What do you disagree with the article on?

The answer will shock approximately nobody on Lemmy, but it’s a good, careful look at what history can show us, from a pedantic historian.

Totally. My comment was just a note that if you screw up like the alt text is humorously referring to, that’s a good tool for fixing it.

FYI if you get mojibake like in the alt text, this is a great tool for automatically fixing it:

https://github.com/rspeer/python-ftfy

You’re in good company. Steam even managed to do it for a whole bunch of people:

https://github.com/ValveSoftware/steam-for-linux/issues/3671

The way I’m imagining it, it wouldn’t be microblogging, but I’m probably not describing it well. You’d still have communities with threads, unlike Mastodon. You’d just wouldn’t have people posting “to” those communities (unless maybe you intentionally wanted to).

It’s mostly a way to get at the same thing as merged comment threads, just in a way that feels like it would have fewer edge cases to me.

IMO copying communities from Reddit as-is was a mistake long-term, but was maybe necessary short-term so that people wouldn’t be confused. If I had my druthers, I’d make a new system where communities are uniquely identified purely as !UUID@lemmy.instance (though still with a human-friendly display name). You don’t get to create a community that namesquats something like !gaming@lemmy.world. All posts would be made with hashtags like Mastodon, and then each community would just configure “Include all posts with this tag in our community”. The big issue then is who moderates tags? I think a system like Bluesky has would work well, as you mention. People can moderate tags and other people can follow their work, or not.

If that was combined with seamless account/community migration, that would solve a lot of moderation issues. If you mod a community and the admins suck, just move it to a new instance. If the mods of a particular community suck, start your own. They won’t be able to monopolize a common name, so it’s much easier to get traction.

On the long-ago internet, there were many, many different software options that supported the same protocols, and they were also a lot more configurable generally speaking

Lemmy is pretty good about that, actually. It’s interoperable with Mastodon via ActivityPub, and there’s other projects like MBin that work nicely with Lemmy.

I was also curious, here’s a good answer:

https://unix.stackexchange.com/questions/670199/how-is-dev-null-implemented

The implementation is:

static ssize_t write_null(struct file *file, const char __user *buf,
              size_t count, loff_t *ppos)
{
    return count;
}

Neat! Do you pick one instance to load comments from? I notice that this comment isn’t showing up immediately, so wondering if there’s federation delay or the like.

You’re both on .world, which isn’t federated with hexbear, which is the most annoying instance. They’ll brigade other communities, for example the recent thread over at https://jlai.lu/post/11504685 (view it from that instance to see the hexbear comments)

I browse all sometimes from an instance federated with hexbear and I roll my eyes quite a bit whenever I do

Not sure, sorry. I don’t really use Mastodon all that much, maybe somebody else knows?

Neat, thanks!

It would be nice if communities that are similar enough could “share” a comment thread, so you don’t end up with comments scattered over many different communities for the same link. The mods could toggle something in the settings and say “This other community is good and we’ll be OK sharing posts with them”. You also wouldn’t have to explicitly crosspost.

Some apps will collapse those into a single post, but not all of them, and not all the time. It would be nice if that were better.

It would be nice if there was a way to handle instance/user migrations. If an instance gets their domain name taken away, there’s no way AFAIK for the admin to say “Here’s our new location, with a verifiable signature”. Likewise there’s no way for a user AFAIK to move their account with a verifiable signature that the new one is still them. Ideally this could all happen automatically with signatures getting synced automatically and all that.

I’m sure it would be a lot of work and no idea if ActivityPub would get in the way, but it would give people a lot more assurance that they didn’t pick a server that will screw them over by going down.

FYI the link requires login because it’s for edit mode. Might be good to also have a “What is Ibis?” bit here, instead of requiring people to follow the link.

At any rate, looks neat! Has there been any thought given to what happens if the Conservapedia or similar people want to get onto the network? Is it instance blocking like Lemmy?

It really couldn’t decide on pixelated vs line art

The whole “it’s just autocomplete” is just a comforting mantra. A sufficiently advanced autocomplete is indistinguishable from intelligence. LLMs provably have a world model, just like humans do. They build that model by experiencing the universe via the medium of human-generated text, which is much more limited than human sensory input, but has allowed for some very surprising behavior already.

We’re not seeing diminishing returns yet, and in fact we’re going to see some interesting stuff happen as we start hooking up sensors and cameras as direct input, instead of these models building their world model indirectly through purely text. Let’s see what happens in 5 years or so before saying that there’s any diminishing returns.

Gary Marcus should be disregarded because he’s emotionally invested in The Bitter Lesson being wrong. He really wants LLMs to not be as good as they already are. He’ll find some interesting research about “here’s a limitation that we found” and turn that into “LLMS BTFO IT’S SO OVER”.

The research is interesting for helping improve LLMs, but that’s the extent of it. I would not be worried about the limitations the paper found for a number of reasons:

• There doesn’t seem to be any reason to believe that there’s a ceiling on scaling up
• LLM’s reasoning abilities improve with scale (notice that the example they use for kiwis they included the answers from o1-mini and llama3-8B, which are much smaller models with much more limited capabilities. GPT-4o got the problem correct when I tested it, without any special prompting techniques or anything)
• Techniques such as RAG and Chain of Thought help immensely on many problems
• Basic prompting techniques help, like “Make sure you evaluate the question to ignore extraneous information, and make sure it’s not a trick question”
• LLMs are smart enough to use tools. They can go “Hey, this looks like a math problem, I’ll use a calculator”, just like a human would
• There’s a lot of research happening very quickly here. For example, LLMs improve at math when you use a different tokenization method, because it changes how the model “sees” the problem

Until we hit a wall and really can’t find a way around it for several years, this sort of research falls into the “huh, interesting” territory for anybody that isn’t a researcher.

Gary Marcus is an AI crank and should be disregarded