Iâd be eager to see anything that speaks to Leverageâs past or present research activity: what have they been trying to find out, what have they achieved, and what are they aiming for at the moment (cf).
As you know from our previous conversations re. Leverage, Iâm fairly indifferent to âtheyâre shady!â complaints (I think if people have evidence of significant wrongdoing, they should come forward rather than briefing adversely off the record), but much less so to the concern that Leverage has an has achieved extraordinarily little for an organisation with multiple full-time staff working for the better part of a decade. Showing something like, âAh, but see! Weâve done all these things,â or, âYeah, 2012-6 was a bit of a write-off, but hereâs the progress weâve made sinceâ, would hopefully reassure, but in any case be informative for people who would like to have a view on leverage independent of which rumour mill they happen to end up near.
Other things Iâd be interested to hear about is what you are planning to work on at Leverage, and what information you investigated whichâI assumeâleads to a much more positive impression of Leverage than I take the public evidence to suggest.
Thanks for the message and for engaging at the level of what has Leverage achieved and what is it doing. The tone of your reply made me more comfortable in replying and more interested in sharing things about their work so thank you!
Leverage are currently working on a series of posts that are aimed at covering what has been happening at Leverage from its inception in 2011 up until a recent restructure this year. I expect this series to cover what Leverage and associated organisations were working on and what they achieved. This means that I expect Leverage to answer all of your questions in a lot more depth in the future. However, I understand that people have been waiting a long time for us to be more transparent so below I have written out some more informal answers to your questions from my understanding of Leverage to help in the meantime.
Another good way to get a quick overview of the kinds of things Leverage has been working on beyond my notes below is by checking out this survey that we recently sent to workshop participants. Itâs designed for people whoâve engaged directly with our content so it wonât be that relevant for people to fill in necessarily but it gives an overview of the kinds of techniques Leverage developed and areas they researched.
What did Leverage 1.0 work on?
A very brief summary is that the first eight and a half years of Leverage (letâs call this âLeverage 1.0â as a catch-all for those organisations before the restructure) was at first a prioritisation research project looking at what should people work on if they want to improve the world. Leverage 1.0 later came to focus more on understanding and improving people as their psychological frameworks and training tools developed but they still conducted a wide range of research.
This means that in the very early days they were thinking a lot about how to prioritise, how to make good long term plans and just trying a bunch of things. I get the impression that at this stage almost nothing was ruled out in terms of what might be worth exploring if you wanted to improve the world. This meant people investigating all sorts of things like technological intelligence amplification, nootropics, and conducting polyphasic sleep experiments. People might be researching what caused the civilisational collapse that led to the dark ages, the beliefs of a particular Christian sect, or what lead to the development of Newtonian physics. Leverage felt this was important for research progress. They wanted researchers to follow what motivated them. They thought that it was important to investigate a lot of areas before deciding where to focus their efforts because deciding what to prioritise is so important to overall impact. This felt particularly important when investigating moon-shots which had the potential to be extremely valuable even if they seemed unlikely at the outset.
Some of the outputs of these early days of research included training sessions on:
PlanningâHow to build and error-check plans for achieving your goals
Expert assessmentâhow to determine if someone is an expert in a given domain when you lack domain knowledge
Learning to learnâhow to improve and expand the scope of your learning process
Theorizingâhow to build models and improve your model building process over time
Prioritisation and goal settingâhow to find your goals, back chain plans from them etc
This is far from everything but gives you a flavour.
Geoff had developed a basic model of psychology called Connection Theory (CT) so this was a thing that was investigated alongside everything else. This involved spending a lot of time testing the various assumptions in CT.
Through experimenting with using CT in this way, Leverage eventually found they were able to use ideas from CT to make some basic predictions about individual and group behaviour, help individuals identify and remove bottlenecks so that they could self improve and perhaps even identify and share specific mental moves people were using to make research progress on particular questions. This made the team more excited about psychology research in particular (amongst the array of things people were researching) as a way to improve the world.
