I think that the missing link for the fediverse is the user interface that most users see.

This is oxymoronic given that the original Reddit looks eerily similar to Lemmy today, but it’s not just looks I’m talking about.

Moderation and usability tools, bots, blocks, filtering and spam control need to go through several iterations before we can actually grow this community.

Search is another issue, as is post deletion. Right now a post vanishes, but all the stuff hanging off it is still there. This makes for a complex user experience.

Finally, Lemmy appears to be run by developers who appear to be interested in their own issues and regularly appear to dismiss issues raised by users. This is not sustainable.

I consider myself a user of the fediverse before I’m a Lemmy or Mastodon user. We have a way to go before this settles down.

I am part of the Reddit exodus. I’m here because I have no interest in promoting or supporting the atrocious policies that now govern Reddit.

The pace here is different, but the interactions feel more measured.

Based on being online since 1990, I’m comfortable with being an “early adopter”, even though I’ve only been here for a few months and Lemmy is five years old.

Will Lemmy survive? Who knows. The horse and buggy didn’t, neither did Yahoo!, MySpace or Google+, but here we are nonetheless.

I like it here.

Also, the word you’re looking for is: “headless”, as in, “headless install”

The traditional way is to use a serial console from another device.

What will it take until people get it through their thick skulls that ChatGPT isn’t intelligent, doesn’t learn and is a tool that can only generate plausible gibberish.

Using the same tools to detect such gibberish will give you more gibberish.

Garbage in, Garbage out has been true since the difference engine, it’s just that today the garbage smells like English words, still garbage, but not knowledge, intelligence or anything like it.

The machine learning approach for building models, used to produce so called large language models like ChatGPT is also used to create weather forecasting models that are bigger, better and orders of magnitude faster than available until now.

The tools have changed life, but I’m unconvinced that it’s a suitable, sustainable or realistic way to create artificial intelligence, despite claims to the contrary.

I suspect that this will require backend support to implement. Raise an issue in the Lemmy repository.

Back propagation happens during the creation of the model, not after it’s deployed.

You used to be able to run Apple Music on Android. I used it for a while. Not sure if it still exists.

DRM is one potential reason, but not the only one.

Content is licensed under specific conditions, resolution, audio tracks, closed captions, etc. Two organisations might have licensed the same title, but not the same conditions.

You can see this clearly during the Olympics where some channels only have secondary rights, or only certain events, but only free to air, not online, etc.

Added to that are marketing and exclusively deals and in the end it’s anyone’s guess what you actually end up with.

The underlying issue with an LLM is that there is no “learning”. The model itself doesn’t dynamically change whilst it’s being used.

This article sets out a process that gives the ability to alter the model, by “dialling up” (or down) concepts. In other words, it’s changing the balance of the weight of concepts across the whole model.

Altering one concept is hardly “learning”, especially since it’s being done externally by researchers, but it’s a start.

A much larger problem is that the energy consumption is several orders of magnitude larger than that of our brain. I’m not convinced that we have enough energy to make a standalone “AI”.

What machine learning actually gave us is the ability to automatically improve a digital model of things, like weather prediction, something that took hours on a supercomputer to give you a week of forecast, now can be achieved on a laptop in minutes with a much longer range and accuracy. Machine learning made that possible.

An LLM is attempting the same thing with human language. It’s tantalising, but ultimately I think the idea applied to language to create “AI” is doomed.

As opposed to the real apps that … steal your data?

Is this not a slightly selfish action? It solves the problem for you, but doesn’t make the community better for everyone. I feel like blocking users should be reserved for issues like harassment, not spam.

This is an aspect that I had not considered. Even thinking about it now leaves me unsure of the best way forward.

Specifically, whilst it’s a valid argument that blocking the user only solves this for me, and not blocking would help me see if the issue was dealt with, I feel that leaving the user free to roam across my screen is impacting me directly and if I’m not a moderator in a community, it’s not my place to second guess their decision to leave such a user and post in place.

In other words, I’m stating to a moderator that I think that this post is spam and should be dealt with accordingly, but if you leave it alone, that’s your choice.

I moderate several communities outside of the fediverse and spam in my communities is a one-strike ban. That’s not what everyone does.

Having now thought through this again, now in more detail, I’m comfortable with blocking the user.

I’ve noticed a sharp increase in spam and I’ve been reporting each one simply as “spam”.

I then block the user

Many of these posts have dozens of down votes.

Several go back months, which I discover when a new variant turns up.

I’m unsure if what I’m doing is helping or not, and as an ICT professional, I’m not sure why this obvious spam isn’t caught earlier.

I don’t know, but I doubt that the frequency response of a mobile phone microphone is either linear or consistent across sound level.

I don’t even think you could compare two sounds with different frequencies, but I don’t know.

I suspect that calibration of any such thing would require a whole lot of infrastructure, consider for example the angle of the phone in relation to sound and the impact of holding the phone in how it affects vibration and noise damping.

You might be able to use a calibrated sound level meter and pair it via Bluetooth with your phone, but I think that’s going to be as close as you might get.

In the past I’ve tried a wired USB microphone, but the OS isn’t real-time, so the jitter was horrendous. A pi would give you a more consistent result.

Note that there is no calibration of audio hardware, so the level of usefulness of any such software would be strictly limited.

Nothing quite like different shades of the same colour to get your point across.

Perhaps you might select distinct colours instead.

Shades of colour are best used for indicating a scale, like on a heatmap for example.

You can change how long a phone rings for. Talk to her telco for both landline and mobile.

In my experience, if someone doesn’t want to answer the phone, strapping it to their arm is unlikely to make any difference and in my experience they’re more likely than not to leave it on the charger.

Long battery life and tiny battery are on opposite ends of physics. Pick your poison.

Health monitoring is unlikely to be transmitted to emergency services, except iOS fall detection.

iOS and Android are both tracking as much as they can get away with.

Remote management is likely only with devices used in corporate settings.

Yeah, I was hoping to avoid that, but it’s been heading that way for a few days now :}

Yeah, it’s already on a pi, connected to my LAN and the USB port of the CNC. The switch is on a gpio pin.

I need to automate the calibration of the three axis. In other words, tell the CNC to move a specific distance, then figure out how far it actually moved, update the number of steps per mm, rinse and repeat.

To implement this, I have a known calibrated distance, a set of three 1-2-3 blocks, so I actually need to move until the switch closes, then ask the CNC how far it thinks it moved.

I intend to run this several times because right now, doing it manually is giving me weird results and I’m trying to figure out the root cause of the error.

So, I need to move an axis, interrupt the move if the switch is closed, and keep moving until the switch is closed.

That’s several recommendations for expect. I’ll start digging. Thank you.