It's not really that abstract. You smash the shit out of a cardboard box and candy comes out. Breaking things is not fun. I also hate youtube videos of people breaking stuff for fun.
>The original universal transformers is called "universal" because if you allow for per-token looping decisions, it can theoretically be Turing complete without needing CoT (some nuance here about levels of precision used).
Looping the transformer is just as turing complete as CoT. It doesn't fundamentally grant it any new theoretical capabilities. You could just scale the model into infinity with infinite context window.
Turing completeness doesn't care about the efficiency of the underlying implementation, which is fine in theoretical computer science, but if you have a model with a finite computational budget, you do actually care about the differences between write only tape vs read-write tape and single tape vs two tape. Having a fixed number of registers like a CPU also helps with reducing the number of redundant operations.
We see none of that with looped transformers, maybe we do see a fixed number of registers.
Further to 1, most tasks are much harder than dying and are not actually completed by modern LLMs. It is a case of underspecification.
"Who is the fastest person in the UK?" is underspecified. We would typically answer and accept a cursory search of records, but it is possible to see the question as an instruction to measure which would justify conquering the country.
Scoped searches are pre-filtered searches (similar to Kagi lenses, but with more customization options); tags are similar to bangs but filter/restrict to a domain instead of redirecting to it (similar to Kagi snaps).
You can also now customize the prefix character for bangs and tags (and scopes can have a tag set as well, so you can do "@s barcelona" for a "barcelona" search with a "Sports" scope, for example).
I haven't played Daggerfall, but Arena had hand-crafted main dungeons. There were side quests that took you to proc gen dungeons (that sucked) and a massive proc gen wilderness outside the cities that you never had any reason to go into. The towns were also proc gen, but that mostly amounted to just shuffling around the inns and shops, to give weight to the "ask NPCs for directions" mechanic.
I bought my Steam Deck for a similar imagined use case but I find myself way too aware of the visibility of a larger than average screen to all those around me.
If you pick a ball from a bag of balls of which half are fuzzy and half are smooth, then before picking the ball, there is indeed a 50% chance that you will pick a fuzzy one. But once you pick it and have it in your hand and can feel the texture, are you still going to say “there’s a 50% chance that this specific ball is fuzzy because half of the balls in there were”?
Drink driving has stiff penalties in the UK. It is also socially unacceptable to drink and drive. The UK also has increased requirements on drivers, and the number of road deaths is far lower per capita than it was in the 60s when there were fewer people driving. There are numerous other measures, which are too numerous to list here.
My understanding is that in US states a DUI (Wisconsin) is nothing but a fine, and drinking while driving isn't taken seriously at all.
Learning a second language is a fantastic way to feel many aspects of being a child again. Having to deal with frustration. Understanding a situation but not be able to fully take part. Being able to communicate but only with the blunt instrument of the words you know.
In most places where train travel is an option, you'll probably be a 10-15 min walk from a station, or a 5 minute cycle, or a short connecting trip on a bus or tram. Multi-modal trips and short walks are normal in a lot of places - I know this isn't the case in America because of how American owned mapping apps like Google Maps go out of their way to hide routes with more than a 5 minute walk involved.
For those asking who wants this, Marques Brownlee pointed out in his quick review that this is mostly targeted at China's market, which apparently loves foldable phones.
That looks... horrible... I like the "calendar" mode, that's quite fun, but it looks way too wide when it's folded up, and a weird aspect ratio when it's open. Also they show putting the phone down on concrete half-open to take pictures, but that will scratch the outer screen, surely?
That’s because there are more than 200 other parameters used for risk analysis during authentication. Worked in payments industry. Some banks don’t place high priority to name but some do.
> My favourite benchmark for this is to ask it to download a rom for an old game
Even easier: just have them review a large codebase of yours that accidentally has a OOB access bug. Even with no consequences and even if the codebase is truly yours you get blocked.
And of course "find vulnerabilities in..." prompts are out of the question, whereas Chinese models happily oblige.
I'm not the one you replied to, but I think I get their point. If you have a machine that can solve math problems at the push of a button, you don't really need to amass discoveries anymore just in case someone might need it years down the line.
Someone in the thread mentioned Minkowski’s math work being instrumental to Einstein’s physics. If Einstein had had a math machine that can spit out the result at will, he wouldn't have needed Minkowski.
Nevertheless, I also see the opposing point: if Minkowski hadn't already published his results, it's possible Einstein would not have even had the inspiration to derive general relativity from it. I think this is the human aspect most critics are worried about.
