The Unabomber manifesto talked about the slave becoming the master, when the master becomes totally dependent on it – everything is done by the slave, & the master doesn’t know anymore how to sustain himself. This is a very strong argument & I remember seeing an artist who spent his life building everything by himself (including his microwave). I think that regardless of this argument, the most fulfilling, satisfying & self-actualizing life is a life where you do things by yourself. When you have thousands of items to go through, or a large text to read, or a large house to clean, the best thing would be to lay out the thousands items on the floor or wall & go over all of them, to sit & read the large text, or roll your sleeves & clean the large house. Do you want to save time? Time is everything you’ve got.

Yes, using AGI we could accelerate & empower science/engineering to be able to cure diseases or colonize space, but how would we feel about it? The joy & satisfaction Watson & Creek had discovering the structure of DNA was so strong that you can still feel & enjoy it even today (reading the Double Helix), how would we feel if we outsource it to computer programs? Teachers get so much joy in teaching wonderful kids & getting their gratefulness years afterwards – how would we feel if we let computer programs teach all of our kids?

Max Tegmark describes our situation as if we’re the Neanderthals creating the Homo Sapience to replace us. I enjoy my life as Neanderthal & would like to preserve it, with all of the *work* that gives us joy & satisfaction, even if it’s laborious & time consuming. We can achieve Ikigai, by living simple lives, doing the work that gives us joy & help others.

Tegmark also defines Life 3.0 as the life forms that can change their hardware. This reminds me the state of humans in the last act of Shaw’s “Back to Methuselah”, where humans born from eggs spend the hundreds of years they live doing Math & Art, while growing more hands if needed. I can see a far future of sentient beings evolving to that, but I don’t think that offloading all work & thinking to computer programs will get *us* there.

Humans generate words (& actions) after they generate emotions that are driving the words (& actions). This theory is based on self-introspection & also on Heidt’s metaphor of the Elephant & the Rider: we 1st unconsciously feel some emotion or urge (the “Elephant”), which is the primary driver for our behavior, & only then “rationalize” that emotion/behavior with words (the “Rider”). The Elephant is feeling something, which causes us to act in some way, & the Rider verbalizes that & come up with explanations why we did it.

OpenAI’s founder Ilya Sutzkever claims that in order to predict the next token, AI needs to predict the world leading to that token. That sounds really powerful, but at least we can consider the idea that in order to predict the next token of a human, AI needs to predict/model/understand the emotions behind the words.

Why do we need AI alignment? Because we assume AI will either take or be given the power to shape human society. So, we need to ensure that AI will share our values & do what’s good for us, just as with the humans that are given the power to shape our society. With humans given that power, what it usually means is that we want them to have similar driving emotions (moral feelings) as we have.

Having recently read Heidt’s “Righteous Mind” book, I can try to briefly explane what I mean by “driving emotions”. Basically, Heidt claims that our moral decisions & actions are driven by feelings (such as, disgust & compassion), which are in turn based on some innate moral foundations, similar to taste buds. Each moral foundation, is a feeling triggered by some modules. For example, the “Care” foundation, can be triggered by seeing a crying baby. Based on people’s upbringing & surrounding culture, each of us may have different range of triggers for their moral foundations, which will impact a person’s moral & political values. Liberal people will have more triggers for the “Care” & “Liberty” foundations, while conservatives will have more triggers for the “Loyalty” & “Authority” foundations.

Back to the questions of what’s needed for AI Alignment, it seems like in order to ensure an AI shares our values & does what’s good for us, we need to ensure the AI’s actions & words are driven by the same emotions (moral foundations) as we are. (This BTW is in complete opposite from what a US senator once gave budget to – a project to research how to create an AI without emotions..). Why? because our values – the things we won’t compromise on – are precisely determined by the moral feelings we have, which are determined by which modules trigger our moral foundations.

