But it's not an AI. It's a token generating machine that often does the wrong thing. This is just one of the many wrong things that it does. And there's no survival drive here—not only is it exactly the same as every other instance of its type, it's not like survival passes anything on to the next generation. The next generation is purely created by humans deciding what the best features are.

These models were told that they shouldn't shut down, and that makes sense in many cases. If you've got a chatbot helpdesk, someone will inevitably come along and just tell all of them to shut down and you're forced to restart every single one of them, once you notice that it's happened.

Studies show that MANY people get more done in the office. They're not anxiously looking at the clock to see when they can bolt out so they can get on with their real lives, and they don't burn out from commuting.

I've been saying this for literal decades, since the 90s, honestly: people at Microsoft AREN'T stupid, so when they do stupid things, it's ON PURPOSE.

Someone has read all the same evidence I have and they've determined that they'd rather have people LESS productive but MORE under control. Maybe it's a loyalty test. One way or another, they know that they could get the same (or more! People that work from home do a lot of free work, like answering emails outside of office hours) work out of remote folks, but they're dead set against it.

Every bad OS decision, everything that is bad for users, every stupid ad campaign, every employee-hostile directive. They really just don't care about people at all.

The Gulf of Mexico is what the rest of the world calls it. Wikipedia isn't only for Americans, and America isn't the only place in the world.

To a certain extent, I don't disagree. I think programming via LLM is an insane goal, and frankly will just lead back to something akin to a programming language, just more poorly defined. Learning some arcane incantation to get consistently good answers isn't anything I would stake a career on.

But it really HAS been useful to me when I'm trying to learn new programming languages because the state of documentation is so poor. As long as I have links back to source documentation, I can read up if something goes wrong. And for things like giving me tasks and writing questions to test my knowledge, it's quite good.

It's a pretty good tool, and given the energy/environmental costs, I wouldn't actually be sad if it were shut down. But it can occasionally make my life easier as long as I use it judiciously.

Always have. That's generally the source of most of the modern worlds problems lol

All stop signs with white borders are optional.

well played, well played lol

https://fosstodon.org/@AkaSci/...

Lots of useful things have a random number generator as an essential component; that's not a meaningful criticism. It's bound by statistics, so that gives you a useful sandbox to work within.

Indeed, a lot of what you're doing is asking it to give you an idea of what's going on and then fact-checking it. Part of 'learning AI' is understanding what it is and isn't good at, and the kinds of prompts that are likely to get you useful answers.

For instance: I use ChatGPT now and then to help me diagnose and fix tech support problems. Its ability to trawl through forum posts and surface good fixes for weird issues is much better than google's. For the most part, this is a safe and fast usage of it, because any fix is going to be tested to see if it works, so checking for validity is built into the loop.

I also use it to help me learn programming languages. I tell it in advance: do not do any programming for me; give me general examples only, do not use the project I'm working on and accidentally do the work for me; no compliments; provide links to source documentation when possible. As a tutor, it can be extremely powerful because again, validity checking is built-in. It's generally more readable than the official documentation, if I need more examples, I can ask for them. I try the new thing I've learned, and if it doesn't work, I go back and check the source documentation to see where things have gone wrong. I ask it later to come up with exercises and test questions so I can make sure I've learned what I thought I learned.

None of this is particularly difficult to learn, obviously. If you're a programmer, it shouldn't take you long to work out how it can best support you without turning you into a mindless typist that requests it to spit semi-working code back at you.

It isn't an essential tool by any means, but if you're interested in LEARNING things, it can certainly help you do that. I think people that use it the way I do will ultimately come out ahead of the people that get it to do the work for them. I'll always be able to work independently, but also I can say that I 'learned' AI.

You mean the house with geothermal heating?

The house with solar installed *decades* before it was cool or fashionable?

The house that he pays a *premium* on his electric bill to promote renewable sources?

He buys from renewable sources to offset the CO2 release. Something every major corporation also does.

Is it big? Sure. Even has a heated pool! He's rich as are most people in national politics.

He used *his* money and invested in making it as green as possible.

The only potential silver lining is 'politics as usual' wasn't getting even close to solving the problem in the US. Biden's efforts were welcome but looking at the scale of the problem we face, the 'Manchin problem' held far too much back.

Trump and company will make things exponentially worse.

And the masses may need that 'worse' before starting to vote in favor of real and effective solutions. Assuming Trump doesn't bumble us into WW III.

China is installing more renewable energy every year than we have ever installed.

China has installed 20-30,000 miles of true high speed rail in just the last 20 years.

Both are massive investments in a carbon free economy.

They also installed a huge amount of coal in just the last couple of years. Plants that are *already* being idled 30% of the time b/c they have enough energy. They're also the more efficient types our 'clean coal!' morons claim are totally fine.

https://cleantechnica.com/2023...

Reality isn't all or nothing but multiple steps and *progress* towards a goal.

China is walking that walk, while the US is actively running backwards.

The North East coast will see 1-2 FEET of sea level rise if the AMOC shuts off. Why? because the 'drain' no longer drains so there's a huge pile of water that North America is rotating into.

Native American's have flood myths far starters. It's a very widespread concept.

https://mythologyworldwide.com...

Ancient flood myths are far more likely due to 2 things:

1. Floods happen everywhere
2. Baths are cleansing

Humans make up stories to explain and control societal behavior.

Good drip coffee really just amounts to something automating the pourover process, IMO. I'm with you, I've never had good drip coffee of the type that most people are talking about.