From there they (alongside the newly founded Paradigm Academy) developed some of the research into things like
one on one and group training,
A catalogue of different mental procedures individuals use when conducting research so that they could be taught to others to use to tackle different research and other problems. One example intellectual procedure (IP) just to give a sense of this, is the Proposal IP where you use the fact that you have a lot of implicit content in your mind and your taste response to inelegant proposals to speed up your thinking in an area.
Specific training in strategy, theorizing, and research
A collection of specific introspection and self-improvement techniques such as:
Self-alignment (a tool for increasing introspective access which handled a class of cases where our tools previously werenât working)
Anti-avoidance techniques (making it so you can think clearly in areas you previously didnât want to think about or had fuzzy thoughts in)
Charting (a belief change tool that has been modified and built out a lot since itâs initial release)
Mythos (tool for introspection with imagery, helpful for more visual people)
Integration and de-zoning (tools for helping people connect previously separate models)
What is Leverage doing now?
As for what Leverage is currently working on, once we have posted our retrospective weâll then be updating Leverageâs website to reflect its current staff and focus so again a better update than I can provide is pending.
The teaser here is that from the various research threads being pursued in the early years of Leverage 1.0, Leverage today has narrowed their focus to be primarily on two areas that they found the most promising in their early years:
Scientific methodology research
Psychology
We also continue to be interested in sociology research and expect to bring on research fellows (either full time or part of future fellowship programmes) focusing on sociology in the future. However, since weâre relaunching our website and research programme we want to stay focused so weâre punting building out more of our sociology work to further down the line.
The scientific methodology research involves continuing to look at historical examples of scientific breakthroughs in order to develop better models of how progress is made. This continues some of our early threads of research in theorising, methodology, historical case studies and the history of science. Weâre particularly interested in how progress was made in the earlier stages of the development of a theory or technology. Some examples include looking at what led to the transition in chemistry from Phlogiston to Lavoisierâs oxygen theory or the challenges scientists had in verifying findings from the first telescopes. We aim to share lessons from this research with researchers in a variety of fields. In particular, we want to support research that is in its earlier, more explorative stages. This is more of a moon-shot area but this means it can get less attention while being potentially high reward.
Our psychology research aims to continue to build on the progress and various research threads Leverage 1.0 was following. While this is quite a moon shot style bet, if we can improve our understanding of people then we potentially improve the ways in which they work together to solve important problems. At this point, we have developed tools for looking at the mind and mental structures that we think work fairly well on the demographics of people weâve been working with. I got a ballpark estimate from someone at Leverage that Leverage and Paradigm have worked with around 400 people for shallower training, and about 60 for in-depth work but treat those figures as a guess until we write something up formally. Weâve focused in the last few years on improving these tools so they work in harder cases (e.g. people who have trouble introspecting initially) and using the tools to find common mental structures. Moving forward with this research we want to test the tools in a more rigorous way, in particular by communicating with people in academia to see whether or not they can validate our work.
One thing I personally like about the plans for psychology research is that it also acts as a check on our scientific methodology research. If the insights we gain from looking at the history of scientific progress arenât useful to us in making progress in psychology then thatâs one negative sign on their overall usefulness.
Who works for Leverage and Paradigm?
The team is much smaller and the organisation structure slightly more defined (although there is a way to go here still). There are four researchers (including Geoff who is also the Executive Director) and Iâll be joining as a Program Manager managing the researchers and helping communicate with the public about our work. So four in total at the moment, five once I start.
While Leverage Research in its newer form is getting going it still receives a lot of help from its sister organisation Paradigm Academy. This means that while they are two separate organisations, currently Paradigm staff give a lot of time to helping Leverage, particularly in areas like operations and helping with PR and Communications like the website relaunch. This helps allow the researchers to focus on their research and means the burden of public communication wonât all fall on their newest employee (me). Once a lot of that is done though we expect to make the division between the two organisations clearer. Paradigm currently has nine employees including Geoff.
