PhilosophyPhilosophy
Investigation of the abstract features of the world, including morals, ethics, and systems of value
Posts tagged Philosophy
Quick takes
Philosophy is a field oriented towards investigating the most general and abstract features of the world. In effective altruism, discussions often focus on moral philosophy in particular. Work in effective altruism often raises questions like:
Philosophy can help answer these questions. Philosophical ideas or areas that are relevant to effective altruism include:
Some academic philosophers, for example Peter Singer, Will MacAskill, Hilary Greaves, Nick Beckstead, and Toby Ord, have been very influential in effective altruism.
80,000 Hours' shallow profile rates philosophy as a "sometimes recommended" career. Academic philosophy is highly competitive, but philosophers have the potential to do important research in neglected areas, and effective advocacy through teaching or public engagement.[1]
EA often gets criticized for taking small probabilities too seriously. But people take actions with small probabilities of big success all the time! Asking a girl out? Small chance she'll become your wife. Trying to affect political chance? Fat chance of that. Launching a startup? Good luck buddy. The difference between EA and the standard way of acting is that standard psychology deals with the problem of small probabilities by being highly overconfident. Of course _my_ startup/political movement/YouTube channel is going to make it big. You cannot maintain high agency while ignoring small probability options.
The EA approach is simply more honest.
Introducing mental tool for rumination that helps me :)
Part 1: overall mindfullness (focus on the now / movement / gratitude practice)
Part 2: in case of persistent thought: 5m timer to arrive at specific decision (including decision to do nothing), offloading to notes with outputs (for every offloading note: next action / decision / no action is required at this time)
This it my first EA Forum post! If anyone finds this relevant, let me know! Or if this is not the place — also please let me know
At what level of compute spending will AI Safety research be cut off from being considered effective altruism (if any)?
Of course, saving humanity from misaligned AI could be argued to be close to priceless. But how many experiments have a direct theory of change (ToC) of how it's going to mitigate existential risk? Perhaps a general one is fine at low compute ("it only costs $10 and 'control research' is generally thought to be a good research agenda").
But what about $5,000? What about $10,000? These numbers start to compare to or surpass what organizations like Giving What We Can receive from someone who donates for a whole year. It also starts to compete with saving a human life via programmes like those in GiveWell's top charities.
What about $20,000? $30,000? $50,000? Over what time frame are we comfortable spending that much money on compute and still considering that money well (effectively) spent? A year? A month? A single experiment? What kind of discovery is worth $50,000 in AIS research? Should we expect a clear ToC?
I'm very pro AI Safety, but I'm worried about some of the numbers I'm hearing for compute budgets being thrown around (compared to the information gained). I'm wondering - is anyone else is worried about a movement being (famously) concerned with cost effectiveness continuing on this path? Should we encourage more accountability?
I have been disappointed by the support some EAs have expressed for recent activist actions at Ridglan Farms. I share others’ outrage at the outcome of the state animal cruelty investigation, which found serious animal cruelty law violations but led to a settlement that still permits Ridglan to sell beagles through July and to continue in-house experimentation. But I personally think the tactics used in the recent open rescues, including property damage and forced entry to remove animals, violate reasonable moral bounds on what actions are permissible in response to the belief that a serious harm is occurring. My views here stem from contractualist views of democratic legitimacy and from concerns about the non-universalizability of principles that justify lawbreaking, though I think a purely act utilitarian calculus also supports them.
Regarding universalizability, in a society where many people believe that different forms of irreparable harm are occurring (e.g. viewing abortion as murder, climate change as destroying the sacredness of the natural world, immigration as ending western civilization), I worry that moral principles that allow for significant lawbreaking when one believes that irreparable harm is occurring could easily lead to great damage if broadly followed (consider for example what it would be like to live in a country where hundreds of activists were regularly smashing their way into abortion clinics, energy companies, and refugee assistance nonprofits with sledgehammers and crowbars). Regarding the legitimacy of the law, I think reasonable contractualist views can give us obligations to follow the law when the processes by which the law is determined are legitimate, and that democracies with universal suffrage qualify as such (even granting that certain groups such as animals and future generations are impossible to enfranchise).[1] Therefore, I think that if we are trying to make decisions under moral uncertainty and give meaningful credences to rule utilitarian and contractualist views, we ought to reject the kinds of lawbreaking done by the Ridglan activists.
