Question for either James or Julia: Is this specifically for lead policy or just policy advocacy in general? And can you elaborate why?
anormative
Karma: 173
This is awesome! Any details you can share on how this whole thing came together? It could be really impactful to try to aim for more coalitions like this for other cost-effective opportunities.
To clarify, I agree that that the ways you can be liable mostly fall into the two categories you delineate but think that your characterization of the categories might be incorrect.
You say that a developer would be liable
if you developed a covered model that caused more than $500M harm
if you violated any of the prescribed transparency/accountability mechanisms in the bill
But I think a better characterization would be that you can be liable
if you developed a covered model that caused more than $500M harm→ if you fail to take reasonable care to prevent critical harmsif you violated any of the prescribed transparency/accountability mechanisms in the bill
It’s possible “to fail to take reasonable care to prevent critical harms” even if you do not cause critical harms. The bill doesn’t specify any new category of liability specifically for developers who have developed models that cause critical harm.
To use Casado’s example, if a self-driving car was involved in an accident that resulted in a person’s death, and if that self-driving car company did not “take reasonable care to prevent critical harms” by having a safety and security protocol much worse than that of other companies, it seems plausible that the company could be fined 10% of their compute/have to pay other damages. (I don’t know if self-driving cars actually would be affected by this bill.)
I think the best reason this might be wrong is that courts might not be willing to entertain this argument or that in tort law “failing to take reasonable care to avoid something” requires that you “fail to avoid that thing”—but I don’t have enough legal background/knowledge to know.
Thanks for your reply! I’m a bit confused—I think my understanding of the bill matches yours. The Vox article states “Otherwise, they would be liable if their AI system leads to a ‘mass casualty event’ or more than $500 million in damages in a single incident or set of closely linked incidents.” (See also eg here and here). But my reading of the bill is that there is no mass casualty/$500 million threshold for liability like Vox seems to be claiming here.
Kelsey Piper’s article on SB 1047 says
This is one of the questions animating the current raging discourse in tech over California’s SB 1047, newly passed legislation that mandates safety training for that companies that spend more than $100 million on training a “frontier model” in AI — like the in-progress GPT-5. Otherwise, they would be liable if their AI system leads to a “mass casualty event” or more than $500 million in damages in a single incident or set of closely linked incidents.
I’ve seen similar statements elsewhere too. But after I spent some time today reading through the bill, this seems to be wrong? Liability for developers doesn’t seem to be dependent on whether “critical harm” is actually done. Instead, if the developer fails to take reasonable care to prevent critical harm (or some other violation), even if there is no critical harm done, violations that cause death/bodily harm/etc can lead to fines of 10% or 30% of compute. Here’s the relevant section from the bill:
(a) The Attorney General may bring a civil action for a violation of this chapter and to recover all of the following:
(1) For a violation that causes death or bodily harm to another human, harm to property, theft or misappropriation of property, or that constitutes an imminent risk or threat to public safety that occurs on or after January 1, 2026, a civil penalty in an amount not exceeding 10 percent of the cost of the quantity of computing power used to train the covered model to be calculated using average market prices of cloud compute at the time of training for a first violation and in an amount not exceeding 30 percent of that value for any subsequent violation.
Has there been discussion about this somewhere else already? Is the Vox article wrong or am I misunderstanding the bill?
SummaryBot hallucinated an acronym! UGAP is the University Group Accelerator Program, not the “Undergraduate Priorities Project.”
What are those non-AI safety reasons to pause or slow down?
Checked to see if it had been released, but it looks like the release date has been pushed back to August 23!
At a gut-level, this feels like an influential member of the EA community deciding to ‘defect’ and leave when the going gets tough. It’s like deciding to ‘walk away from Omelas’ when you had a role in the leadership of the city and benefitted from that position. In contrast, I think the right call is to stay and fight for EA ideas in the ‘Third Wave’ of EA.
