Sure, that's your definition. You and I don't get to define what is appropriate for others.

Who are we to say what is an "appropriate" or "inappropriate" use of AI? Appropriate is in the eye of the beholder. If somebody finds use for AI in a way that others find "inappropriate" then for them, it is appropriate.

Having recently read the history of IBM in the form of a biography of the T. J. and Thomas Watson, I disagree with your characterization that early computers were not successful. True, they weren't widespread because of cost, but they were absolutely successful, from ENIAC on. IBM didn't invent them computer, they just made it widely available.

AI has already gone through these early struggles. ChatGPT stands on the shoulders of many earlier, less-available incarnations of AI. We used to call them "neural networks." And even then, there were some useful things done with them. OpenAI did not invent AI, they just made it widely available.

No, I didn't say that.

Here's what I *am* saying. Just as:
- Farming without machinery is outdated
- Construction without power tools is outdated
- Accounting without computers is outdated

Now, with AI
- Hand-coding without AI assistance is outdated
- Research without AI assistance is outdated
- Medical diagnostics without AI assistance is outdated
- Composition without AI assistance is outdated

It's true that some of these statements are more in the "becoming true" category, but all are either true or quickly becoming true.

Crypto does have a cult following, indeed. AI has been adopted by just about everybody. Old people, young people, leaders, followers, doers, thinkers, artists, engineers, you name it. Crypto never, ever had that kind of uptake.

What kind of real-world problems does AI solve? There are too many to count. But here are a few.
- It's making software cheaper to produce. Coding assistants have come a *LONG* way in the past few months. They're no longer unreliable code-completers, they can now write entire swaths of end-to-end code. No, you blindly, implicitly trust it. But it does speed up the process by an order of magnitude. I have experienced this not only at work, but in my own hobby projects.
- It's beginning to improve medical diagnostics. In recent tests, AI has outperformed human doctors in terms of accuracy of reading diagnostic images. https://erictopol.substack.com...
- It greatly speeds up research. Combing through stacks of documents is laborious. With AI, you can ask for it to highlight specific themes, or help you find specific kinds of statements that you are looking for.
- It can greatly help with idea generation.

There are a ton more. Go ahead, shake your fist at the sky and curse the sunrise. But the sun is coming up regardless. My advice: wake up and find ways to leverage this new tool. Do that, or find yourself one of those who gets displaced because they are hanging onto their old, outdated skills.

This assumption is quickly becoming a thing of the past. In many fields, the output of AI has become more trustworthy than that of humans. Medical diagnostics is one example. https://erictopol.substack.com... Also, today's coding agents often built better code than I would. All I have to do is scan it and test it, I've come to realize that it generally gets the details right.

We humans are used to machines that do tasks better than we do. That's why we have machines in the first place.

That might have been largely true 3 years ago, but people are quickly figuring out how to use AI. Sure, there are still many who don't, and many who have unreasonable or unfounded expectations of what it will or can do. But those who get things done, are getting things done with AI.

Also, you don't *have* to know how to use it, to get value out of it. AI is such a versatile tool, it can take an improperly formed request, and do something logical with it.

Crypto solved no real-world problems.

AI solves a million real-world problems.

The two aren't even in the same category, other than that they both use GPUs a lot.

I presume you are referring to Azure. I'm not sure what missing logs your are referring to. Windows servers keep _tons_ of logs, as does Azure itself. Could a hacker break in without leaving a log trace? Sure. But it would be a lot harder than doing so in an on-premise server cluster.

And here I thought chatbots couldn't be understood or controlled. Hmmm...so people have to make the software do what it does after all???

And many sysadmins think that their on-premise servers are immune from hacking. In my experience, self-hosted servers are much less protected than cloud servers. This is because most companies that self-host, refuse to spend the money required to implement real security measures.

McAfee has been getting scammier and scammier. Their ads make it look like you are in fact infected with a virus, and you need to sign up with McAfee immediately to take care of the problem. Uninstalling McAfee doesn't make the ads go away, and this story finally explains why.

Ransomware is indeed an entirely different problem. In this context, the worry of the author was that AI might accidentally delete things it shouldn't. AI is highly unlikely to act like ransomware, so ordinary backups are likely sufficient.

I'm a fan of cloud backup systems like OneDrive or CrashPlan. These automatically keep multiple versions of all backed up files for a period of time, generally for at least 30 days. This should be enough time to notice a ransomware attack and remediate.

Most people and businesses who suffer severe damage from ransomware, have failed to take even basic precautions against loss. It's true, there are extreme scenarios that can cause loss even with best-practice precautions. But these are rare and difficult to avoid. They certainly aren't caused by AI agents run amok.

And by the way, your fancy git commands are inherently dangerous. Do you really *have* to squash those intermediate commits? It's better IMO to work on features in a branch, then when you merge your PR, squash on merge. This pattern eliminates the need to perform risky commit removals.

Agreed, the real issue is lack of backups.

I've had software that is not AI, and humans that are not AI, delete very important folders. The solution is the same, regardless of the source of accidental deletions: back up everything that you don't want to lose.

Don't these programmers use Git?

Git has this nifty little feature: it can undelete files that were accidentally deleted.

Are these programmers skipping Git too? Are they letting AI commit and push changes without even looking at them? If they are, that's stupid humans, not bad AI.

Really??

Every AI coding assistant can delete files, and they do so frequently. The question is whether they do so inappropriately.

GitHub Copilot, my personal favorite, deleted a file in my project just yesterday. Yes, I wanted it to do that, it was a duplicate file. But I don't know what it even means to say that OpenAI "acknowledged" that file may "accidentally" be deleted.

There's a solution to this: git. If a file is deleted but shouldn't be, git easily and effortlessly restores it. What's the problem here exactly? Are people letting AI commit and push changes without even looking at them? If they are, that's a stupid human, not a bad AI.