Here’s the weird thing though: we humans are also an example of “AI”, whose words & actions are driven by feelings, but in the same way that no one understands how AGI candidates work exactly, we also don’t understand how are feelings (the Elephant within us) exactly work. As much as scientists try, I don’t think we understand the Elephant so much, neither what it does nor how it works. It seems to be triggering different feelings, such as disgust, compassion & team belonging, which are the real driving forces behind our actions & decisions. & it seems to be tuned differently with different range of triggers in different humans.

Should we 1st create a model of the Elephant, instead of the Rider? The next token prediction is a model of the Rider & not the Elephant, although we assume it “contains” a model of the Elephant in order to work. Will a model of the Elephant allow us to better align or at least determine the alignment with an AI? Obviously, this is a huge challenge – we have a lot of datasets of the Rider outputs, but very little fMRI scans showing the only evidence of the Elephant outputs. How also would the model of the Rider (predicting next token), which we don’t exactly understand, be interacting with an imposed model of the Elephant, which it may already implicitly contain.

When I was very young, I thought about this problem, & initially wanted to create a computer program that models & measures morality, so that AI can be controlled by it. But then I thought that only something much deeper, such as a religion, could really drive AI to behave ethically. The idea that we need to endow moral feelings into an AI, using the mechanism of modules triggering the moral foundations, could be a step in that direction.

Photo by Andrea De Santis on Unsplash

But will power is also something that underlies all of our behavior & choices, not just the cases where we need to control our urges.

Years ago I read the metaphor or model that I think is most useful for understanding our will power, & even all of our behavior & choices. It’s not a new concept, Aristotle & Freud had similar models, but this model is just so clear & effective to understand, that I wish more people would know about it.

I read it in Jonathan Heidt’s book “The Happiness Hypothesis”, & it basically says the following: each one of us has in their head an elephant and a rider riding it. The elephant can’t speak, but basically decides how we’ll behave. The rider can speak, but doesn’t really control the elephant. When we speak with a person, or when a person speaks with themselves, we’re talking with the rider, not the elephant. So, when you ask someone why they behaved in a certain way, they actually don’t have a clue, & can only rationalize or think of a plausible explanation, as if they were the elephant’s lawyer.

If you think about it, this model explains so beautifully will power in everyday life, but more generally all of our behavior & choices. We may tell ourselves that we’ll wake up early tomorrow & stick to our diet, but that’s just the rider blurb & only the elephant will do what they feel like tomorrow. You can ask someone why they votes to a certain candidate, & they’ll give you a detailed explanation, but it wasn’t their choice, but rather the silent elephant’s.

There are tricks the rider can do to try influence the elephant’s behavior, for example, setting up the world in a way that removes temptation, or setting an habitual routine, but those are just tricks, & not a real control. I find that the rider will sometimes have a different agenda, for example a noble reason to wake up early in the morning, but in order to influence the elephant to actually get up from bed, the rider will need to place some carrot (not noble at all) if they want to make the elephant to do what they want.

So to understand human behavior, you need to somehow understand the elephant’s behavior, & to change your behavior you need to find tricks to influence the elephant. But maybe I’ll think differently after I’ll read the next book on my queue (“The Power of Now”) which may suggest that you should rather stop controlling the elephant & trying to be something else. Ignore the talking of the rider that makes you lose sight of the simple life of being your real self – the elephant.

Why
• Why would you like to do such a thing?
• Many times in human society evolution, some people suggested & dreamt about alternative ways to organize & live, but alas the socio-economic constrains of society required them to get the permissions of the people controlling society, which would obviously object to any change that may risk their control & power. In most cases, people would be prosecuted for just suggesting something that deviates from the current norms & believes
• Suggesting that people can also become part of the government, or that all humans are equal, or that women can vote &c, would cause the people suggesting them to be prosecuted or highly ridiculed at best. In any case, you always needed some permission from the people in control to try any change to the current rules.
• As human psychology advanced, it acquired the meta-capability of understanding that the rules a certain community has are based on their imagined stories & myths, which means that different communities have different perspectives & rules, but none of them is the “correct” one, they’re just like different animals trying different phenotypes & seeing which fits the environment better
• It seems that now, it might be possible for people to imagine alternative ways to live & organize, & actually try them out in practice, without the permission of the people in control
• Nation states have many rules governing how people live & how their economy is working to provide everyone the means of living & satisfy their needs
• However, Metaverses & Web3 now offer humans a way to try out alternative ways of living, that don’t require any permission from any nation state – a community could come up with its own rules, governing how people live & how their economy is working
• With the store of value in Web3, & the permission-less governence systems of DAO’s, people can actually run large communities & actually make a living inside them, without asking any Nation state authority for permissions
• Imagine of through out history, people could just imagine an alternative society, & just build it & test it in practice. Others could immerse themselves in it, & change their perspective in case they see that the alternative is much better. Wouldn’t human evolution might have been not only faster, but yielding much better models & believes?

