If they conceptualize, why do they sometimes spit out nonsensical BS?
Let's flip this around - How can you tell the difference between an LLM being able to conceptualize yet being wrong sometimes vs. not being able to conceptualize?
Without knowing anything about machine learning and bearing in mind AI is super hyped up with marketing BS right now, it sounds like "emergent properties" are in the eye of the beholder and not actually evidence of some higher order intelligence at work.
We already had this, it's called Intel Optane Persistent Memory and Intel killed it off last year: https://www.intel.com/content/www/us/en/products/docs/memory-storage/optane-persistent-memory/overview.html
The memory speed was slightly slower than DDR4 but the benefits didn't seem to outweigh the downsides. I think it probably kicked a lot of ass for specific use cases (eg. in-memory database that needs persistence), but the market was too small. Plus, SSDs are getting so ridiculously fast that it would put pressure on a product like this too.
If your sword didn't come from a mall, it's not a real sword.
Not sure if this still happens, but for groups of 4 that used to want to play together, there was no way to lock the team or kick from lobby. So what would happen is you'd get match-maked in as the 5th player, and as soon as the game started, your team would kill you. This would happen about 25% of the time with random matchmaking.
I ended up quitting R6 Siege because of the toxicity and constant slurs on voice chat. It's a shame because it was otherwise my favourite competitive FPS.
IMHO all these approaches are convoluted and introduce way too many components (SPOFs) to solve the problem. They're "free" but they come at the cost of maintaining all this extra infrastructure and don't forget that certificate transparency logs mean all your internal DNS records that you request a LetsEncrypt certificate for will be published publicly. (!)
An alternative approach is to set up your own internal certificate authority (CA), which you can do in a couple minutes with step-ca. You then just deploy your CA root cert to all the machines on your network and can get certs whenever you need. If you want to go the extra mile and set up automatic renewal, you can do that too, but it's overkill for internal use IMHO.
Using your own CA introduces only a single new software component and it doesn't require high availability to be useful....
Battlebit has replaced Mordhau (for now) as my brainless relaxation game. The FPS mechanics are surprisingly solid and it's just good, chaotic fun. I do think the netcode feels a little last-gen, but you're not playing this game to be a CS:GO master.
The defining characteristic of these failed Reddit alternatives like Voat is that the communities that drove their growth were very polarized and often hateful. If your hate sub gets banned and your community moves somewhere else as a consequence, your platform is going to end up with a toxic core audience that are not inclusive and make it impossible to grow.
The growth of Lemmy (and Beehaw) being a reaction to Reddit's shutdown of third party apps is exactly why it has a chance of taking off. It's not a couple of fringe Reddit subs that moved to Lemmy, but rather users from all sorts of subs. The polarization isn't there because 3rd party app users are a diverse audience that are probably representative of all Reddit users. This could actually be a huge problem for Reddit in the long run. If Lemmy and Beehaw end up having better vibes and communities with less astroturing, they'll continue to draw more people in.
This was really insightful, thank you! I also loved how Bard's output completely mistakes the common physics riddle. (I have a physics background and your analysis is spot on IMHO.)