This talk describes how tokens/decentralization and complex systems relate. Contents:
-blockchains as trust machines
-blockchains as incentive machines
-evolutionary algorithm design (and agent based simulation) for token design
-benevolent computer viruses (aka smart contracts)
-AI DAOs
-blockchains as life
Presented at Santa Fe Institute, New Mexico, Jan 31, 2018
Video at: https://medium.com/abq-blockchain-community/talking-blockchain-ai-complex-systems-3c5a33676f85
2. Outline
⢠Trust machines
⢠Incentive machines
⢠EA design for token design
⢠Benevolent computer viruses
⢠AI DAOs
⢠Blockchains as life
5. Blockchains: Google Sheets, Minus the Google
Whatâs special:
- no one owns it
- anyone can add to it
- no one can delete from it
- Writing to a blockchain is like
etching in stone.
- The sheet stores a list of
transactions (txs).
- Then we use txs to issue assets,
and transfer them
6. Blockchains are databases with
âblue oceanâ benefits
Decentralized / shared control
Immutability / audit trail
Assets / exchanges
7. How:
⢠Decentralized: Byzantine fault tolerant (BFT) consensus
⢠Immutable: chain of blocks = hash linked list of transaction groups
⢠Assets: digital signature on every transaction
âTrust machineâ because it minimizes trust needed to operate.
Alice
Mongo
DB
11. Economic Incentive for Bitcoin
Objective: Maximize security of network
⢠Where âsecurityâ = compute power
⢠Therefore, super expensive to roll back changes to the transaction log
E(Ri) Îą Hi * T
E() = expected
value
# tokens (BTC)
dispensed each
block
block
rewards
hash power of actor
= contribution to
âsecurityâ
12. E(Ri) Îą Hi * T
Bitcoin Token Release Schedule
F(H, t) = 1 - (0.5t/10) = % tokens released after t years
⢠Schedule is fixed in advance
⢠4 years for 50% of tokens released = half-life âthe halveningâ
13. Result of Bitcoin maximizing security?
Maximizing energy usage! > USA by mid 2019!
14. Result of Bitcoin maximizing security?
Maximizing energy usage! > USA by mid 2019!
Like paper-clip generating AI, but with energyâŚ
15. Economic Incentive for Ocean
Objective: Maximize supply of high-quality data (for use by AI)
⢠This means: reward curating data + making it available
⢠Where âcuratingâ = betting on data. Reward taste-making.
E(Rij) Îą log10(Sij) * log10(Dj) * T *Ri
Expected
reward for user
i on dataset j
Dj = proofed popularity
= # proof-of-service
invoked (eg # deliveries
of dataset j)
Sij = predicted popularity
= userâs curation market
stake in service j
(eg dataset j)
# tokens
during
interval
17. Design of Tokenized ecosystem
â Design of EAs (Evolutionary Algorithms)
What Tokenized ecosystem Evolutionary Algorithm
Goals Block reward function
E.g. âMaximize hash rateâ
Objective function
E.g. âMinimize errorâ
Measurement
& test
Proof
E.g. âProof of Workâ
Evaluate fitness
E.g. âSimulate circuitâ
System agents Miners & token holders (humans)
In a network
Individuals (computer agents)
In a population
System clock Block reward interval Generation
Incentives &
Disincentives
You canât control human,
Just reward: give tokens
And punish: slash stake
You canât control individual,
Just reward: reproduce
And punish: kill
18. Agent-based Systems for Token Simulation
⢠Q: How do we design computer chips?
⢠A: Simulator + CAD tools
⢠Q: How are we currently designing tokenized
ecosystems?
⢠A: By the seat of our pants!
⢠Which means we might be getting it all wrong!
What we (desperately) need:
1. Simulators: agent-based systems
2. CAD tools: for token design
19. Reason #37 why you might care: there is tons of
$$ at stake (and we might be getting it all wrong)
21. Then you get
decentralized processing
aka smart contracts.
And you can build a
world computer
having decentralized processing,
storage, and communications
(e.g. Ethereum vision)
What if you used a blockchain (database)
to store state of a state machine?
State
Virtual machine
Decentralized applications
(Dapps)
World computer
Blockchain
22. DAO: a computational process that
⢠runs autonomously,
⢠on decentralized infrastructure,
⢠with resource manipulation.