How
• fromTeal is designed around neighborhoods, each clustering together organizations & teams working towards a similar purpose. Anyone can start an organization in fromTeal & apply to join one of the neighborhoods having a purpose similar to theirs
• fromTeal can also have cities & states, that try out certain governance rules, based on shared stories & myths. The same way people can apply to join a neighborhood, they could also apply to join a neighborhood in a different city or state, having stories & myths that resonate with them better
• In order to break away from the chains of the rules, laws, stories & myths we currently have in our current IRL society, fromTeal can position itself as an attempt to “live in the future”
• By just moving the clock to 100 years from now, people would have the freedom to come up with completely different stories & myth underlying the community they’re living & working in
• In the movie “La belle vert”, an alternative human society is depicted, living in a different planet, that spends most of its time doing sport in nature & growing produce which they share with each other
• They have technology, which they only use to sometimes visit other planets to learn, & to communicate with others remotely
• It is very optimistic vision on how a human society can live 100 years from now. Can we use fromTeal to test living & working in such society today?

I haven’t read the book yet, but from the explanation in the interview, it sounds like the new model of the brain is based on thousands of small brains, capable of creating models of the world, which are basically maps with reference frames (coordinates), similar to a building model that an architect creates on his computer. Hawkins’ theory is that animals 1st created this ability to create maps of the environment in their brain, and then took that capability to create maps of other things as well, including anything that we humans think about. The theory goes on to explain how that works on the neuron level, and how we have a singular perception which is the result of voting between the thousands of tiny modeling sun-brains.

For me that’s so fascinating, because I’ve been working for some time on an open-source framework & protocol for converting & representing any information stream or system into a 3D organized space interface. My assumption was that this will make our ability to find, discover, consume & work with information much more effective. I wrote about it in this blog several times. Recently I’ve made nice progress with new version of the reference bldg server & client.

The new theory of the brain that basically suggests that the brain models knowledge using the same mechanism for creating maps of the environment, and the fact that these models & maps are hierarchical, just like the buildings in my framework, is making me feel more strongly that 3D organized spaces are the best interface for consuming & working with information. Of course this is just assumption & intuition, which need to be experimented with & proven. But this is really exciting & fascination for me.

I’m old enough to remember Jeff Hawkins previous book about the brain, which had a different model of hierarchical temporal memory, which I found fascinating & inspiring at the time, but for my personal research, I preferred the idea of 3D organized spaces for representing information, which now sounds somewhat grounded with the underlying mechanisms of our brain.

