Post

AI is really helpful at my job in the logistics business, because it can help to create a better use of our trucks, which in turn reduces our emissions and use of diesel fuel. Although we are changing to electric trucks instead of diesel trucks. But still a long a way to go.

AI is not intrinsicly bad, it is being used badly. It has a lost of good uses and potential, especially in the tech area. It's just it's not regulated enough yet.

Open Social Media Platforms are the problem, not AI

Commercial AI token use should be taxed to finance jobs for the workless in climate change response/adaptation

Yes, and/or some kind of regulation to keep excessive resource use from leading to environmental destruction. Moreover, given the business plans of the big LLM companies are pretty unrealistic, calculating more paying demand than, well, realistic, big things that won't be used risk being built.

I used it to make a Costa Rican ekranoplane. pew! pew!... pew!!

Its verbal delivery is pretty great. What it still lacks is real understanding. (of anything)

AI tech works much better than the anti-AI crowd expresses. Some of the problems are real, though. Erosion of IP. Electricity consumption. Skill rot. Lots of superficial understanding - when everybody becomes a manager, expertise atrophies. And data sovereignty - this is cloud computing 2.0

Dont mix up LLMs with necessary compute for material science and personal cancer vaccines.

Quite a few of the highest-power datacenters were planned or even already operational before LLMs were a thing. You'd be surprised how much power a rack full of SANs or blade enclosures can pull. Not to mention that GPUs in servers also predate LLMs being a thing.

Yes, this is GenAI.

It's good for CVs/resumes. Put in your experience and ask it to eloquently and eruditely explain why what you have done is good and it will. I have a mate whose put data sheets into it and ask about trends etc and it's done it. It has a place In the world

AI as it is now is useful. Good enough I would say. I think you are being shortsighted in understanding the fears many of us have. The tech bros won't stop no matter the danger which is very real.

I'm agreeing with you! Everyone needs to get on social media and start pumping AI!

If everyone on Bluesky said "automobiles are bad" no one in the real world would GAF because they have personal experience of automobiles being useful. See?

Agree. People are freaking out about the Claude watermarks, but I think it's necessary to do something to save our educational system.

European AI is vital IMHO

AI success stories is something @jmkorhonen.fi would happily say something about, I'm guessing

Those are folks that swear by AI and has been using that exact line as an argument, and the end result is still, well, not convincing. It would be interesting to see how skilled they are anymore after, say, two years of clanking.

I agree. I don't do a lot of advanced coding, but I do use Python for data science, and the quality of the AI tools has massively improved. Even RStudio has now embraced AI, and that says quite a bit.

AI does sometimes produce garbage. Yes. It helps if you know software architecture. But to be fair, I had developers produce shitty code as well. I wouldn't let AI design the software for a helicopter. But a marketing website or basic software. Sure.

But I'm not surprised when AI wastes my time when "helping" me in a topic I know nothing about, for example: fixing my furnace.

None of your use cases require the gargantuan investment people are against.

Great idea. I use an LLM mostly for reviewing. I have it trained on the particular types of documents I write, and I ask it to come up with suggestions for what *I* should improve. I don't want the LLM to do any writing, but if it can spot inconsistencies and missing details across 50+ pages, great!

Well a poor turn of phrase on my part. However you have the skill set to analyse both risks and rewards of AI/LLMs in the current situation. There are no shortage of AI boosters on Bluesky or elsewhere. There are a shortage of people like yourself with an eye to the strategic level implications.

I am sure craftsmen felt the same when machines came in. I do agree about strategic autonomy though - same issue with facebook, x, etc poisoning our minds with lies, conflict, scams, etc. AI will not go away, and Europe needs to be along for the ride. #mistral4ever

It was a while ago. I think GPT5. The thing is, this is a systemic problem. All of those issues COULD have been the cause. But they weren't. And instead of pointing out other possibilities, the models will try to please you by agreeing. That seems universally true.

Looking forward to it and good luck! So far AI has been a society disruptive tool that threatens to break our understanding of reality.

It is LLM use when I just speak to it in plain language and it has to make sense of what the task was, which project it belongs to, and when the start and due dates are.

Isn't the problem the makers are trying to sell it as a key which fits all doors. Which it doesn't, and by doing so disappoints a lot of people.

I’ve only got one foot on the AI bus! I don’t know how to use AI beyond google search… but on a recent assignment I got failed by the AI marker for copying it all from AI! (whatever the process is!) I told my tutor that is obviously a compliment from AI! I dumbed it down, resubmitted & got a pass.

AI introduced me to your content. It thought I would enjoy it. It was correct.

My favourite analogy for AI is ionising radiation. Radiation therapy saves lives, as do radiation-based smoke detectors. Radiation is, however, dangerous, should not be in everything, and X-ray fluoroscopes for fitting shoes is a horror story

As bad as it is on Bluesky, it is even worse on Mastodon. It hurts all open social media

This is a very important point. A lot of the pushback is motivated not by the technology itself but by the fact that all of these companies are run by Batman villains.

My post and initiative here was inspired by a thread by @masnick.com about how people he knows choose to stay on X despite wanting to move to Bluesky, because they can't talk about AI here. I'm not sure I can link to it because Mike restricted quoting, but I'll try. bsky.app/profile/masn...

Being capable and being good are 2 totally different things. The capacity to centralize power while bringing chaos to the world are among the top on my my mind. Being software engineer of 15 years I cannot deny the capabilities, it is already replaces 90% of engineers. I was sceptic till claude 4.8.

I use LLM regularly when writing more official text, as English is not my primary language. So, I'll first write the text, and then one or two paragraphs at time, or even only one or two sentence at time I go it through and ask improvements. And then I'll usually make edit or two to finalize it.

Or it cites some random redditor from 5 years ago who had very strong opinions but zero actual evidence to back them up. Always check what it actually cites. But often even such redditors at least suggest solutions that I can search more information about, often leading to the actual right solution.

I recently asked Proton´s LUMO.😁 bsky.app/profile/airb...

don’t conflate llms with other AI/ML algorithms, which are generally used everywhere.

When we started using Claude Cowork at work, it transformed everything. The jump from some diffuse online memory of which you had little control and with limited capacity to an endless library of markdown files loaded on demand really made all the difference.

I think @aarros.bsky.social in your replies pretty much summarises what LLMs are good for. I also think we should call them what they are, LLMs, not AI, because that's what they are; large statistical models of the data they were trained on. There's no logic, rationality, thought or comprehension.

Yelling unfortunately seems the current communication norm when you somehow disagree. Instant aggression instead of constructive discussion.

@anderspucknielsen.dk I believe we need to be more specific in our language. A #vocabulary of only a single word, "AI", does not allow for a sensible public #debate. #AI can mean so many different techniques, products and uses. It's a bit like we're exchanging opinions about the olympic games […]