Actually useful uses for LLMs:

1. Text checking - being a language model LLMs can look over written text and provide fixes traditional tools might have a problem with.
2. … I guess answering very general or language based questions, the kind that doesn’t require specifics or is specifically about language and understanding meaning. (eg. summerization, definitions etc.)
3. … … Entertainment? although an actual human being would almost certainly do a better job.
4. Can’t think of any more. I don’t even trust it enough to structure speech into a machine readable output.

In reality I think it actually just needs to be known that LLMs cannot reason and are just guessing and relying on their output is stupid. At that point it becomes really hard to profit off of them as all of the above can be achieved with small models running locally, which you can’t really monetise.

It’s a bubble. It will pop, and everyone laid off because of it will be rehired.

This got me thinking that Valve is the best competitor to Xbox. Instead of getting to release some walled garden, single store handheld they have to open it up to other stores because otherwise people would just buy a SteamDeck.

But everyone who cares about handheld gaming already has a steam deck and this is going to be more expensive so I don’t see the point anyway.

Don’t worry it’s going to launch with silksong. Don’t really know what’s the point of releasing a game along a cross-platform console but I guess the publicity is good.