CS, AIS, PoliSci @ UC Chile.
Milan Weibelš¹
Karma: 174
LANAIS (Latin AmerĀiĀcan NetĀwork for AI Safety) kick-off
Just noticed the feature is deployed already. Thanks!
High-variance. Most people seem to have created an account and then gone back to being mostly on (eX)twitter. However, there are some quite active accounts. Iām not the best to ask this question to, since Iām not that active either. Still, having the bluesky account post as a mirror of the twitter acccount maybe isnāt hard to set up?
Do you actually have the skills to be competitive as a data scientist in a tech startup (or in some other role?), or would you need to get a few years of training and then complete alongside fresh grads?
This is a crux.
Another factor to consider is that junior software engineering roles seem to be getting more competitive. Iām not sure how much of this is attributable to macroeconomic cycle stuff vs AI automation reducing demand vs increasing supply of people qualified for these jobs.
EA Forum feature request: Can we get a bluesky profile link button for profile pages?
General-purpose and agentic systems are inevitably going to outcompete other systems
Thereās some of this: see this Gwern post for the classic argument.
People are trying to do this, and I just havenāt noticed
LLMs seem by default less agentic than the previous end-to-end RL paradigm. Maybe the rise of LLMs was an exercise in deliberate differential technological development. Iām not sure about this, it is personal speculation.
Left-progressive online people seem to be consolidating on an anti-AI position; but mostly derived from resistance to the presumed economic impacts from AI art, badness-by-association inherited from the big tech /ā tech billionaires /ā ātechbroā cluster, and on the academic side from concern about algorithmic bias and the like. However, they seem to be failing at extrapolation. āAI badā gets misgeneralized into skepticism about current and future AI capabilities.
Left-marxist people seem to be thinking a bit more clearly about this (ie extrapolating, applying any economic model at all, looking a bit into the tech). See an example here, or a summary (EDIT 2025-02-08: of the same example piece) here. However, the labs are based in the US, a country where associating with marxists is a very bad idea if you want your policies to get implemented.
These two leftist stances are mostly orthogonal to concerns about AI x-risk and catastrophic misuse. However, a lot of activists believe that the publicās attention is zero-sum. I suspect that is the main reason coalition-building with the preceding two groups has not happened much. However, I think it is still possible.
About the American right: some actors have largely succeeded in marrying China-hawkism with AI-boosterism. I expect this association to be very sticky, but it may be counteracted by reactionary impulses coming from spooked cultural conservatives.
Weird off-the-cuff question but maybe intentionally inducing something like experimenter demand effects would be a worthwhile intervention? After figuring out a way of not making recipients feel cheated or patronized, of course.
Probably not, or to a much lesser extent.
I would expect those to be the same person if AI turns out to not be a huge deal, which for me is about 25% of futures.
While I agree that strong founder effects are likely to apply if SpaceX and/āor NASA succeed in establishing a Mars colony, I expect that colony to be Earth-dependent for decades, and to be quite vulnerable to being superseded by other actors.
To put my model in more concrete terms: I expect whoever controls cislunar space in 2050 to have more potential for causal influence over the state of Mars in 2100 than whoever has put more people on Mars by 2040.
I think it would be a major win for animal welfare if the plant-based foods industry could transition soy-based products to low-isoflavone and execute a successful marketing campaign to quell concerns about phytoestrogens (without denigrating higher-isoflavone soy products).
I think it would be really hard (maybe even practically impossible) to market isoflavone-reduced products without hurting demand for non-isoflavone-reduced products as a side effect.
If the plant-based food industry started producing and marketing isoflavone-reduced soy products, I am quite confident that it would counterfactually lower total demand for soy products in the short term, and I am very uncertain about the sign of impact over the long term.
Hi Juliana! Thank you for your response, it indeed answers my question quite clearly.
Short reĀview of our TenĀsorTrust-based AI safety uniĀverĀsity outĀreach event
I love how thorough this post is. However, Iām not sure why you chose to research the production of vitamin D in an ASRS over other nutrients Pham et al. 2022 found would be deficient given adequate ASRS responses, such as vitamins E and K. ĀæAre the effects of vitamin D deficiency worse, or maybe it is more feasible to produce than vitamins E and K?
However, endorsing this view likely requires fairly speculative claims about how existing risks will nearly disappear after the time of perils has ended.
A note on this: the first people to study the notion of existential risk in an academic setting (Bostrom, Ord, Sandberg, etc.) explicitly state in their work many of those assumptions.
They chiefly revolve around the eventual creation of advanced AI which would enable the automation of both surveillance and manufacturing; the industrialization of outer space, and eventually the interstellar expansion of Earth-originated civilization.
In other words: they assume that bothThe creation of safe AGI is feasible.
Extremely robust existential security will follow, conditional on (1.).
Proposed mechanisms for (2.) include interstellar expansion and automated surveillance.
Thus, the main crux on the value of working on longtermist interventions is the validity of assumptions (1.) and (2.). In my opinion, finding out how likely they are to be true or not is very important and quite neglected. I think that scrutinizing (2.). is both more neglected and more tractable than examining (1.), and I would love to see more work on it.
Why not print the pdf?
I think it is very likely that the top American AI labs are receiving substantial help from the NSA et al in implementing their āadministrative, technical, and physical cybersecurity protectionsā. No need to introduce Crowdstrike as a vulnerability.
The labs get fined if they donāt implement such protections, not if they get compromised.
Humans could use AI propaganda tools against other humans. Autonomous AI actors may have access to better or worse AI propaganda capabilities than those used by human actors, depending on the concrete scenario.
It would probably be good AI Safety orgs explicitly and prominently endorsed the CAIS statement on AI Risk:
āMitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.ā