I’m sure you mean this in good faith, but I think we should probably try to consider and respond meaningfully to criticism, as opposed to making ad hominem style rebuttals that accuse betrayal. It seems to me to be serious epistemic error to target those who wish to leave a community or those who make criticism of it, especially by saying something akin to “you’re not allowed to criticize us if you’ve gained something from us.” This doesn’t mean at all that we shouldn’t analyze, understand, and respond to this phenomenon of “EA distancing”—just that we should do it with a less caustic approach that centers on trends and patterns, not criticism of individuals.
Thank you!
Question: what’s the logic behind organizations like Good Ventures and people like Warren Buffet wanting to spend down instead of creating a foundation that exists for perpetuity, especially if they have such large amounts of money that they struggle to give it all out? Is it because they believe some sort of hinge of history hypothesis? Are they worried about value drift? Do they think that causes in the future will generally be less important (due to the world being generally better) or less neglected (“future people can help themselves better than we can?”)
Is there any writing from RP or anywhere else that describes these flaws in more depth, or actually runs the numbers on EV calculations and x-risk?
I’ve often found it hard to tell whether an ideology/movement/view has just found a few advocates among a group, or whether it has totally permeated that group. For example, I’m not sure that Srinivasan’s politics have really changed recently or that it would be fair to generalize from his beliefs to all of the valley. How much of this is actually Silicon Valley’s political center shifting to e/acc and the right, as opposed to people just having the usual distribution of political beliefs (in addition to a valley-unspecific decline of the EA brand)?
I’m pretty confident that a majority of the population will soon have very negative attitudes towards big AI labs.
Can you elaborate on what makes you so certain about this? Do you think that the reputation will be more like that of Facebook or that of Big Tobacco? Or will it be totally different?
Will it be focused on GCRs again?
It feels like in the past, more considerateness might have led to less hard discussions about AI or even animal welfare.
I think factory farmed animals is the better example here. It can be pretty hurtful to tell someone you think a core facet of their life (meat eating) has been a horrendous moral error, just as was slavery or genocide. It seems we all feel fine putting aside the considerateness consideration when the stakes are high enough.
Can you elaborate on what you mean by “the EA-offered money comes with strings?”
But the AIs-with-different-values – even: the cooperative, nice, liberal-norm-abiding ones – might not even be sentient! Rather, they might be mere empty machines. Should you still tolerate/respect/etc them, then?”
The flavor of discussion here on AI sentience that follows what I’ve quoted above always reminds me of, and I think is remarkably similar to, the content of this scene from the Star Trek: The Next Generation episode “The Measure of a Man.” It’s a courtroom drama-style scene in which Data, an android, is threatened by a scientists who wants to make copies of Data and argues he’s property of the Federation. Patrick Stewart, playing Jean Luc-Picard, defending Data, makes an argument similar to Joe’s.
You see, he’s met two of your three criteria for sentience, so what if he meets the third. Consciousness in even the smallest degree. What is he then? I don’t know. Do you? (to Riker) Do you? (to Phillipa) Do you? Well, that’s the question you have to answer. Your Honour, the courtroom is a crucible. In it we burn away irrelevancies until we are left with a pure product, the truth for all time. Now, sooner or later, this man or others like him will succeed in replicating Commander Data. And the decision you reach here today will determine how we will regard this creation of our genius. It will reveal the kind of a people we are, what he is destined to be. It will reach far beyond this courtroom and this one android. It could significantly redefine the boundaries of personal liberty and freedom, expanding them for some, savagely curtailing them for others. Are you prepared to condemn him and all who come after him to servitude and slavery?
This seems awesome! Thanks for sharing.
We’ve tried to do research-based meetings in the past, but we’ve found that people tend to just focus on debating abstract or shallow topics and we haven’t been able to sufficiently incentivize diving in to the more nitty-gritty details or really digging for cruxes. This might not have worked for us because we tried to have too much control over the research process, or because we presented the activity as a debate, or maybe because of the makeup of our group.
Some questions: Did all of the meetings go well? Did you notice any of the issues I mentioned (if not, any idea why)? How many people did you do this with? Were they all post-Intro fellows or selected for in some other way? How much progress did you make on the questions?
Can you tell us a little bit about how this project and partnership came together? What was OpenPhil’s role? What is it like working with such a large number of organizations, including governments? Do you see potential for more collaborations like this?