I expect all of this will generate more questions than it answers at the moment and while my answer is to wait for Leverageâs formal content to be published I can see why this is frustrating. I hope my examples give a small amount of insight into our work while we take the time to write things up. You have every reason to be sceptical about Leverage posting content given various promises made in the past. I think given our track record on public communication that scepticism is valid. All I can perhaps offer in the meantime is that I personally am very keen to see both the retrospective and the new Leverage website published and the get sh*t done spirit that you and others on this forum know me for is part of the reason theyâve offered me a job to help with this in the first place.
Why I chose to work at Leverage
As for my personal reasons for choosing to accept an offer from Leverage, I expect this to be hard to transmit just because of inferential distance. My decision was the result of at least five months of discussions, personal research and resultant updates all of which is built on various assumptions that caused me to already be pursuing the plans I was at CEA.
Iâll attempt a short version here anyway in case itâs helpful. If thereâs a lot of interest Iâll consider writing this up but Iâm not sure itâll be sufficiently useful or interesting to be worth the time cost.
Broadly speaking, I created a framework (building off a lot of 80Ks work but adapting it to suit my needs) to use to compare options on:
Impact - comparing potential career plans by looking at the scale and likelihood of success across:
the problem being tackled (e.g. preventing human extinction),
the approach to solving that problem (e.g. develop AGI in a way thatâs safe)
the organisation (e.g. DeepMind)
and what I personally could contribute (e.g. say in a role as Project Manager)
Personal happiness (personal fit with the culture, how the job would fit into my life etc)
Future potential (what skills would I build and how useful are they, and what flexible resources such as useful knowledge or network would I gain)
I decided that I was willing to bet at least a few more years of my career on the more moon shot type plans to build a much better future (something like continuing personally to follow CEAâs vision of working towards an optimal world).
This narrowed my focus down to primarily considering paths related to avoiding existential risks and investing in institutions or advances that would improve humanityâs trajectory. In exploring some options around contributing to AI safety in some way I came away both not feeling convinced that I wouldnât potentially cause harm (through speeding up the development of AGI) and less sure of the arguments for now being a particularly important hinge on this. It, therefore, seemed prudent to learn a lot more before considering this field.
This left me then both wanting to invest more time in learning more while also not wanting to delay working on something I thought was high impact indefinitely. In terms of impact, the remaining areas were advances or institutions that might improve humanityâs ability to tackle global problems.
Iâd had plenty of conversations with various people about Leverage (including many Leverage sceptics) in the past and interacted with Leverage and Paradigm directly to some degree, mostly around their introspection techniques which I personally have found extremely useful for self-improvement. I knew that they were interested in psychology initially as a potential way to improve humanityâs trajectory (but didnât yet understand the scope of their other research) so I reached out to chat about this. I found that many of the people there had already thought a lot about the kinds of things I was considering as options for improving the long-term future and they had some useful models. Those interactions plus my positive view of their introspection techniques led me to think that Leverage had the most plausible plan given my current uncertainty for improving the long-term future and was likely to be by far the best option for me in terms of self-improvement and gaining the knowledge I wanted for making better future plans. Their recent restructure, desire to establish a more structured organisation and plans to publish a lot of content meant they had an opening for my particular skill set and the rest, as they say, is history.
Leverage are currently working on a series of posts that are aimed at covering what has been happening at Leverage from its inception in 2011 up until a recent restructure this year.
This aged well⊠and it reads like what ChatGPT would blurt, if you asked it to âsound like a convincingly respectful and calm cult with no real output.â Your âAnti-Avoidance,â in particular, is deliciously Orwellian. âYouâre just avoiding the truth, youâre just confused...â
I was advocating algal and fish farming, including bubbling air into the water and sopping-up the fish poop with crabs and bivalvesâback in 2003. Spent a few years trying to tell any marine biologist I could. Fish farming took-off, years later, and recently they realized you should bubble air and catch the poop! I consider that a greater real-world accomplishment than your âtraining 60+ people on anti-avoidance of our pseudo-research.â Could you be more specific about Connection Theory, and the experimental design of the research you conducted and pre-registered, to determine that it was correct? Iâm sure youâd have to get into some causality-weeds, so those experimental designs are going to be top-notch, right? Or, is it just Geoff writing with the rigor of Freud on a Slack he deleted?
