AI alignment is rapidly scaling funding, and this means grantmakers will be stretched thin and less able to reliably avoid giving money to people who are not mission aligned, not producing useful research, or worst of all just want to extract money from the system. This has the potential to cause massive problems down the road, both by producing distracting low-quality research and by setting up expectations which will cause drama if someone is defunded later on while there's still a lot of money flowing to others.

An attack-resistant EigenKarma^[1]-like network for alignment researchers would, if adopted, allow the bottleneck of grantmaker time to be eased and quality of vetting to be improved, by allowing all researchers to participate in the process of vetting people and assessing the quality of their work.  The ideal system would:

• Allow grantmakers to view any individual's score with an arbitrary initial trusted group, so they could analyze how promising someone seems from the perspective of any subgroup, with a clean UI.
• Allow people to import their own upvote history from any of AF/LW/EAF to seed their outflowing karma, but adjust it manually via a clean UI.
• Have some basic tools to flag suspicious voting patterns (e.g. two people only channeling karma to each other), likely by graphing networks of votes.
• Maybe have some features to allow grants to be registered on the platform, so grantmakers can see what's been awarded already?
• Maybe have a split between "this person seems competent and maybe we should fund them to learn" vs "this person has produced something of value"?

Rob Miles has some code running on his Discord with a basic EigenKarma system, which is currently being used as the basis for a crypto project by some people from Monastic Academy, and could be used to start your project. I have some thoughts on how to improve the code and would be happy to advise.

I'm imagining a world where researchers channel their trust into the people they think are doing the most good work, which means that grantmakers can go "oh, conditioning on interpretability-focused researchers as the seed group, this applicant scores highly" or "huh, this person has been working for two years but no one trusted thinks what they're doing is useful" rather than relying on their own valuable time to assess the technical details or their much less comprehensive and scalable sense of how they think the person is perceived.

Obviously some pre-research would be to make a sketch of how it'd work and ask grantmakers and researchers if they would use the system, but I for one would if it worked well (and might provide some seed funding to the right person).

1. ^{^}

   This is Rob Miles's description of his code, I hear the EigenTrust++ model is even better, but have not read the paper yet to verify that it makes more sense here.

An easier means of starting up an EA-aligned, accredited private school.

As EA matures, there will be more and more parents. Kids of self-identified EAs are (going out on a limb) likely to be very smart and neuroatypical, and to struggle with the default schooling system .

As remote work options free up location choice, there could be major EA community-building gains from making it more straightforward to start an aligned micro school.

Speaking from personal experience, schooling considerations are a major limitation on my family's ability to relocate and get more intimately involved with the EA community. I'm sure we're not alone.

SchoolHouse is very much the kind of thing I'm thinking of. I have no idea what competition is like on the tech side of this, only that finding an actual school that works for my kids is still super challenging.

• It's listed in the list of software projects above, but I really think that Coordination Software (https://www.lesswrong.com/s/vz9Zrj3oBGsttG3Jh) has a lot of potential, if only somebody could find the right angle -- how to define the group that is voting to change its norms or take an action, what killer applications of the software or particular audiences would be a good place to start, etc...

• I think that remote-working software is very positive for civilization; allowing people to work on top tech projects while not living in exploitative NIMBY cities (or even living in the developed world at all!) will do a lot to increase economic growth and distribute more resources to poorer areas. Of course after covid everyone is working on this, but almost none of the efforts are targeted overseas. Combine Wave-style currency transfer with translation and other tools, and maybe the right legal framework to make it easy for US companies to hire/outsource to people abroad, and you've created a kind of "virtual immigration" allowing thousands of people in India or Latin America to work remotely and get paid western wages for working on software projects.

• There are a lot of good prediction market angles. One of the best remaining unexplored niches, IMO, is creating flexible software that corporations can use to set up their own internal prediction markets and forecasting tournaments (and ideally integrate those with other tools like their scrum boards and issue tracking and etc). See this link for more detail on my idea about creating "the Gitlab of prediction markets": https://forum.effectivealtruism.org/posts/dQhjwHA7LhfE8YpYF/prediction-markets-in-the-corporate-setting?commentId=eaLBFJJJgiDGY5QHM

Do these have to be software ideas to suit the technical skills of the CTO? I can imagine some more constructiony things that could potentially make money and be big:

• Become a real estate developer, start an affinity city / intentional community along the lines of Arizona's car-free 'culdesac' neighborhood, but experiment with new social and governing institutions like quadratic public goods funding, or urban design built around remote work, or something something crypto property rights, or liquid democracy for residents to make collective decisions, or an assurance contract to get people to move there initially... basically do a Charter City but as a suburban development within the USA. Get rich by owning the land that will hugely appreciate in value when the neighborhood does well. Hope that the successful institutions in your town will be adopted by larger cities / states / nations and thereby improve civilization's overall decisionmaking abilities.

• Make better ventilation / UV-light sanitation / PPE tech, then sell the ventilation tech to new buildings and sell the PPE to hospitals and national stockpiles.

• Get good at building especially deep, airtight, large bunkers that can serve the EA desire for long term civilizational resilience from pandemics, asteroid strikes, nuclear wars, etc. Sell your premium bunkers to governments and eccentric billionaires. (For more on these last two bullet points, see here: https://forum.effectivealtruism.org/posts/u5JesqQ3jdLENXBtB/concrete-biosecurity-projects-some-of-which-could-be-big-1)

BOAS (boas.co) is the world's first marketplace that donates all profits to save lives and we need a CTO to help us scale and get funding (we're pre-seed).We launched in January 2022 and we’ve grown to more than 1.500 products (almost 100% MoM) and growing our active users at around 50% MoM and we spent no money on advertising.You can the role and detailed tasks in our jobs page. You can apply there or send me a DM with any questions! Even if this is not you, please share the message with someone who might be a great CTO. Thanks in advance!