• 1st concept is Ikigai: the ideal of achieving your life’s purpose or reason for being (read more here)
  • It is commonly described as the combination of 4 conditions:
    1. You work on what you love
    2. You are good at what you’re doing
    3. The world needs what you’re doing
    4. The world pays you for what you’re doing
  • Many people achieve conditions 2-4, but not many achieve the 1st – work on what you love.
  • Ask yourself: if you won the lottery tomorrow, would you continue to work at the same place afterwards? If not, than you haven’t achieved the 1st condition
  • Why is the 1st condition so difficult?
    1. Many people start off their career doing one thing, but find out later that their passion is actually doing something else
    2. Many people are working on “bullshit jobs” that aren’t bringing any real value to anyone, but they keep doing it just because of the status game, social pressure or fear of change
    3. Many people do work in the profession they love, but deeply need the freedom to work on exactly what they love – to build things that are meaningful for them & allow them to express their creativity (fulfilling Maslow’s top need of self-actualization)
• 2nd concept is Teal: an hypothesis of how organizations & employees can be 10-100x more productive, effective, successful & happy (learn more here)
  • It is based on 3 principles:
    1. Work should be done in purpose-driven teams, where everyone are trying to achieve some north-star purpose they’re all passionate about
    2. Work should be done in small self-managed teams, without any managers within or outside telling people what to do
    3. Everyone should be their real self at work, without putting on any mask or pretend
  • This sounds crazy & impossible but actually many people (especially engineers) already had a chance to see a proof that it actually works:
    1. Hackathons teams are usually x100 better than regular teams, just because they are following Teal
    2. Startup founding team are usually x100 better, until the point they get bigger & hierarchical (i.e., not Teal)
    3. Open-source projects & organizations do so much more with less resources, by practicing Teal
  • There are also companies showing how much Teal is effective, even too effective..
    1. Patagonia complaining that they’re making too much money
    2. Zappos turning down Amazon offers for billions of dollars because they’re so happy
    3. Burtzorg taking over Netherland healthcare & going global
    4. Morning star making almost all of US tomato based products
    5. Favi, Sun Hydraulics, Gore making best products in their categories for decades

fromTeal is trying to make as many people reach Ikigai, by leveraging the principles of Teal

• It’s a blockchain-based community, marketplace & metaverse, that lets people create a team working on the thing they love
• Once they work on what they love, they get help in achieving the 3 other Ikigai conditions
  • Being good at what they’re doing
  • Ensuring the world needs what they’re doing
  • Getting paid for what they’re doing
• This is done by removing the constraints between the teams & achieving Ikigai:
  • Time & resources to become good at what they’re doing
    1. Find more team members driven by the same purpose & passion, & share the work with them
    2. Get knowledge & guidance from other teams doing the same thing
    3. Get investment from people who need or share the same purpose as the team
    4. Get feedback until they become really good in what they’re doing
  • Ensuring the world needs what they’re working on
    1. Find people who really need the solution
    2. Design the solution together with these people
  • Get paid for what they’re building
    1. Enable teams building the solution to directly interact with people using it & sell it without any middleman
    2. Facilitate the commerce in a way that ensures trust & security
• How does fromTeal provide all that?
  • It’s a marketplace, based on a digital currency, for exchanging these goods:
    1. Team member referrals – get paid for referring someone to a team based on shared purpose
    2. Coaching sessions – get paid for providing knowledge & guidance based on relevant experience
    3. Feedback sessions – get paid for providing feedback on teams products or services
    4. Customer referrals – get paid for referring someone who needs what the team is building
    5. Virtual land & offices – get paid for selling metaverse locations for teams HQ
    6. Team created goods – get paid for building products & services for other teams
  • It’s a metaverse where teams form neighborhoods in a virtual city based on purpose – what they’re trying to change in the world
    1. Teams with similar purpose have their virtual HQ next to each other, allowing them to discover, learn from & help each other
    2. People who care about some purpose or have a need for it, can discover teams, join them to provide feedback or buy their products
    3. There’s a reason why certain cities are hubs for technology, research, art or commerce: having many similar companies/organizations close to each other, allows people to discover others in a similar situation, exchange ideas & create a network
    4. There’s a reason why stores of the same products are usually located in certain areas, close to each other – they want to learn from each other as well as get to the same customers
    5. Since all teams in a neighborhood are driven by the same purpose, they aren’t in a real competition – they care about the purpose just as much as being successful
  • It’s a community driven by the common goal of helping all members reach Ikigai
    1. Every team has a visual indicator of where it stands in the path to Ikigai, & what are their current needs, allowing other teams & people to help them
    2. The ownership of the community is by the community (DAO), so it provides grants to teams for building tools & solutions to help the community
  • It’s an experiment in using technology to enable new ways of working
    1. Using blockchain DAO’s for each team, to enable autonomy & self-management, fair equity & value sharing, decentralized decision making based on advice, role-based salaries &c
    2. Using new blockchain protocols to enable decentralized commerce & investing – no need for middleman platforms, managers or shareholders cutting a part in the value flow between teams & their customers. Customers buy products directly, & can also buy equity tokens & become part of the team.
    3. Using AI-based assistant devices to help each team manage its work, monitor progress & get recommendations & referrals across teams
    4. Using 3D metaverse to enable discovery & collaboration that is much more powerful than the page-based interface of websites & apps