Hello Larissa,
Iâd be eager to see anything that speaks to Leverageâs past or present research activity: what have they been trying to find out, what have they achieved, and what are they aiming for at the moment (cf).
As you know from our previous conversations re. Leverage, Iâm fairly indifferent to âtheyâre shady!â complaints (I think if people have evidence of significant wrongdoing, they should come forward rather than briefing adversely off the record), but much less so to the concern that Leverage has an has achieved extraordinarily little for an organisation with multiple full-time staff working for the better part of a decade. Showing something like, âAh, but see! Weâve done all these things,â or, âYeah, 2012-6 was a bit of a write-off, but hereâs the progress weâve made sinceâ, would hopefully reassure, but in any case be informative for people who would like to have a view on leverage independent of which rumour mill they happen to end up near.
Other things Iâd be interested to hear about is what you are planning to work on at Leverage, and what information you investigated whichâI assumeâleads to a much more positive impression of Leverage than I take the public evidence to suggest.
Hi Greg,
Thanks for the message and for engaging at the level of what has Leverage achieved and what is it doing. The tone of your reply made me more comfortable in replying and more interested in sharing things about their work so thank you!
Leverage are currently working on a series of posts that are aimed at covering what has been happening at Leverage from its inception in 2011 up until a recent restructure this year. I expect this series to cover what Leverage and associated organisations were working on and what they achieved. This means that I expect Leverage to answer all of your questions in a lot more depth in the future. However, I understand that people have been waiting a long time for us to be more transparent so below I have written out some more informal answers to your questions from my understanding of Leverage to help in the meantime.
Another good way to get a quick overview of the kinds of things Leverage has been working on beyond my notes below is by checking out this survey that we recently sent to workshop participants. Itâs designed for people whoâve engaged directly with our content so it wonât be that relevant for people to fill in necessarily but it gives an overview of the kinds of techniques Leverage developed and areas they researched.
What did Leverage 1.0 work on?
A very brief summary is that the first eight and a half years of Leverage (letâs call this âLeverage 1.0â as a catch-all for those organisations before the restructure) was at first a prioritisation research project looking at what should people work on if they want to improve the world. Leverage 1.0 later came to focus more on understanding and improving people as their psychological frameworks and training tools developed but they still conducted a wide range of research.
This means that in the very early days they were thinking a lot about how to prioritise, how to make good long term plans and just trying a bunch of things. I get the impression that at this stage almost nothing was ruled out in terms of what might be worth exploring if you wanted to improve the world. This meant people investigating all sorts of things like technological intelligence amplification, nootropics, and conducting polyphasic sleep experiments. People might be researching what caused the civilisational collapse that led to the dark ages, the beliefs of a particular Christian sect, or what lead to the development of Newtonian physics. Leverage felt this was important for research progress. They wanted researchers to follow what motivated them. They thought that it was important to investigate a lot of areas before deciding where to focus their efforts because deciding what to prioritise is so important to overall impact. This felt particularly important when investigating moon-shots which had the potential to be extremely valuable even if they seemed unlikely at the outset.
Some of the outputs of these early days of research included training sessions on:
PlanningâHow to build and error-check plans for achieving your goals
Expert assessmentâhow to determine if someone is an expert in a given domain when you lack domain knowledge
Learning to learnâhow to improve and expand the scope of your learning process
Theorizingâhow to build models and improve your model building process over time
Prioritisation and goal settingâhow to find your goals, back chain plans from them etc
This is far from everything but gives you a flavour.
Geoff had developed a basic model of psychology called Connection Theory (CT) so this was a thing that was investigated alongside everything else. This involved spending a lot of time testing the various assumptions in CT.
Through experimenting with using CT in this way, Leverage eventually found they were able to use ideas from CT to make some basic predictions about individual and group behaviour, help individuals identify and remove bottlenecks so that they could self improve and perhaps even identify and share specific mental moves people were using to make research progress on particular questions. This made the team more excited about psychology research in particular (amongst the array of things people were researching) as a way to improve the world.