TL;DR: Subscribe to this comment to get notified about the most promising ideas

"the most promising" is according to my own judgement.

Please don't reply to this comment yourself, but feel free to start your own list

How to subscribe:

Click the "..." on the top right:

And then "Subscribe to comment replies":

Also, 

If you care about promising EA Software projects, consider getting in touch with me 

Even just to let me know you're a person that cares about this thing

Adding: This comment was requested by yet another potential EA CTO who I didn't know beforehand and am happy to hear from

Biological Safety Reporting System:

Despite the effectiveness of safety reporting systems (see, e.g., Aviation Safety Reporting System) at identifying weak points in safety protocols and reducing risks of accidents, there does not appear to be an analogous system for biological risks.  ^[1]

Desired Features: 

• Anonymous reporting
• Optional identity verification of reporters (for credibility and whistleblower protection)
   - Encription of reporter identity (secure multi-party computation?) for increased whistleblower security
• Submission of bioinfohazard sequences to a biological risk sequence surveillance organization (e.g., SecureDNA) 
• Encription of reports containing bioinfohazards

EA Advantages:

• Appropriate respect for infohazards and tail risks
• Synergy with efforts to strengthen the biological weapons convention

1. ^{^}

   The Federal Select Agent Program only applies to a short list of pathogens and does not apply to novel hazards.

A better Swapcard:

The current Swapcard  is buggy and imperfect. The world maybe could use a better one?

A few alternatives exist but they don't seem so good (if one of them does seem good, please tell me or CEA !).

Your first client would be the EA community, and you could continue to build role out your startup and/or put some effort into integrating with the EA forum, which would be wonderful too

Something I've considered making myself is a Slackbot for group decision-making: forecasting, quadratic voting, etc. This seems like it would be very useful for lots of organizations and quite a low lift. It's not the kind of thing that seems easily monetizable at first, but it seems reasonable to expect that if it provides valuable, it could be the kind of thing that people would eventually have to buy "seats" for in larger organizations.

Are they familiar with Charity Entrepreneurship? They research high-impact nonprofit ideas (which you can find on their website) and they have an incubation program 

Hello Yonatan and EA community, I am currently exploring the idea of a EA fundraising videogame, with 3 objectives: 1/fundraising for EA charities 2/Advertising for EA charities and 3/EA education.

I see it as a new way of enlarging the EA community but it may already have been done and I am simply unaware of it. Are there already EA video games designed by the community?

Would someone in the community having experience in building video games be motivated to get in touch with me and start a project?

A Decentralized Peer to Peer Donation/UBI Platform

--------

- Everything controlled automatically by smart contracts on the blockchain

- Donors donate crypto funds into a central treasury wallet on the blockchain

- A "smart contract" automatically distributes a percentage from this treasury direct to the wallets of verified recipients one-time or periodically (maybe weekly?)

- Recipients could be people in dire need, have medical bills or unable to fend for themselves.

- A limit could be placed on how long recipients are to keep receiving the payments (or whether one-time)

- Donors have a dashboard where they can monitor funds in the treasury and the payments made to recipients.

- All payments will be made using a green cryprocurrency and climate-friendly blockchain

- A fee can be taken from donations to cover platform maintenance.

- OPTIONAL: A DAO (Decentralized Autonomous Organization) to manage the project. People who hold a specific minimum amount of tokens will be able to make proposals and vote on issues concerning the project.

Funding Required: $39,000 > Time to delivery: 3 months

I believe, a system can be designed for distribution of money that lacks the most "problematic" features of UBI and streamlines the act of donations-making in general. The basic idea is to not give money away, but pay people for work - any work they want to do. Poetry writing, musical band practicing, gardening, picking trash,... They would have to report the work with a believable proof (photo, video, url,...) and have it approved by two other System-selected workers. They would receive a fixed amount of System Currency for every reported/approved hour. The Currency for the reward is created by the System at the moment the work gets approved. The Currency would be freely exchangeable for national currency (and back). A fixed exchange rate would be maintained by the donor and/or by inflow of national currency from other sources, like own economic activity of the workers in the System, investments into projects run by the workers in the System,... Products and services created in the System must be possible to purchase with the Currency. When purchased, the buyer pays the System, which at that moment destroys the Currency paid and awards the creator/seller of the purchased product with a numerical, reputation-like reward instead. The reputation reward is equal in amount to the paid Currency. A reputation score of a worker is non-transferable. Separation of the monetary and the skills/reputation-signaling rewards is the key to making the system both - equitable economically and efficient at the same time (workers have different reputation "signatures")*.

The described system exists in a proof of work-level implementation and has a number of other, non-listed features. It was originally developed to make volunteer products consumption and production fair, but it will be used also for the use case described above. It would be great to find somebody who really understands technology to make the system scalable. Also, it will be easier to get funding with a team that has more than one founder -  I think.

The POC runs at https://rovas.app and theoretical foundations are described in a paper I wrote about this.

* The separation of economic and reputation rewards might sound crazy, but it was suggested recently for different use case

Something like give well (evidence based, reasoning transparency) but…

• Rates a broader set of charities, not shying away from cross-cause and “US versus internationally operating” charity comparisons

• Explicitly models (calibrated) uncertainty

– makes a pitch to a broader audience, and prominently comparing impact ratings of popular charities To each other, and to more effective ones… To get people thinking about this and to get charities to offer better evidence and compete on this basis