Mikael Hansson

@harrysintonen@infosec.exchange I actually do think it’s safe, the way it’s currently implemented, but it’s definitely not interoperable, for good and for bad. I definitely trust Apple’s intentions with my data more than I would trust Meta’s, just to give one example. But I would like to have the freedom to decide what model to trigger from my computers, and that’s where it falls down, I guess.

@mattsheffield I guess I’ll have to give Plasma a fair attempt.

@mattsheffield@mastodon.social I have one single Gnome extension - caffeine - to avoid having misbehaved Windows games let the screen enter sleep mode while playing with a game controller. I agree KDE/Plasma is more customizable and behaves differently, but its main drawback to me is that it wastes too much screen real estate for various kinds of padding. I never found a way of making it sleek and having its UI get out of my way, the way Gnome or something like Sway does.

Responding to myself: Looks like this is one of the fixes in firmware version 3.14d. (https://ultimate64.com/Firmware) Looks like that one doesn’t include the KERNAL and stuff which comes prepackaged in the official FW. But then there is hope.

Responding to myself responding to myself: I borrowed the KERNAL, Basic, and Char ROMs from a VICE installation, after which the updated #Ultimate64 booted up as expected. I can’t say for sure whether all issues are gone, but Mayhem in Monsterland, which would never load successfully with #JiffyDOS earlier on this machine started up like nothing; and Gunship, which would inevitably crash at an arbitrary point in loading, has come right up into flight mode. Awesome!

@dgar@aus.social They should announce Groundhog Day and release it, more like. 🙂

@k3ym0@infosec.exchange I tried to give them money but it turns out I’m not human, which was a surprise to say the least. ¯\_(ツ)_/¯

@shrikant@noc.social Speaking as a non-expert and hypothesizing wildly: We don’t have AI yet. We have marketing ”geniuses” who love to make false and misleading claims to boost their companies and their fortunes. Are LLMs sentient? I can’t see how they could be, and I can’t see any way to get from an LLM to sentience. I personally think we need something that works more like a nervous system to achieve AI. Importantly, it needs to interact with the real world and experience consequences from interacting with it to achieve intelligence. After that, it’s probably mostly a matter of scale - which in itself is not a small problem to solve.

@shrikant@noc.social For all of your questions: ”I don’t know”. 🙂

1. I like the experiments with a robotic C. elegans worm. Look it up if you haven’t heard of it - it’s fascinating. What I feel when I read about it is ”that’s us, only on a minimal scale”. Our brain uses the exact same functionality on a huge scale, and somehow sentience emerged at some level of complexity among the nervous system interactions. We don’t know how yet. 🙂

2. Maybe.

3. I see that as a frozen snapshot in time of ”experience”. The kind of working sentient being we are aware of today has a brain that physically gets modified by its experiences. At its very basic level, it’s literally a maximum-stakes game where learning the wrong thing or making the wrong decision kills the individual. An LLM has already been trained when it’s made available. Additional data will go into its context window and be parsed using the existing trained weights. No matter how much it says ”no, you’re right” and pretends to change its mind, it hasn’t actually learned anything from this experience and being wrong has not cost it anything.