From there they (alongside the newly founded Paradigm Academy) developed some of the research into things like
one on one and group training,
A catalogue of different mental procedures individuals use when conducting research so that they could be taught to others to use to tackle different research and other problems. One example intellectual procedure (IP) just to give a sense of this, is the Proposal IP where you use the fact that you have a lot of implicit content in your mind and your taste response to inelegant proposals to speed up your thinking in an area.
Specific training in strategy, theorizing, and research
A collection of specific introspection and self-improvement techniques such as:
Self-alignment (a tool for increasing introspective access which handled a class of cases where our tools previously werenât working)
Anti-avoidance techniques (making it so you can think clearly in areas you previously didnât want to think about or had fuzzy thoughts in)
Charting (a belief change tool that has been modified and built out a lot since itâs initial release)
Mythos (tool for introspection with imagery, helpful for more visual people)
Integration and de-zoning (tools for helping people connect previously separate models)
What is Leverage doing now?
As for what Leverage is currently working on, once we have posted our retrospective weâll then be updating Leverageâs website to reflect its current staff and focus so again a better update than I can provide is pending.
The teaser here is that from the various research threads being pursued in the early years of Leverage 1.0, Leverage today has narrowed their focus to be primarily on two areas that they found the most promising in their early years:
Scientific methodology research
Psychology
We also continue to be interested in sociology research and expect to bring on research fellows (either full time or part of future fellowship programmes) focusing on sociology in the future. However, since weâre relaunching our website and research programme we want to stay focused so weâre punting building out more of our sociology work to further down the line.
The scientific methodology research involves continuing to look at historical examples of scientific breakthroughs in order to develop better models of how progress is made. This continues some of our early threads of research in theorising, methodology, historical case studies and the history of science. Weâre particularly interested in how progress was made in the earlier stages of the development of a theory or technology. Some examples include looking at what led to the transition in chemistry from Phlogiston to Lavoisierâs oxygen theory or the challenges scientists had in verifying findings from the first telescopes. We aim to share lessons from this research with researchers in a variety of fields. In particular, we want to support research that is in its earlier, more explorative stages. This is more of a moon-shot area but this means it can get less attention while being potentially high reward.
Our psychology research aims to continue to build on the progress and various research threads Leverage 1.0 was following. While this is quite a moon shot style bet, if we can improve our understanding of people then we potentially improve the ways in which they work together to solve important problems. At this point, we have developed tools for looking at the mind and mental structures that we think work fairly well on the demographics of people weâve been working with. I got a ballpark estimate from someone at Leverage that Leverage and Paradigm have worked with around 400 people for shallower training, and about 60 for in-depth work but treat those figures as a guess until we write something up formally. Weâve focused in the last few years on improving these tools so they work in harder cases (e.g. people who have trouble introspecting initially) and using the tools to find common mental structures. Moving forward with this research we want to test the tools in a more rigorous way, in particular by communicating with people in academia to see whether or not they can validate our work.
One thing I personally like about the plans for psychology research is that it also acts as a check on our scientific methodology research. If the insights we gain from looking at the history of scientific progress arenât useful to us in making progress in psychology then thatâs one negative sign on their overall usefulness.
Who works for Leverage and Paradigm?
The team is much smaller and the organisation structure slightly more defined (although there is a way to go here still). There are four researchers (including Geoff who is also the Executive Director) and Iâll be joining as a Program Manager managing the researchers and helping communicate with the public about our work. So four in total at the moment, five once I start.
While Leverage Research in its newer form is getting going it still receives a lot of help from its sister organisation Paradigm Academy. This means that while they are two separate organisations, currently Paradigm staff give a lot of time to helping Leverage, particularly in areas like operations and helping with PR and Communications like the website relaunch. This helps allow the researchers to focus on their research and means the burden of public communication wonât all fall on their newest employee (me). Once a lot of that is done though we expect to make the division between the two organisations clearer. Paradigm currently has nine employees including Geoff.
