I feel like liability is underdiscussed here.
The technology is so new it seems to be covering an important and simple truth, that OpenAI is committing crimes. They are deploying technology that is hacking other companies. That's against the law. (and if this somehow doesn't count as a crime... shouldn't it?)
If you make the user or company who created the agent liable for every incident like this, I feel like that would do more for alignment than most alignment research. Incentives are powerful things.
It should horrify us to see a state, built in the aftermath of Jewish death and destruction, entertaining the ethnic cleansing of another people.
Ben-Gvir is openly laying out a plan to remove every Palestinian from Gaza. We cannot dismiss this as fantasy or fringe. This vision is now in the political mainstream, and that must be unequivocally rejected.
If a Democratic president had said this, many GOP officials plus an army of hacks, paid and unpaid, would go on TV and social media to pretend to care loudly enough for uninterested people to pay attention and agree it's bad. Dems don't do that.
Not really observations that others before me havenโt made, but:
- Despite the seemingly magical nature of LLMs, reflection over a >3 month timescale suggests my total productivity hasnโt increased by over 100%, or perhaps even by over 50%, and a lot of time is actually wasted because LLMs enable me to spend time on gratifying but low-productivity tasks that in the future turn out to not be useful
- Also, capabilities are incredibly spiky and highly correlated with the degree of investment poured into them, which my earlier tweet about math benchmarks implicitly points out
- From the above, it seems that the nature of LLM intelligence is wildly dissimilar to that of human intelligence and we wonโt trivially get to something superior to human intelligence in all important respects just by scaling up existing approaches with various tweaks; even if AGI Is eventually achievable, this implies a significantly longer timeline
- Benchmark progress is almost definitionally guaranteed to happen because the process of constructing a benchmark is a direct precursor to the process of constructing a training dataset used for hill climbing that benchmark, but the scope of what can be captured in a benchmark is (at least for now) grossly lacking in terms of its relevance to real-world work, with maybe several limited exceptions
- Progress seems highly gated by data but the nature of model training means that each โnext datasetโ is significantly harder to assemble than what preceded it; some wins are possible through synthetic methods but those feel more like โpatching up gapsโ than โpushing the frontier forwardโ
At a higher level, I guess Iโd say thereโs a sort of refusal to think carefully about what models are or are not useful for in a rigorous way which I find personally quite annoying, and instead a reliance on some nebulous notion of being โAGI pilledโ as a replacement for serious thought. I think people are very quick to anthropomorphize LLM intelligence because humans communicate through words and we infer the intelligence of human counterparties through comprehension of their language, but this leads them to wrong conclusions; for example if we observe that a new model proved some incredible mathematical theorem, some will say, โwell, donโt we have AGI now, huh?โ But to me, itโs actually more like, โwell, given how hard it would have been for a human to do these mathematics, and given the limited economic effect of LLMs upon the world so far, isnโt it actually a negative datapoint vis-a-vis the generality of LLM intelligence?โ
As a Jew, I am heartbroken and devastated at other Jews dehumanizing and celebrating attacks and murders of other people. This is a disgrace to humanity.
'At War on the Rocks, we spent thousands of dollars upgrading our office security posture and had to turn to the police. We have reason to believe that Thomas and Jennica Pounds (DataRepublican) and others were coordinating this electronic harassment campaign in terms of messaging themes'
Stumbled onto a new dark-money, conservative digital influence operation. Every month, someone is giving a Latvian company six-figure sums to manufacture support for the right online via the internet manipulation technique du jour: clip farming https://t.co/P6XEQV2ek3
This is a better way to visualize Columbusโ voyages than the traditional way (that shows him traveling through a modern map)..
It looks a lot crazier and more daring when you see it as a trip into the unknown.
I think we're long past that. The Netanyahu government knows. It doesn't care. And part of it -- mostly the Ben Gvir-Smotrich wing -- actually wants an isolated, introverted Israel.
Before right-wingers rush to their defense, try to work through the specific things this government has chosen to do, the details of their failures, and then explain them in any other way. A serious exploration of how this government handled last year's hunger crisis, or the constant violence they've deliberately refused to stop (and which Smotrich has actively defended) in places like Qusra, or the endless posturing to the base with insane rhetoric that each time, bit by bit, chipped away at the ability of even our best friends to expend any more political capital on standing with us, even when they knew in great detail the nature of our enemies and the need for us to defend ourselves -- all these phenomenon were either allowed to continue or were deliberately driven by Netanyahu and his ministers.
When I said these Israeli extremist assailants are far fewer in number than the anti-Israel propaganda claims but are still a real and disturbing phenomenon, that this violence is textbook terrorism and is having a terroristic effect on the Palestinian populations they're targeting -- each time I made these points, right-wingers, usually from overseas, rushed in to explain to me that it's all a mirage, a propaganda fiction, a libel.
Some even offered to introduce me to right-wing Israeli leaders to hear firsthand how everything is okay and there's nothing to see here. As if I'm new here. As if I get my information from the New York Times.
I know participants in these attacks. I've met and argued with planners of some of these attacks. I've watched this space for years and I have a pretty good handle on their psychology and religious ideas. And I'm telling you this is exactly what I say it is -- and the thousands of emails and angry tweets calling me "leftist" won't make the reality go away. They'll only enable this cruelty to fester and grow.
Reality doesn't care about your feelings. It doesn't notice your fears. The more you try to avoid it, the harder you'll crash when reality eventually claims its due. That's Zionism 101. As I've said every time I've spoken about this, if we don't act against this phenomenon now, aggressively and urgently, the damage will only grow -- to the innocents targeted by it, to the IDF weakened by it, to our ability to hold on to even our closest friends. And then it'll grow more, and then some more. Until eventually none of you will be able to pretend there's nothing to see.
I've written and talked for years about the deep and visceral bigotries of the anti-Israel campaign. It's from that vantage point that I say that this government has nevertheless done and is doing everything possible to make the bigots' claims seem reasonable even to our friends, even to people who know that our enemies are bad and know why we must defeat them.
And no one in this government is even remotely troubled by any of it.
Feel free to flood the comments with insults. But maybe also, if you can, try to squeeze a cogent explanation in there that actually grapples with these failures. We will be paying the costs of this folly for a long time to come. And what portion of those costs comes from the above will be our own damn fault.