1,743 karma · joined October 28, 2021
Jiggling disk heads, modulating fan rates, increasing and decreasing power draw... all are potential information leaks.
The current battle for digital personal rights is the right to private communication and data storage, and thanks to encryption and open source software, that one's not lost yet.
None of the systems I've seen achieve all those goals at once.
YAML, while at first sight a good idea, is irredeemably broken and should be deprecated for further use.
JSONC (https://jsonc.org/) is backwards-compatible with JSON, and a good target for long-term future migration.
.INI format works well as a structured subject-predicate-object tuple store for simple use cases.
We're probably going to have to live with that indifinitely, until someone comes up with a proposal that is better.
Compare this with things like X11 or Win32, where 20 year old programs will still work, or even more so systems like OS/360 and its successors, where 50 year old programs will still work.
I'm not a fan of React - to put it mildly - but something mediocre, but stable, is vastly preferable to the rapidly mutating hellscape of constantly rewritten frameworks driven by innovation for innovation's sake.
Discussion here suggests this may be limited to only a few recent versions, but this is still a nasty exploit:
If. But since the Labour government has a massive Labour majority that doesn't have to hold another election until 2029, that's not going to happen any time soon. Four years is a long time in politics.
For non-European readers, Reform are the equivalent of Germany's AfD or France's RN - or America's MAGA. It's a populist culture war party.
In response, far from doubling down, Labour has swung dramatically to the political right (by UK standards) of their normal political position in terms of the culture war issues like immigration and transgender rights that have been driving support for Reform.
We're having bad times in the UK, and no-one likes the incumbent government during bad times. (Never mind that the bad times are largely due to factors beyond their control, as is generally the case until a party has been in power for long enough to start turning the ship.)
Reform, meanwhile, are currently playing on easy mode; voters can say anything they like at the moment, and Reform are picking up the fantasy-football protest vote based on making wild promises they are unlikely to be able to fulfil.
None of this is to say Reform are not a serious threat. But the concept that their victory is inevitable is part their schtick, and should be resisted.
Providing a BGP feed of such provider network subnets to content providers would then allow them to happily serve content to those subnets without any further checks, safe in the knowledge that they will only be providing service to endpoints controlled by adults.
Details of how this can be done for other services including home broadband omitted - suffice it to say your router would have both adult and child-friendly SSIDs.
This seems both simple and obvious, and protects children without encumbering adults, risking privacy, or forcing a mass censorship regime on everyone.
The LLM thus did two things: first, it used a combination of the prompt, the user input, and the recent history of the chat and relevant extracts from the database state, to update the database with a new state; then after updating the database, it used all of the above to synthesize the output response.
A major component of the system was that, in addition to the usual blocks-world subject-property-object representation of the game state, the system also used the database to store textual fragments that encoded things like thoughts, plans, moods and plot points.
Adding programmatic constraints to the database and its updates made a vast difference to the quality of the output by preventing logical absurdity of the sort that has been described above - attempts by the LLM to do absurd things were simply rejected.
Finally, there was a hierarchical element to the story-telling, not just simulation. There is much more to storytelling than simulation, and experimenting with this became the most interesting thing about the whole enterprise; there is plenty of space for future research in this.
The downside was that all this complexity meant it was _extremely_ compute-intensive, and thus very slow as a result. But the gameplay was fascinating. We will no doubt see more things like this in future as LLMs get faster; it's an obvious application for open source software.
ext4 moved some time ago to 30 bits of fractional resolution (on the order of nanoseconds) and 34 bits of seconds resolution. It punts the problem 400 years or so into the future. I'm sure we will eventually settle on 128-bit timestamps with 64 bits of seconds and 64 bits of fractional resolution, and that should sort things for forseeable human history.
What is interesting about current LLM-based systems is that they follow exactly the model suggested by this paper, by bolting together neural systems with symbol manipulation systems - to quote the paper "connectionism can be seen as a complementary component in a hybrid nonsymbolic/symbolic model of the mind, rather than a rival to purely symbolic modeling."
They are clearly also kludges. As you say, they are built on shaky foundations. But the success - at least compared to anything that has gone before - of kludged-together neural/symbolic systems suggests that the approach is more fertile than any that has gone before. They are also still far, far away from the AGI that has been predicted by their most enthusistic proponents.
My best guess is that future successful hard-problem-solving systems will combine neurosymbolic processing with formal theorem provers, where the neurosymbolic layer constructs proposals with candidate proofs, to submit to symbolic provers to test for success.