I expect all of this will generate more questions than it answers at the moment and while my answer is to wait for Leverageâs formal content to be published I can see why this is frustrating. I hope my examples give a small amount of insight into our work while we take the time to write things up. You have every reason to be sceptical about Leverage posting content given various promises made in the past. I think given our track record on public communication that scepticism is valid. All I can perhaps offer in the meantime is that I personally am very keen to see both the retrospective and the new Leverage website published and the get sh*t done spirit that you and others on this forum know me for is part of the reason theyâve offered me a job to help with this in the first place.
Why I chose to work at Leverage
As for my personal reasons for choosing to accept an offer from Leverage, I expect this to be hard to transmit just because of inferential distance. My decision was the result of at least five months of discussions, personal research and resultant updates all of which is built on various assumptions that caused me to already be pursuing the plans I was at CEA.
Iâll attempt a short version here anyway in case itâs helpful. If thereâs a lot of interest Iâll consider writing this up but Iâm not sure itâll be sufficiently useful or interesting to be worth the time cost.
Broadly speaking, I created a framework (building off a lot of 80Ks work but adapting it to suit my needs) to use to compare options on:
Impact - comparing potential career plans by looking at the scale and likelihood of success across:
the problem being tackled (e.g. preventing human extinction),
the approach to solving that problem (e.g. develop AGI in a way thatâs safe)
the organisation (e.g. DeepMind)
and what I personally could contribute (e.g. say in a role as Project Manager)
Personal happiness (personal fit with the culture, how the job would fit into my life etc)
Future potential (what skills would I build and how useful are they, and what flexible resources such as useful knowledge or network would I gain)
I decided that I was willing to bet at least a few more years of my career on the more moon shot type plans to build a much better future (something like continuing personally to follow CEAâs vision of working towards an optimal world).
This narrowed my focus down to primarily considering paths related to avoiding existential risks and investing in institutions or advances that would improve humanityâs trajectory. In exploring some options around contributing to AI safety in some way I came away both not feeling convinced that I wouldnât potentially cause harm (through speeding up the development of AGI) and less sure of the arguments for now being a particularly important hinge on this. It, therefore, seemed prudent to learn a lot more before considering this field.
This left me then both wanting to invest more time in learning more while also not wanting to delay working on something I thought was high impact indefinitely. In terms of impact, the remaining areas were advances or institutions that might improve humanityâs ability to tackle global problems.
Iâd had plenty of conversations with various people about Leverage (including many Leverage sceptics) in the past and interacted with Leverage and Paradigm directly to some degree, mostly around their introspection techniques which I personally have found extremely useful for self-improvement. I knew that they were interested in psychology initially as a potential way to improve humanityâs trajectory (but didnât yet understand the scope of their other research) so I reached out to chat about this. I found that many of the people there had already thought a lot about the kinds of things I was considering as options for improving the long-term future and they had some useful models. Those interactions plus my positive view of their introspection techniques led me to think that Leverage had the most plausible plan given my current uncertainty for improving the long-term future and was likely to be by far the best option for me in terms of self-improvement and gaining the knowledge I wanted for making better future plans. Their recent restructure, desire to establish a more structured organisation and plans to publish a lot of content meant they had an opening for my particular skill set and the rest, as they say, is history.
Did this series end up being published?
Just wanted to say I super appreciated this writeup.
Thanks Raemon :-) Iâm glad it was helpful.
This aged well⊠and it reads like what ChatGPT would blurt, if you asked it to âsound like a convincingly respectful and calm cult with no real output.â Your âAnti-Avoidance,â in particular, is deliciously Orwellian. âYouâre just avoiding the truth, youâre just confused...â
I was advocating algal and fish farming, including bubbling air into the water and sopping-up the fish poop with crabs and bivalvesâback in 2003. Spent a few years trying to tell any marine biologist I could. Fish farming took-off, years later, and recently they realized you should bubble air and catch the poop! I consider that a greater real-world accomplishment than your âtraining 60+ people on anti-avoidance of our pseudo-research.â Could you be more specific about Connection Theory, and the experimental design of the research you conducted and pre-registered, to determine that it was correct? Iâm sure youâd have to get into some causality-weeds, so those experimental designs are going to be top-notch, right? Or, is it just Geoff writing with the rigor of Freud on a Slack he deleted?