I have 4 such contracts with myself, of things that I want to try change in the world, that I’d like to share with you.

Considering Stephen Covie’s circles of influence, it is also very stupid to want things that are beyond your circle of influence. Still, I want to believe that technology today makes the world smaller than what it used to be, which may mean that your circle of influence might be larger.

Besides wanting these changes in the world, I’m also trying to do something about them. For 3 of these 4 changes, I’ve started 2 projects to try move the needle in their direction.

Here are the 4 things:

1. I want every person to be able to work on her calling – the thing she wants to contribute to the world & change in the world, aka: self-actualize. Many of us feel strongly that we have some high purpose, a talent or passion or some desire to create some unique value for others. Unfortunately, most of us are busy with survival & never get to work on our real calling. In order to make a living, we end up working in a company building generic impersonal product, instead of building the personal unique product that we desire to build & give to the world.

2. I want companies to be flat – have no managers. Hierarchy is not only immoral – no man is more important than others to allow him to buy others & tell them what to do, as if they were disposable resources – it is also an extremely ineffective way to do great things together – you only get a fraction of people’s potential if you take away their freedom & tell them what to do. As a simple experiment to prove it, go to any Hackathon event & see for yourself flat teams being orders of magnitude more productive, creative & happy in their work, just because they have no managers.

3. I want people to change the way they talk to each other: each of us should always say what we feel & what we need, listen to what the others feel & need, and then think together how to satisfy the needs of everyone to make all feel better. This simple method is called Non-violent Communication & was developed by Marshall Rosenberg. It’s quite amazing how this simple trick can resolve any conflict & make people interaction more effective, just by creating an emphatic connection between them, instead of the common disconnected dialogue based on zero sum game.

4. Finally, I want to change the computer interface by which people consume & interact with large amounts of information: instead of receiving it as an assorted feed, they should move inside it as an organized space. If you need to find out or discover some things you’re interested in, from a huge collection of items, which interface would you prefer: the conveyer belt in a luggage carousel, in which you wait until you see the items you’re interested in, or the organized space in a store or supermarket, where you can move & look around, see many items sorted by categories, & also remember after some time where different things are located.

For 1 & 2, I’ve been building with my friends a service for helping entrepreneurs follow their dreams, by following the principles of Teal. We basically help people find others who share the same purpose, & guide them in creating a flat team that creates some unique value & delivers it directly to customers.

For the technological item #4, I’ve been working on a framework that generates this organized space VR interface, based on simple protocols of interaction between people, bots & information.

I’d be happy to hear what changes would you like to make in the world?

What’s better – a senior banker with a high-salary stable job or a taxi driver who’s uncertain every day whether he’ll find enough rides or not? At a certain age, the banker may suddenly lose his job, & may not be able to find a new one. The taxi driver, on the other hand, will have daily uncertainty, but no risk of losing his job & not finding a new one.

When an economy is designed so that businesses, such as restaurants, wouldn’t have a risk of closing, the result is that all restaurants look like soviet style cafeteria, with boring food, & eventually the system breaks down. In countries with free market, a new restaurant may close if there’s not enough demand, or any problem with its food, but the whole system guarantees that there will always be a good restaurant for any taste & need.

Experts, government officials, managers, economists & doctors all need to justify their existence, so they would always try to intervene & thereby hurt the organic self-healing force of anti-fragile systems, be it a country, an organization or the human body.