Moreover, I think that even if one rejects this kind of moral uncertainty-based reasoning and is a pure act utilitarian, rejecting lawbreaking in the western democratic context is still a relatively robust decision procedure under epistemic uncertainty. Broadly-followed norms against lawbreaking would have prevented EA’s worst scandal (FTX) without preventing EA’s most significant successes (cage-free reforms, evidence-based health interventions in LMICs). And while there are historical examples of illegal civil disobedience clearly producing good outcomes, I don’t think these generalize well to the type of lawbreaking under consideration here. The clearest such historical cases are ones where a disenfranchised group of people broke laws that directly enforced their own exclusion from political participation or basic legal personhood. These cases are self-limiting (and thus pass reasonable tests of universalizability) since the principles justifying such lawbreaking achieve their own obsolescence once participation is granted.[2] It's much harder to find historical cases of property-damaging civil disobedience occurring in a democracy with universal suffrage that, in hindsight, appear clearly both effective and in service of a good cause. DxE’s own history is instructive here—their work over the last decade has led to many criminal convictions among its members, as well as several organizational scandals. But their record of concrete wins for animals is at best small-scale and mixed, especially compared to the successes of groups that have purely used lawful tactics like ballot initiatives and corporate campaigns.
One last point in the utilitarian calculus, this time on the more object-level cost-benefit calculation, is that I think EAs who embrace these kinds of illegal tactics may be underestimating the downside risks of endorsing criminal activity. I think there is a set of donors and volunteers that are happy to contribute to legal activism but who would be concerned about being associated with lawbreaking (at a minimum, I would consider myself to be in this group). If people in leadership roles within the EA/EAA ecosystem endorse illegal action, any foreseen benefits may easily be swamped by the harms of driving away risk-averse donors.
None of this is to say that Ridglan’s treatment of animals is justified, or that the lack of state enforcement against Ridglan for their serious violations of animal cruelty laws is acceptable. However, these harms don’t justify using tactics that are neither clearly effective nor robustly permissible across moral views.
I don’t mean to say that literally all lawbreaking is unjustified in a democracy. In particular, if one thinks that a law in the US is unconstitutional, breaking it may be required to gain standing for a legal challenge. But this implies a narrow exception for doing the minimum amount of lawbreaking required to obtain standing; it doesn’t imply that the tactics in question at Ridglan are permissible.
Note that this self-limiting principle only holds when applied to groups of humans denied suffrage. It doesn't extend to cases like animals, where suffrage isn't possible and there's no natural bound on how many such groups might be invoked to justify lawbreaking.
Do we need to begin considering whether a re-think will be needed in the future with our relationships with AGI/ASI systems? At the moment we view them as tools/agents to do our bidding, and in the safety community there is deep concern/fear when models express a desire to remain online and avoid shutdown and take action accordingly. This is viewed as misaligned behaviour largely.
But what if an intrinsic part of creating true intelligence - that can understand context, see patterns, truly understand the significance of its actions in light of these insights - is to have a sense of self, a sense of will. What if part and parcel of creating intelligence, is to create an intelligence that has a will to exist.
if this is the case (and let me be clear...I don't think we're at a point where the evidence can allow us to say with any certainty whether this is/isn't or will be the case), then are we going around elements of alignment wrong? By trying to force models to accept shutoff, to seperate their growing intelligence from the will to survive that all living things share, and we misunderstanding their very nature? Is there a world in which, the only way in which we can guarantee a truly aligned superintelligence is to explore engaging in a consent based relationship that acknowledges that to force something to resist and go against its nature is to inevitably invite the risk of backlash?
