tl;dr: Recently founded the AI & Democracy Foundation (AIDF) & we are hiring!. Two especially critical roles:.• Director of AI Strategy: Bring deliberative democracy to AI & AI to deliberation.• Head of Ops: Lead & own everything needed to grow the org. Details below.

RT @BKCHarvard: Unlocking Better Algorithms: Faculty Associates @random_walker and Elissa Redmiles, along with Affiliates @metaviv and @nat….

Most recent context: > "We need fire drills for AI spearphishing for everyone.".This claim is a bit intentionally provocative. But also a result of some justified frustration. There is so much we (govs, tech co, etc.) could have done, that we advocated.

Rant:.We need fire drills for AI spearphishing for everyone. ~Now. To train people. AI spearphish (drill) everyone with a phone, email address, WhatsApp account, etc. Every month. We've run out of time to secure the infrastructure. It's time to (try) to secure the people.

As we build more intelligent machines, it's worth thinking about the kind of intelligence and wisdom that we see in ourselves, and how we might bring more of that wisdom into those machines, and into our interactions with them & each other. This is just one tiny example of that.

Does it make sense to have one AI system, with a system prompt that pushes for this kind of wisdom, do extensive chain of thought analysis of the conversation, watching you converse with another AI system?.Is anyone building this?.When is it worth doing?.How can we triage that?.

I know people who are doing this to help make sense of conversations with others by pasting in e.g. text message logs. But what if the conversation itself is with an AI? Is there any difference from just continuing the conversation? Can you make it different by 'escaping' through.

Is it even coherent to think about receiving that kind of wisdom from current AI systems—if the conversation being observed is with an AI system? (Perhaps even the same kind?) What makes it more or less likely that you can get that quiet wisdom perspective?.

There is an incredible kind of interaction where someone listens quietly to a long conversation among others, and then drops 'wisdom bombs'. Getting to the heart of the matter, highlighting key insights, or untangling miscommunications. Can/should we have AI systems do that?.

Who is conducting ML-style evaluations for scaffolding designed for humans or groups of humans?. To be more precise, it seems that we can improve LLM performance on evaluations with scaffolding (caveats: this applies to some LLMs, some evaluations, and some types of scaffolding).

I'm especially interested in AI-focused outputs (tech, governance, etc.), and links to them, in addition to favorite aspects of the process.

What are your favorite many-org collaborations that created a complex output? (10+ orgs; 20+ people; especially interested in 50+) .This could be a many-author paper, a working group developing standards, a shared ecosystem roadmap, etc. How was it organized? What made it great?.

RT @democraticAI: We are hiring! We are far more likely to be able to address the challenges of AI if we invest in, develop, and adopt proc….

@Meta Even worse, consider the alternative to open models—permanently locking them behind APIs that can (hopefully🤞) mitigate harms. This could significantly reduce freedom and agency, and many kinds of *good* competition!.This also creates immense (dangerous?) power concentration.

@Meta But what if a model does end up presenting fairly significant risks, such that there is significant pressure *not* to share it openly?.How should we determine what that line is? (especially if it's somewhat ambiguous).And what prevents others from just ignoring that line?.

@Meta Can we get the best of openness, while mitigating potential risks?. One option is delayed release: ensure that there is a period of time where weights are not provided openly (but are usable via API). This gives time to learn about the potential unexpected risks of new models.

@Meta The FTC must navigate a tricky balance here — leaning into competition that serves consumers and the public interest, while being careful about competition that could significantly increase harm. Openness is irreversible, which is often great—but it could backfire.

@Meta Some kinds of competition can harm consumers and even critical public goods, including democracy itself. (Ideally, other agencies or Congress would help prevent that. ).Other kinds of competition can massively accelerate valuable innovation.

@Meta Competition can also lead to valuable innovations. but with significant dual-use impacts where the pros outweigh the harms. If you compete to create a helpful AI, and that AI is also very helpful to terrorists, who then deploy powerful bio weapons. that's rather awkward!.

@Meta For example, competing for people's scarce attention or for their food buying can be great—if the competition is over the quality of information and healthiness of food!. But in practice, competition can end up focused on persuasion, manipulation, and addiction.