That's totally expected. The field of large-scale generalist AI is entirely novel and experimental. It doesn't and can't have rigor of more mature disciplines that had decades to develop.
That said, there's no cargo cult in blindly using heuristics for certain fundamental LLM phenomena that have tons of good studies backing them (e.g. have no extra distractors, group and delimit pieces of the context, etc). If you want quantitative rigor, perform correct evals on your specific task and model.
>It also tackled illegal immigration, war criminals in Bosnia, those behind 9/11 and other terrorist plots, as well as the Russian black market in weapons-grade plutonium.
Ah, the "25kg of Pu-239 in a closet" guy. [1] That was a hoax, not even a scam he fell to but a deliberate fake entirely made up by him.
I'm not quoting anyone, and being extremely familiar with the era in question, I do not believe the ITC, regardless of what evidence they had. That episode is a ludicrous, physically impossible hoax filmed with a "concealed" camera that contradicts everything that was ever known about nuclear arms control at that time (yes I know about Project Sapphire), internal struggle between siloviki, KGB leftovers' disarray in early 90's, and especially the Solntsevskaya gang itself. Every single illegal activity on that scale at that time was uncovered in the Russian press and countless documentaries were filmed about those, but this case didn't even have any smoke. Suspicion that they fell for a scam (rampant in 1993) doesn't align with a well funded investigation, I have zero doubts they just pulled everything out of their asses and outright faked it, hiring the actors, possibly even from the gang itself. Things like this were charitably called "cranberries" in 90's lingo, only outsiders could believe something like this.
The article is all over the place and fails to convey author's point. I think he's just confused about what "LLM" (as opposed to "AI", what???), "deterministic", and "probabilistic" mean, let alone nuances of any of these.
It's usually implemented as a keyword the model is trained to react to. In API-only models like Claude you don't have the access to it, as it's inserted into a masked area. The reason the cache might break is that it's inserted at the start, and if you change it the rest will be invalidated.
Most fatbikes have a much more powerful motor than a regular e-bike, and can be used without pedaling at all. They are listed as 250W, but it's actually a software-limited 750W motor. Local vendors provide unlock instructions along with the bike.
It's just a motorcycle in disguise. Then, to make matters worse, they weigh 30-40KG + two 12-year-olds on top - that, combined with reckless riding, can do a lot more damage than a boring cycling incident.
I mean ordinary as non-electric. I know what an electric fatbike can do of course. The article makes it ambiguous as it's talking about the common limit for all cyclists.
These are undrivable as a city commute, not to mention the stealing risk as they're usually expensive. I knew some people who attempted and gave up after few weeks.
People drive fatbikes from solely selfish reasons. They want something heavier and larger than others, they want others to give way and be afraid of collision.
Nah, its almost for sure xenophobic and classist. I think fat bikes are the best thing that happened to amsterdam because all the scooters disappeared. All the poor people that used to ride scooters are now riding fat bikes. Its better for their health and their wallet. Yes they tend to be a bit less educated and socialised, but they were like that when they drove scooters as well. Banning fat bikes is gesture politics.
Games three-letter agencies play are the same everywhere and have zero relation to the culture. 2016 meddling did happen of course. It was also negligible and led to a huge overreaction, extremely similar to the US meddling in Russian elections in 1996 where Clinton admin indirectly prevented Nemtsov from running by supporting unpopular Yeltsin (and NGOs did a ton of "work" which barely affected anything, the main reason Yeltsin won was Filatov running the campaign, oversized spending and collusion with the media aka Xerox affair, and the "admin resource" he had).
Stupid generalization, just like propagandist stereotypes about obese redneck Americans and perverted Europeans.
Only propagandists and their victims can throw rocks like this without any discernment.
Any nation is a collection of vastly different personalities, out of whom one could of course pick a group with sought after characteristics, but taking one such group and extending its characteristics to the whole nation is beyond stupid.
It's OK if you implied "Russian government", but that better be explicit.
Exactly. Russians seem to understand the word "Nazis" is a less-literate way than the rest of the world.
"Nazi" -- in Russia -- means anybody that Russia hates, resents, is jealous of, or otherwise dislikes. Anybody not explicitly in Russia's thrall is "anti-Russian", considered an enemy, and is hence labelled a "Nazi".
Coincidentally, this "Whoever isn't with us, is against us" thinking is a hallmark of fascist regimes like the actual Nazi Germany, where it was an article of faith.
>outrageous copyright infringement
>unethically scrapped data
Hahahahaha
reply