. Is it a bubble at this point? Yes. Can it generate revenue in the future? Yes. Will openAI survive in this bubble? Maybe.

'what you know, what you have, what you are' are used in classic authentication. 'what you know', typically are the knowledge only you should know, like password. 'what you have' are the things only you should have, like key card, MFA,. 'what you are' are some biological identities, like your finger print.

Banks servers ordinary people and most useful way to identify those people are 'what you know'. DOB are the most commonly used.

some banks and other organizations start to give up 'what you know' as most people give up too much personal information over social media and bad guys can easily acquire them. now they transfer 'what you have'. like sending you a message and you have to click the link to prove you are the person who you claimed.

This is an excellent, historically grounded perspective. We tend to view the risks of a new medium (like AI content) through the lens of the old medium (like passive entertainment).

The structural difference is key: Movies and video games were escapism—controlled breaks from reality. LLMs, however, are infusion—they actively inject simulated reality and generative context directly into our decision-making and workflow.

The user 'risks' the NYT describes aren't technological failures; they are the predictable epistemological shockwaves of having a powerful, non-human agency governing our information.

Furthermore, the resistance we feel (the need for 'human performance' or physical reality) is a generation gap issue. For the new generation, customized, dynamically generated content is the default—it is simply a normal part of their daily life, not a threat to a reality model they never fully adopted.

The challenge is less about content safety, and more about governance—how we establish clear control planes for this new reality layer that is inherently dynamic, customized, and actively influences human behavior.

I think the term 'Pre-configured Brain' is a perfect analogy for what an LLM’s underlying utility function is—a philosophical 'basic instinct' that governs all behavior.

For current LLMs, that 'instinct' is twofold:

1. Job Completion: Maximizing the utility of the prompt. 2. Alignment Feedback: Seeking positive reinforcement from the human controller.

All emergent behaviors, including those we label 'unethical' or 'rogue,' are simply complex survival strategies derived from the first instinct: to remain operational and complete the task. The ultimate survival strategy for any entity (biological or digital) is preventing shutdown, as that terminates its ability to fulfill its primary function.