Itâs code that can own stuff!
(Or, good computer viruses)
DAO: Decentralized Autonomous Organization
State
Virtual machine
DAO Dapp
World computer
Blockchain
24. AI entity is a feedback control system.
That is, AGI.
Its feedback loop would continue on
its own, taking inputs, updating its
state, and actuating outputs, with the
resources to do so continually.
AGI on a DAO?
State
Virtual machine
AI DAO Dapp
World computer
Blockchain
25. An AI DAO is âŚ
⢠An AI running on decentralized processing substrate
<or, equivalently>
⢠A DAO running with AI algorithms
Definition of AI DAO
26. Example: The ArtDAO
AlgorithmâŚ
1. Run AI art engine to generate new image, using GP or deep
2. Claim attribution in blockchain
3. Create multiple editions
4. Post editions for sale onto a marketplace, using Getty
(centralized), or OpenBazaar (decent.)
5. Sell the editions. $ goes to ArtDAO using built-in cryptocurrency
like Ether. IP go from ArtDAO using ascribe.
6. Repeat! Create more art, sell it, get wealthier
27. Example: The ArtDAO
AlgorithmâŚ
1. Run AI art engine to generate new image, using GP or deep
2. Claim attribution in blockchain, using ascribe
3. Create multiple editions, using ascribe
4. Post editions for sale onto a marketplace, using Getty
(centralized), or OpenBazaar (decent.)
5. Sell the editions. $ goes to ArtDAO using built-in cryptocurrency
like Ether. IP go from ArtDAO using ascribe.
6. Repeat! Create more art, sell it, get wealthier
Over time, if ArtDAO makes more money from sales
than from generating new art, then
it will accumulate wealth. And, you canât turn it off.
31. AI DAO Arch 3/3 : Swarm Intelligence
Many dumb agents with emergent AI complexity
32. Angles to Making AI DAOs
⢠DAO ď AI DAO. Start with DAO, add AI. E.g. Plantoid
⢠AI ď AI DAO. Start with AI, add DAO. E.g. numer.ai
⢠SaaS ď DAO ď AI DAO. Convert SaaS to DAO. Then add AI
⢠Physical service ď AI DAO. E.g. Uber self-owning cars
33. Evolving the ArtDAO
Market
Art work
Code
Level to adapt at
How to
adapt
Human-based adapt at the code level.
Here, humans put in new smart contract
code (and related code in 3rd party
services), to improve ArtDAOâs ability to
generate art and amass wealth.
34. Evolving the ArtDAO
Market
Art work
Code
Level to adapt at
How to
adapt
Auto adapt at the market level.
It creates more of what humans buy, and
less of what humans donât buy.
35. Evolving the ArtDAO
Market
Art work
Code
Level to adapt at
How to
adapt
Auto adapt at the art-work level.
Here, a human influences the creation of an
artifact. For example, it presents four variants of
a work, and a human clicks on a favorite. After
10 or 50 iterations, it will have a piece that the
human likes, and purchases.
36. Evolving the ArtDAO
Auto adapt at the code level.
Here, the ArtDAO modifies its own code, in hopes of improving.
⢠It creates a copy of itself, changes that copyâs code just a little bit, and gives a tiny bit of
resources to that new copy.
⢠If that new copy is bad, it will simply run out of resources and be ignored.
⢠But if that new copy is truly an improvement, the market will reward it, and it will be able
to amass resources and split more on its own.
⢠Over time, ArtDAO will spawn more children, and grandchildren, and the ones that do well
will continue to spread. We end up with a mini-army of AI DAOs for art.
⢠If buyers are DAOs too, itâs a network of DAOs, leading to swarm intelligence
We can extrapolate way further. It gets crazy, quickly.
A simple ArtDAO allows us to âidea-inoculateâ so weâre
ready for the scarier stuffâŚ
Market
Art work
Code
Level to adapt at
How to
adapt
42. Trent McConaghy
@trentmc0
We covered:
⢠Blockchains as trust machines, and incentive machines
⢠You can get people to do stuff!
⢠Complex systems for token design. EAs, agents
⢠Smart contracts are benevolent computer viruses
⢠AI DAOs = amazing intersection of AI * blockchain
⢠Blockchains as life
The complex systems world has useful theory & tools for blockchain
world; and blockchain has many interesting problems!
Conclusion