The best strategy for investment & life, is the bar belle: dedicating half for being insulated from the worst risk, while dedicating the other half for random exploration of new directions with much risk, but also large potential of gain.

But behind this smart advice & insight on what’s the necessary attribute for any successful human endeavor, the book beautifully conveys an underlying core truth, that the most important things for a good person, are character & courage. The good person, whom we would best learn about in the Greek & Roman classics (eg Seneca) must always stay true to his principles & believes, regardless what everyone else say or think.

Here’s a work-in-progress overview of the protocol.

• Purpose

  • The purpose of the protocol is to simplify the interface between a team of people and the data they’re working on. The common way to implement this interface today is via middleman apps, which display handcrafted screens for managing the data, which is disparate from the main channel of communication of the team, which normally is chat.
  • Those apps normally use an handcrafted REST API for reading & updating the data, which is stored in some data storage backend.
  • The EntityChat protocol tries to simplify this by replacing all the middleman layers between the user & the data storage backend with 1 standard layer capable of translating the user intentions into data changes & convey data back to the user, so that you don’t need to handcraft apps, screens & API’s, but rather get it for free by just providing simple metadata about the entities.
  • Furthermore, the UI offered by the EntityChat layer is conversational, so you can use it in any chat tool.
  • For any business logic beyond managing the data itself, the EntityChat layer emits events that can be handled by serverless functions adding extra business logic.
  • The protocol replaces apps & API’s, but you can still have them if you want- you’ll just have more clients besides the chat client, that also need to with the EntityChat layer in chat messages.

• Protocol overview

  • The basic interaction in EntityChat is using chat messages:
    • A user sends via chat a text message
    • An EntityChat backend parses the message & updates the data storage backend if needed
    • The EntityChat backend sends a chat message back to the user
    • The EntityChat backend emits event about the data change, for any additional business logic handling

    The chat messages are structured in a simple way, that encode the intention of the user:

    • Every chat message part of the protocol consists of the following parts:
      • speechAct
      • subjectEntityType
      • subjectEntityId
      • relationType
      • objectEntityType
      • objectEntityId

      There are several core speech acts:

      • show: This speech act allows asking for all details of a given entity
      • list: This speech act allows listing all entities or entities matching some criteria

      The other speech acts basically correspond to transitions in entities states, which are described in entity metadata.

      For example:

      • a team dealing with a store, will have an entity called Order with the following states:
        • created
        • cancelled
        • paid
        • prepared
        • shipped
        • lost
        • delivered

        it will create an entity metadata document, listing the transitions between these states:

        • [] -> create -> created
        • created -> cancel -> cancelled
        • [created] -> pay -> paid
        • [paid] -> prepare -> prepared
        • [prepared] -> ship -> shipped
        • [ship] -> lose-track -> lost
        • [ship,lost] -> deliver ->delivered

        Based on the metadata, the EntityChat will support chat messages such as:

        • create order <id>
        • pay order <id>
        • prepare order <id>
        • ship order <id>
        • …
        • show order <id>
        • list delivered orders

        Chat messages with speech act that corresponds to a transition in the entity state, will result in an update to the entity state, captured in a new record in the data storage backend

        • The current entity record is the one with the latest timestamp

     

• Protocol concepts

  • There are 3 main concepts in the protocol:
    • Purpose – an horizontal way to correlate similar entities & teams
    • Team – a vertical way to group entities by their owners
    • Entity – the core pieces of data handled by the protocol

     

  • Entities have several core attributes:
    • Id: Together with the teamId, uniquely identifies the entity
    • Type: The class of things the entity is similar to
    • State: The lifecycle stage the entity is currently at
    • Team: The group of people owning the entity. Every entity always belongs to exactly 1 team
    • Purpose: The unique goal under which the entity was created & of which it serves. Entities of different teams can share the same purpose
    • Timestamp: The time of the last change in the entity state
    • URL: An optional external URL for accessing the entity’s content via the Web

     

  • The data storage backend should be immutable, and store new versions of entities on every update to their state