Will Of the Wizard

The Star Trek title card generator is over here:

The sitemaker (Josh Mayfield at https://bsky.app/profile/bean525.bsky.social) has asked for people to come in and kick its tires so he can work out any bugs they turn up. So do give him a hand, if you feel inclined. :)

(also, deeply amused by this one...)

NASA operations are grounded in science and truth. They continue to follow long established safety procedures knowing the cost if they deviate.

It’s been such an emotional breath of fresh air when the rest feels so suffocating.

Edit: I went and found a high res version direct from NASA instead of my screen cap.

All of us together in this shot.

--

It produces slop. Sometimes entertaining slop, sometimes marginally useful slop, but slop.

It cannot produce anything better than slop. This is, perhaps, the key point. The proponents claim otherwise, without any evidence to support them.

Some of the slop is actively dangerous. (See: Suggestions of safe mushrooms, inaccurate legal advice, and whatever the hell is going on with the character chatbots that teens are talking with.)

People who auto-produce slop, fail to learn the skills they would have developed by writing/drawing/coding/singing/etc.

Gen-AI is utterly useless as a teaching tool. It is worse than useless: it actively discourages people from developing their own skills.

It's an incredible resource drain. Exactly how bad gets debated, but the point is: In order to have AI slop not be the random 5-legged horse with green mush tree backgrounds, you need a LOT of computer power.

The only reason we have the current level of entertaining-to-useful slop (it's still slop) is venture capitalists sinking billions into a pipe dream that is never going to happen. AI will never produce insightful, entertaining fiction; it will never produce a coherent movie; it will never produce blueprints for a house or an Ugly Christmas Sweater knitting pattern. The "hallucinations" are never going away; they're part of the process. Again: Proponents claim otherwise, despite computer experts saying, no, that isn't how this can work.

Part of that "investment" includes widescale data & content scraping, some of which is definitely illegal, much of which is immoral and might be illegal, and, importantly, they are very aware that paying market value for the content they're using would tank any chance of profitability.

There might be useful potential aspects to gen-AI. (The same way there might be potentially good uses for blockchain.) Those are not the aspects the venture capitalists are sinking money into, because those are not "replace human creativity with a robot" features.

And once the pro-AI contingent has loudly declared that none of that matters and the haters are wrong because any day now, AI is going to be able to make apps that meet HIPAA security standards, or good knitting patterns based on particular types of yarn, or a chatbot that can act as a therapist without suggesting you get divorced or commit suicide, or beautiful artwork that doesn't hit the uncanny valley and definitely isn't copyright infringement, I promise bro, just wait, the NEXT version will fix this problem...

The anti-AI contingent has to fall back on "AI is stupid and it sucks and I hate it" because the people who are invested in it succeeding have blithely ignored, not refuted, every single problem with it.

The most reasonable analyses I've seen, say we've got another 2-3 years of it at most before the bubble collapses.

Because, and again, this is important:

It cannot produce better than slop.

Another cat and turtle that appreciate the Warm Lamp.

First, he said, you write the methods. You’ve been doing the methods for months, you know what goes there. You know what questions you’re trying to answer and how you went about answering them.

Now that you’ve done that, write the results. Do your analysis. Take some time to bask in having done that and chew it over.

Next: the introduction. What information do you need to set up the conclusion? Do you need to write the six worst sentences known to man as a draft conclusion so you can go back that up in the introduction? Do that.

Write that conclusion. Tie things up, you’ve done amazing. It’s looking good! Unfortunately. The hardest part is yet to come:

The abstract. An unholy melange of introduction, results, and conclusion. The thing that 90% of readers are going to stop after reading. The ultimate test of your ability to communicate your science. Thank goodness you have all this prep work done and aren’t trying to create it out of nothing! Thank goodness you didn’t try to write it first because it was the first thing on the page!

These three things are different:

* the order of the words on the page

* the order the reader will read the words

* the order you should write the words.

If you try to write the words in the order of the page or the order they are read first, there will be PAIN.

(You should of course think about the other two as you write, but that’s not the same thing.)