10,958 karma · joined February 28, 2014
A well-known anticompetitive tactic is for a big player with a vast bank account to move into an area and eat losses while waiting for their competitors to starve. Well observed in the 90's movie rental market, so certainly real.
AI is shaping up to be a major concern amongst the non-tech segment of the population. Some of the arguments are good, some are dumb, but people largely are forming up against AI and datacenters. Yes, you should expect politicians to act on that; it's how the system is meant to work.
When all of your factory bits are randomly wearing down, causing speed mismatches and stochastic outright failures, large and complex chains become limited by your ability to run around and diagnose and fix things. Things also tend to fail in novel and unexpected ways; redesigns can help mitigate some failures, while new failures may be introduced. And as reliability improves, you start seeing failures that occur on longer timescales, which more trivial failures would initially mask.
This could hopefully happen somewhat faster with Israel, given that the government isn't a literal Nazi remnant, as the South African government was. (Though, ironically, the main arms supplier to South Africa during the arms embargo was Israel, often reselling weapons from the United States. Sigh.)
"By the late 1980s, the South African state was in profound crisis, crippled by international sanctions, a stagnant economy, and widespread resistance. Recognizing the unsustainable nature of the perpetual conflict, the government under P.W. Botha began tentative, secret contact with the imprisoned Nelson Mandela and the exiled African National Congress (ANC) leaders."
https://en.wikipedia.org/wiki/History_of_South_Africa_(1948%...
The real question for education is whether we can get 'tutor-quality' education from ai interactions. One on one education is leagues better than classroom education; it's the one-true-result in education research. There's a massive opportunity here to improve outcomes, and in a generally available way. You can't clone a good teacher, but you can give everyone an ai tutor.
Additionally, this all happened after mythos was held back due to cyber security concerns (April, 2026).
#mmt bay-beeee!
The whole world of fuzzing is about brute forcing exploits by exploring unlikely inputs. Fuzzing a system which hasn't been previously fuzzed will almost certainly turn up a pile of bugs, some of which may be exploitable.
So, both are true. Pretty dumb exploration is very likely to find bugs and even exploits. It seems unsurprising to me that an agent swarm could do better than a fuzzer, even as a better, more directed but still broad exploration.
You should have a fscking air gap.
Treat it like nukes when you're turning the safety filters off. This is very much OpenAI screwing up, running obviously unsafe tests.
I'm looking forward to making a Factorio themed war game table. Star Wars Legion on the factory planet...
It was the hardest task (alongside godwit call types, bat species (pitch shifted into the audible range), Marine mammal species, and frog species), but still worked!
Ada Palmer's terra ignota. It takes 'the past is another country' and really sends it forward. The world she draws is fundamentally different from our own, as ours is from medieval France. Absolutely fascinating books, with their own set of Utopians trying to make a better world for the future...
And I'm the vibey solar punk direction, Becky Chambers' 'Monk and Robot' is a fascinating pair of novellas about humanity reestablish ing contact with robots after they decided to just leave 250ish years previously. It is very much a book about a Good Future, where people still have people sized problems.
Any way you cut it, this is a major achievement for AI, besotted with human drama over whose prompt should be recognized by the history books.
Now: Could Astra agents have hacked their way into the OpenAI logs to find human mathematicians with a good lead on the problem to build upon? Certainly doesn't seem impossible.
Palmer's handling of the different groups in Terra Ignota takes both sides, in every case. The case for why people join the imperial group has to be made, and made honestly. But problems with that approach are also pretty clear, and well explored: at the very least, elites get weird if they aren't responsive to the people, and the elites in Terra Ignota get very, very weird. I think she takes a bit more time setting up the positive case for the imperials exactly because it's pretty alien to us. Or, at least it was before our current global swing towards authoritarianism.
Europeans were also once pretty obsessed with religion, as you might recall. They may be less so now, but who's to say the pendulum won't swing? This is one of the things I loved about the series -- the future or build is very different because the pendulum has swung in ways that aren't obvious to us, from our modern perspective. Just as the people of Florence would find some of our obsessions pretty strange and unexpected.
Distillation is alive and well... Earlier work on model printing also found that it's pretty easy to find smaller sets of parameters which can replicate the behavior of the entire network with pretty good fidelity.
Large parameter counts give space to explore, and give routes out of what would be local minima in a lower dimensional space.
In other words, there's no guarantee that any given trained model is a minimal representation of its training set.
Timothy Mcveigh (and Terry Nichols) weren't members of a terrorist organization, but they were terrorists. Meanwhile, the ku klux klan has actual initiations, membership, and organizational structure. As do Isis and al qaida.