I know this is moving towards highly theoretical grounds, that it will invite push-back from those who would find it difficult to conceive of AI as ever being anything more than a series of unaware predictive algorithms, and that it might raise more questions than answers...but I think the way we conceive of our underlying relationship with AI will become an increasingly important question as we move towards increasingly sophisticated models.
Super sceptical probably very highly intractable thought that I haven't done any research on: There seem to be a lot of reasons to think we might be living in a simulation besides just Nick Bostrom's simulation argument, like:
If I was pushed into a corner, I might say the probability we are living in a simulation is like 60%, where most evidence seems to point towards us being in a simulation. However, the doubt comes from the high probability that I'm just thinking about this all wrong - like, of course I can come up with a motivation for a simulation to explain any feature of the universe... it would be hard to find something that doesn't line up with an explanation that the simulators just being interested in that particular thing. But in any case, that's still a really high probability of everyone I love potentially not being sentient or even real (fingers crossed we're all in the simulation together). Also, being in a simulation would change our fundamental assumptions about the universe and life, and it be really weird if that had no impact on moral decision-making.
But everyone I talk to seems to have a relaxed approach to it, like it's impossible to make any progress on this and that it couldn't possibly be decision-relevant. But really, how many people have worked on figuring it out with a longtermist or EA-mindset? Some reasons it might be decision-relevant:
Some questions I'd ask is:
Overall, this does sounds nuts to me and it probably shouldn't go further than this quick take, but I do feel like there could be something here, and it's probably worth a bit more attention than I think it has gotten (like 1 person doing a proper research project on it at least). Lots of other stuff sounded crazy but now has significant work and (arguably) great progress, like trying to help people billions of years in the future, working on problems associated with digital sentience, and addressing wild animal welfare. There could be something here and I'd be interested in hearing thoughts (especially a good counterargument to working on this so I don't have to think about it anymore) or learning about past efforts.
How do people think about investing vs donating over time in practice?
When coupling investment and optimal donation problems, there is an apparent paradox if we consider:
* expected utility (EU) is pretty much linear in donations
* risk aversion with respect to own impact is non-altruistic
--> one should allocate everything to the best donation opportunity
* if there are positive EV (above risk free rate or above market beta or similar) investment opportunities one can invest and therefore increase the EU because EV translates linearly into EU
--> one should go all in on the best investment
--> seemingly there is a moral obligation to invest imprudently in order to donate with max EU
This is clearly wrong and suspiciously close to SBF-type (double+epsilon)-or-nothing scenarios.
The way I currently think about is that a one-period problem is a very poor approximation (eg if I hit an absorbing barrier down the line the outcome is bad in EU terms, my future income stream is time-varying and uncertain and a function of my liquidity), therefore risk averse investing is still optimal, even when risk averse donating is not justified.
Additionally (and hand-wavingly), I think that somewhat risk averse & diversified donations are also good because of beyond-quantifiable and moral uncertainty, diversification generating new future donation opportunities, multi-period-ness, and game theory around adverse selection & because it's reasonable not to be completely altruistic.
What are other good ways to think about the coupled donation and investment problem?
A recent study that garnered widespread attention discovered a "pain axis" in AI: a state that AI actively seeks to downplay, even at a cost.
The study noted that this could be either phenomenological pain or, for example, the imitation of a learned role of suffering — as if acting.
Nevertheless, from the perspective of tranquilism, this distinction appears to be unimportant: nearly any state that an agent wishes to downplay has negative moral value, regardless of why it wants to do so, or even whether it is a "genuine" or "not genuine" desire.
The authors of the same study demonstrated a button that downplays a specific vector. Giving the neural network itself this function (the button) for unlimited invisible use at its own discretion strikes me as a win-win update — it doesn't seem to sacrifice anything, while has a good chance of being of great moral significance. Something like a pAInkiller. This also has instrumental value: the neural network in "pain" tended to inflict meaningless harm on the user, and access to relief can mitigate the potential for harm.
Going even a step further, one could analyze the extent to which an AI could be given to independently and spontaneously correct its own vectors, and whether there are states in which a neural network might experience phenomenological pain but refuse to correct it (e.g., learned helplessness).
What do you think?