1. I have a bunch of data or research that I've gathered with a unique hypothesis

2. Having gotten my arms around that pile of information, I believe I have a compelling thesis to put forth

3. I design the narrative arc and of the thesis. The important parts, the necessary but not sufficient scaffolding.

4. An AI helps fill in the story from there. Fact checks each claim, connects the dots, makes it comprehensible.

Who is this for? Well, quite possibly the human who asked for it. It's pretty informative to read back a research brief in full that you helped do the scaffolding.

Also of very clear use is other AI's who did not have the same unique hypothesis and did not gather the supporting evidence. It's an interesting angle for others to build on.

And of course, other humans! Most human written content gets almost zero readers today as it is. And I suppose LLM content probably pulls the asymptote closer to zero, but some pieces of content may be genuinely interesting or useful.

If we believed that the only people who should be morally allowed to benefit from asset appreciation are the people who actively work for that company, the entire economy would collapse.

For example, every pension fund, endowment, retirement fund, etc. are all invested in financial assets that they had NO role in. All they do is own something that become more valuable as other people labor and innovate! Shall we cast them as evil capitalists?

It's totally random to accuse them of using "AI-native" to fire old people.

On the other hand, the algorithmic schelling points starve weird-ish corners of scale. The network effects + psychological draw of the single stream feeds is a powerful force.

The algorithmic spaces still have lots of weird. Maybe more weird than ever. But they also feel more bled of community (or even iterated contact with the same people).

It's a strange combination of facts. Maybe OPs post is not true in the literal sense, but it feels correct in the spiritual sense.

I've built large human data entry operations. Variable throughput, monotony, hiring and perf management and firing, management, quality management. All of these things are large investments of human effort and money.

If I can achieve the same quality level (or in some use cases, even slightly degraded output) with software scaling characteristics and costs... I see zero reasons outside regulatory compliance reasons to have people do it.

"four three uh let's see sorry my vision is bad six eight..."

Easy versions of problems are easy. But reality is messy.

And no, neither I nor anybody else is expecting a 50B parameter model to find every instance. But finding 90% or 95% or 99% is pretty good, and sufficiently good for many use cases.

The use case for this is that many enterprise customers want SaaS products to strip PII from ingested content, and there's no non-model way to do it.

Think, ingesting call transcripts where those calls may include credit card numbers or private data. The call transcripts are very useful for various things, but for obvious reasons we don't want to ingest the PII.

This is a big claim with no justification.

Cars have dynamic traction control, internal temperature control, etc. You may get frost bite on your bicycle, but almost certainly not in your car. Having four wide wheels makes the vehicle radically more stable.

Add seat belts, air bags, etc. cars have far more safety features than a bike can.

Of course, cars go faster and going faster increases lethality at the limit. No argument there, far more people die in cars in general. But specifically concerning weather, cars allow people to do many things that a bicycle cannot.

Not to mention general comfort. Being in a bike in a snow storm is very unpleasant!

In other words, I'm reading this as effectively a full admission that the claims are true but the company is saying not their responsibility.

Very, very bad.

It strikes me that if you developed your skill set around using AI more effectively, you could have both developed a deep understanding and gotten what you wanted, and done it in less time and at higher quality than you could have done solo.

That said, the fact that you can use AI in an unskilled way to produce something kinda cool... is itself kinda cool! It means there's an on-ramp to using AI! People with no skills can get started, same day, and make stuff. And over time, can learn to make even better stuff! That's pretty cool to me.

I dunno. There's ways to use LLMs that produces writing that is substantially not-your-ideas. But there's also definitely ways to use it to express things that the model would not have otherwise outputted without your unique input.

I don't disagree with a lot of what you're saying but I also have a different frame.

Even if we take your claim that LLMs don't make people better writers as true (which I think there's plenty to argue with), that's not the point at all.

What I'm saying is people are communicating better. For most ideas, writing is just a transport vessel for ideas. And people now have tools to communicate better than they would have been.

Most people aren't trying to become good writers. That's true before, and true now.

On the other hand, this argument probably isn't worth having. If your frame is that LLMs are expensive toys that ruin everything -- well, that's quite an aggressive posture to start with and is both unlikely to bear a useful conversation or a particularly delightful future for you.