Whenever I see someone confidently making a comparison between LLMs and people, I assume they are unserious individuals more interested in maintaining hype around technology than they are in actually discussing what it does.
Someone saying "they feel" something is not a confident remark.
Also, there's plenty of neuroscience that is produced by very serious researchers that have no problems making comparisons between human brain function and statistical models.
That's what sustainable air fuel is. There is a lot being done to certify existing engines to run on it. Moving to an incompatible power source (such as hydrogen) isn't feasible because it would require scrapping the current fleets (and well maintained planes live longer than their pilots).
"They were just tracking how well his tweets were doing versus others. "
Yeah, and adjusting it, so he comes out best. That was Musks demand, as the other article shows, that is linked inside, after a Biden tweet performed better than Musk:
They officially boost people, who pay a little bit. Elon payed a lot.
And the source is clearly not the production source and never where in this shape - otherwise why sue someone, who open sourced it?
"But, the release of this source code also comes days after Twitter forced Github to take down other parts of Twitter's source code that was allegedly posted by a former employee without the company's permission. So, clearly, there's still plenty of Twitter that Musk still doesn't want us to see."
Also, you probably missed that:
"Zoë Schiffer of Platformer reported that Twitter actually removed part of the source code that affected the reach of Musk's and other user's tweets before releasing the algorithm to the public."
Which is consistent with quite some other statements, also from Twitter itself and the fact, that the source has not been updated in 8 months.
"But the underlying policies and models are almost entirely missing (there are a couple valuable components in [1]). Without those, we can't evaluate the behavior and possible effects of "the algorithm.""
Well, essentially the libertarian anarcho-capitalist bits.
TBH, I can can just strongly disagree with people like that, but what got me worked up about this pile of high-school grade political science was the way it conflated all that libertarian anarcho-capitalism with being pro-tech, and implying that any concern for the negative effects of developement (or of the economic system) therefore automatically puts you on the "enemies" side of things.
It's a frantastically offensive false catagorisation that I strongly reject.
I've been using SuperWhisper since the start and it's totally changed my workflow. I actually bought a mic because of it and I now mostly narrate my emails and other long form texts. The larger models are very accurate.
I know Neil is working on improving the speed of it using CoreML, which is I think the next big leap forward. Congratulations Neil!
I have a genuine appreciation for how Linear has built this. We have had to build something similar for our note taking application (Reflect). It is very tricky to do and I wish there was more research on this.
In my opinion, what we need is:
1) A client-side performant SQLite database that supports live queries. I.e. you can automatically re-render the page when the queries change. That way your database can drive the UI and be the source of truth in regards to what's displayed on the screen.
2) A separate realtime syncing protocol that syncs database state to client state.
And ideally this is all open source, and that these two endeavors are not coupled tightly.
[1] Wa-sqlite is the best (imo) client-side db - better than than the official Sqlite WASM build (for now) because it had a indexeddb fallback for browsers that aren't the cutting edge Chrome.
[2] cr-sqlite is an interesting project using CRDTs to sync state around. However I still believe that for many production use-cases you want a ultimate server source of truth.
[3] Replicache is still the best closed source solution I know of.
You've probably already seen https://riffle.systems/essays/prelude/ - one of my favorite recent research projects in this space, it is very much in line with what you're describing. Sad that it didn't seem to go anywhere after a couple introductory posts.
(Unfortunately, it is a 1-hour long video and I don't have time to watch it over).
The problem with 2 is often that the database is too far away from the view, hence, what you prioritize to sync might not be the most important thing the user interacts with. That's why there are ideas of syncing via GraphQL API because GraphQL is largely driving by client-side views, so it should have more knowledge on how to prioritize. But that often fails too.
This is not a problem if you are doing multiplayer semi-synchronous collaborations because the states of each player are mostly up-to-date. It is problematic if you are doing asynchronous collaborations and only sync with the server sparsely but have large amount of data to sync.
It is not a well-solved problem, but luckily I think the product surface requires this kind of "prioritized sync" is pretty small.
ElectricSQL are in this space, and have the concept of a central source of truth. SQLite in the browser and Postgres on the server with two way syncing using CRDTs. All open source but backed and funded.
I'm not convinced that separating the two tools is best for usability, loose coupling sure. But some level of integration enables the reactive front end querying to swap between the front end copy and central source depending on if you have the data copied locally.
Also many of the mechanisms needed for reactive re-querying are also needed for two way syncing.
