★ By 2025/2030/2035 will there be a "best AI safety practices playbook" which all leading labs in the US claim to follow?
 

unjournal.org Just released our evaluation package for Schuett et al's "Towards best practices in AGI safety and governance". 

Please see here.

Why not what seems to be the obvious mechanism: the cuts to USAID making this more urgent and imperative. Or am I missing something?

We (Unjournal.org) ultimately decided to de-prioritize this after a second opinion and our internal vote. We might reconsider it if others point out flaws in the reasoning above.

I've shared the reasoning/discussion in the relevant row of our Public Database of research with potential for impact. (Column: "Discussion: Why is this impactful/relevant"). I hope to share more of these discussions in that public database, especially if people indicate interest.

Unjournal.org is piloting an expansion into evaluating legal research. Please contact me if you are interested.

I would lean the other way, at least in some comms. You wouldn’t want people to think that (e.g.) “the career guidance space in high impact global health and wellbeing is being handled by 80k”. Changing branding could more clearly open opportunities for other orga to enter spaces like that.

Policy-focused grad student groups

I think there could be value in running EA groups for niche populations of students who might go on to have high-leverage roles. Policy students strike me as one of the clearer targets for this.
...
Career panels or similar: Career-focused events targeting specific groups, such as late-stage PhD students in economics or public health, could be impactful during critical decision-making periods.

The Unjournal would be interested in helping facilitate and sponsor reading/discussion/evaluation groups for PhD students, especially in economics and policy. This could be compatible with and linked to such grad student groups and career events. 

If you are an academic or PhD student interested in piloting and organizing this, please contact me/us.  You can also peruse our notes on this potential initiative here and add comments.
 

Possibly naive question: are the projects open source, shared code, and forkable?

I expect there are some cohort effects (people in more recent generations have a higher probability of being involved).  In particular, many people get into EA via university groups (although it may not be the place they 'first heard' about it; see David Moss' reply), and these groups have only been around a decade or so. 

But I also imagine some pure age effects (as people age they leave/are less likely to enter), perhaps driven by things like 

1. Homophily/identity/herding: If you only see people unlike you (agewise) you're less likely to think you belong. This leads to inertia. 

2. Cost and family priorities: EA ~expects/encourages people to donate a substantial share of their income, or do directly impactful work (which may be less remunerative or secure). For older people the donation share/lost income could seem more substantial, esp. if they are used to their lifecycle. Or probably more significantly, for parents it may be harder to do what seems like 'taking money away from their children. 

3. Status and prestige issues: EA leaders tend to be young, EA doesn't value seniority or credentials as much (which is probably a good thing). But older people might feel ~disrespected by this. Or second order: they might think that their age-peers and colleagues will think less of them if they are following or 'taking direction' from ~'a bunch of kids'. E.g., as a jr. professor at an academic conference if you are seated at the grad students' table you might feel insecure.

4. Issues and expertise that are relevant tends to be 'new stuff' that older people won't have learned or won't be familiar with. AI Safety is the biggest one, but there are other examples like Bayesian approaches.

(Identity politics bit: I'm 48 years old, and some of this is based on my own impressions, but not all of it.)
 

I expect a lot of people might be attracted to something like PfG now on the following (not entirely utilitarian/EA grounds). This could be a potential catalyst.

1. ~Liberals and moderates in the US and abroad (people who support democratic institutions the rule of law, oppose DOGE shutting down USAID, the corruption, etc.) don't want to buy from any of the companies supporting Trump or ~bribing him. Amazon.com springs most to mind.

2. I think PfG is not about moral purity, it's more about utilitarian 'do the most good'. But companies that donate (most of) their profits to GH&D, environmental, or animal welfare charities are very unlikely to be Trump-backed oligarchs/cronies. In fact, it might be impossible for them to do so because their donations would code them as liberals/globalists/whatever. 

3. So it's a good proxy. Liberals+ could use PfG to find alternatives to Amazon etc.