Note 2024: Now that I am focusing on Sentinel I have zero apetite for new consulting projects.
Shapley Maximizers is a niche estimation, evaluation and impact auditing consultancy run by myself, Nuño Sempere, but increasingly also with support from collaborators. This page presents our core competencies, consulting rates, description of ideal clients, testimonials, and a few further thoughts.
In short, we are looking to support people and smallish organizations who are already producing value, and who want to add clarity and improve their prioritization through estimation, measure and good judgment. You can reach out to to nuno.semperelh@protonmail.com.
Core competencies
Research
Some past research outputs:
- Incentive Problems With Current Forecasting Competitions.
- Hurdles of using forecasting as a tool for making sense of AI progress
- Real-Life Examples of Prediction Systems Interfering with the Real World (Predict-O-Matic Problems)
- A prior for technological discontinuities
- Better scoring rules
- Prediction Markets in The Corporate Setting
- A concern about the “evolutionary anchor” of Ajeya Cotra’s report on AI timelines..
See here for more past projects.
Red-teaming
Do you want my most disagreeable self to poke holes in your proposal or idea? I’m happy to do this.
Programming
Do you want me to build a website or tool? This is something I’d be excited to do more of. For a larger project, see metaforecast, for smaller projects see this EA forum frontend, shapleyvalue.com, this tool to fit a beta distribution to a confidence interval, or this tool to compute proportional approval voting results for cases with more than one candidate. You might also want to check out my Github.
Evaluation and estimation
Do you want me to evaluate a project, estimate its value, and suggest ways you could do better? I am very happy to do this. Some past examples are:
- A Critical Review of Open Philanthropy’s Bet On Criminal Justice Reform
- Shallow evaluations of longtermist organizations
- 2018-2019 Long-Term Future Fund Grantees: How did they do?
- External Evaluation of the EA Wiki.
- An experiment eliciting relative estimates for Open Philanthropy’s 2018 AI safety grants.
- Unflattering aspects of Effective Altruism
For a smaller example, in the past I’ve really enjoyed doing subjective estimates of the value of different career pathways (1, 2).
Forecasting
I am happy to host workshops, or advise on tournament or forecasting platform design. If you are looking for specific forecasts, you probably want to hire Samotsvety instead, which can be reached out at info@samotsvety.org.
Advice to individuals
On occasion, I’ve enjoyed talking with entrepreneurial individuals about how they could make better career decisions, or better prioritization in the work they are already doing. In some cases this has lead to ongoing engagements. If you have a need for incisive advice on your career, feel free to reach out.
Rates
| Size of project | Cost | ~hours | Example |
|---|---|---|---|
| One-off (discounted) | $50 | 1h | You, a student or an early career person, talk with me for an hour about a career decision you are about to make, about a project you want my input on, etc. |
| One-off (well-off professional) | $200 | 1h | You, a well-off proffessional, pick my brain about a project for an hour. |
| One-off (titan of industry) | $500 | 2h | You, a titan of industry, pick my brain about a project for an hour about a topic of your choice, before which I spend an hour of preparation. |
| Forecasting workshop workshop | $1.5k | 2h | I organize a two hour workshop and give it to your underlings. Online, otherwise cost of travel not included. |
| Focused | $2k | 10h | Research that draws on my topics of expertise, where I have already thought about the topic, and just have to write it down. For example, this Bayesian adjustment to Rethink Priorities |
| Major | $20k | 100h | An evaluation of an organization, many shallow evaluations, or an early version of metaforecast, two editions of the forecasting newsletter |
| Custom | reach out | 1000h | Large research project, an ambitious report on a novel topic, the current iteration of metaforecast |
A few notes:
- Something that has worked well for me in the past has been to open a tab on an initial engagement, and then bill the client later, either when I’ve done more work for them, or when I set some dedicated time to do operations more efficiently.
- Each engagement has some negotiation, preparation and administrative burden. This means that I prefer longer engagements.
- Clients not of the landed gentry are hesitant to pay for an hour of advice (+20 mins negotiation, +10 mins bookkeeping, +1h of prep, +chance of a rescheduling, +more things that make the cost of an engagement add up). This means that even the well-off professional price is a bit discounted.
Description of client
My ideal client would be someone or some organization who is producing value in the world, and which wants me to identify ways they could do even better in an open-ended way. Because this context would be assumed to be highly collaborative, they would have a high tolerance for disagreeableness. A close second would be someone who is making an important decision commissioning me to estimate the value of the different options they are considering.
My anti-client on the other hand would be someone who has already made up their mind, and who wants me to rubber-stamp their decisions, like a grifty crypto project1 asking for help writting up a grant to an EA grantmaker, or someone commissioning an evaluation that they will ignore, or a forecast that doesn’t feed into any decisions.
Testimonials
I reached out to Nuño in mid-2021 because I was impressed by his “Shallow evaluation of longtermist organizations”, and wanted him to conduct an evaluation of the EA Wiki, of which I was at the time the editor. I was very pleased with the rigour and thoroughness of the analysis he produced, and would recommend his services as a project evaluator and forecaster unreservedly. Indeed, I can think of very few other people in this area whom I would endorse as enthusiastically as I do Nuño.
—Pablo Stafforini
Nuño has depth of knowledge and a track record to prove it. His work spans AI, nuclear risk and a myriad other topics, which I have consistently found insightful. A generalist of the highest calibre. We hired Nuño to review a complex article about AI and forecasting. He delivered promptly, and we found his work insightful.
—Jaime Sevilla
Nuño took on an extremely underspecified and wicked problem and delivered to everyone’s satisfaction. If you’re reading this page you probably already know about his relentless clarity, but you might not be aware of how much he cares about the work and how good at reading people he is.
—Gavin Leech
Further details
- I am very amenable to taking on projects that would require more than one person, because I am able to bring in collaborators. I would discuss this with the potential client beforehand.
- My thoughts on unpaid work are: I’m pretty averse to it. I understand that there are high transaction costs for small engagements. But there is infinite demand for me working for free. I will very occasionally do it, particularly around mentorship, personal writting, or cool open source projects. If you are easily googleable as a millionnaire, I don’t get why you’d ask me to do unpaid work (counter: II; a perspective would be welcome). Working for free has for me like a pretty high bar (currently: setting up an emergency response team). Many projects don’t meet my bar for promisingness when unpaid, but they do meet my bar when paid such that I can dedicate the ressources I earn to projects of my choice (e.g., that emergency response team).
- Note that I don’t think all crypto projects are grifty, and I in fact view crypto as a pretty promising area of innovation. It’s just that for the last couple years if you wanted to grift, crypto was a good place to do so. And in fact a crypto project that wanted to figure out how to produce more value in the world and capture some fraction of it could be a great client.↩