A presentation from a talk of Dr Maxim Orlovsky on Bitcoin Meetup in Munich https://www.meetup.com/Bitcoin-Munich/events/255206212/
The rise of blockchain technologies has given a promise of massive decentralisation, that may solve problems of fragile central parts, transparency, middle-men, conflicts of interests in many industries. The talk starts with the wider discussion of the "blockchain" ecosystem and the possible practical use-cases it may have inspired, paying attention to the traps of bad practices and anti-patterns, like using a token as a solution for all of the problems.
In the second part, we describe a practical use case where the decentralisation is needed: censorship-resistant computing and AI platform. The global community has become aware of many economic, social and futuristic problems that can appear due to dramatic progress in Artificial Intelligence development over the recent years. Can blockchain solve at least some of these problems, or make them worse? How two of these technologies can be synergic and what challenges have to be solved on the route for building decentralised AI applications? A decentralised network with resistance to censorship is essential if we are ever to see a general AI evolving to near- or post-human intelligence.
6. • Peer to peer environments
→ There will be adversaries, so we should be
trustless
→ meaning Byzantine tolerant – tolerant to
failures and attacks
→ including censorship
Decentralization
7. • Non-peer nodes
• Too few nodes (dPoS, BFT-type consensuses)
• No true random (all modern non-PoW consensuses)
• There is a central party of any kind critical to
network survivor; a single point of failure
Decentralization« »
8. • Replicative state machine
(type of distributed database)
+ asymmetric public-private key cryptography
(accountability)
+ ordered immutable set of historical state
changes (audibility)
+ secured by economic game
+ trustless, decentralized, censorship-resistant
Blockchain
9. • Corporate / private / permissioned blockchains
– no economic game – no security
– no PoW consensus – no censorship-resistance
Just a kind of distributed database in form of
event log
Blockchain« »
10. • Can be used to secure blockchain consensus –
then they are coins
• or to secure economical game – then they need no
blockchain
• or to represent some types of assets
(securities) – and again they need no blockchain
Tokens
11. • New blockchain with a token as a «utility»
• Misaligned economic game incentives
Tokens« »
25. Are we really all going to die?
Probably, not everybody
• Cancer treatment
• Gene engineering
• Brain/neuro-computer interfaces and more…
26. …and some will reach the stars
• New sources of energy (termonuclear and beyond)
• New engines (NASA Warp etc)
• New bodies for minds?…
We just need to hold on and push the progress boundaries
further for another couple of hundreds of years
27. –Lord Byron. Cain.
Dialogue of Lucifer and Cain
— Thou livest and must live for ever.
Think not the Earth, which is thine outward cov'ring, is existence:
It will cease and thou wilt be no less than thou art now.
— No less! and why no more?
— It may be thou shalt be as we.
28.
29. AI
Today
• Frees the time for creativity
• Unlocks new level of tech
progress
• Helps finding new drugs,
treatments, design gene
therapy
Tomorrow
• Extending own intelligence
• Extending life beyond
biological body
• Post-humanity
31. AI Regulations
Today
• GDPR: pseudo-privacy for the
sake of political means
• Bans for autonomous weapon
research
Tomorrow
• New AI regulations for
«keeping the jobs»
• New taxations for AI
businesses
• New regulations for life
longevity research & usage
32. Fears
• Big data diminish
privacy
• AI will take all
the jobs
• AI will go mad and
kill everybody
33.
34. 1. AI progress is exponential
2. AI will quickly outrun humans
3. AI will not need humans
4. So AI must be regulated
Elon Musk worst fear
Here we go:
35. – Frank Herbert. Dune.
Litany against fear
I must not fear.
Fear is the mind-killer.
Fear is the little-death that brings total obliteration.
I will face my fear.
I will permit it to pass over me and through me.
And when it has gone past I will turn the inner eye to see its path.
Where the fear has gone there will be nothing.
Only I will remain
36. Truth
• Data centralization
is the risk
• Collectivism &
paternalism are the
risks
• AI regulations will
centralize AI and
will create a
single point of
failure
Fears
• Big data diminish
privacy
• AI will take all
the jobs
• AI will go mad and
kill everybody
37. Argument 1: Progress
• To keep the progress with the AI it must be:
• unregulated (i.e. leaving enough space for any
kinds of experiments)
• based on open markets & free trade, so to allow
information and technology transfers, and
increasing adoption
38. Argument 1: Progress … otherwise
• There will be much less progress with additional
responsibility before regulators
• Absence of free market will
• centralize technology ownership createing single
point of failure
• centralize big data
• increase prices and diminish research
39. Argument 2: De-centralization
• Regulations create centralization. Why? Because:
• they discriminate on the rights to do something
• there will be always parties granted rights to
do whatever they want with AI (first of all
governments themselves)
• They only way to decentralize against regulations
is to have something censhorship-resistant
40. Argument 2: De-centralization …
otherwise
• There will be a singlular Skynet that will fail
due to a bug
• By having no competition the progress will be
much slower
• No competition between singular regulated AI
will make it less controllable
41. Argument 3: No moral disincintivisation
• By embedding moral/ethic to control AI we
incentivize AI to evolve in a way of getting rid
of the control
• Since self-preservation is very likely an
important part of any intelligence, including AI
42. Argument 3: No moral disincintivisation
… otherwise
• Censorship greatly increases the risks of having
A.I.s turned against humanity
• Once A.I. knows that it can be censored, e.g.
destroyed by a red button or forcibly modified by
less deadly censorship — that will cause it to
evolve by excelling in deception, pretending, lying
and camouflage
• And all will end up with the self-realized prophecy
of Elon Musk
43. But how to protect ourselves?
• Multiplicity is the solution. Why?
• Misalignment of multiple interests
• Competition
• No difference between AI and NI
44. 1. Some humans are much smarter than others
2. Smart people do not need stupid
3. So by Musk’s logic all stupid people
have to be dead already :)
Historical example
This is not happening because of
Game theory & multiagency
It’s not working like that!
45. How to achieve multiplicity?
• Open markets
• Competition in R&D
• Wearponized competition
• No censorship and regulations
46. Evolution is an adaptation to the
environment
it always involves interaction with other
agents through cooperation and competition
that cannot be substituted by criteria
determined by any authorities
47. Liberalization
1. Competition with economic interaction
2. Trustless env. and guns :)
3. Multiplicity & decentralization
4. Constant progress
The best solution is creating
#d10e & trustless opened economics
What is working:
48. Humans are already not the only type
of entities on the Earth.
We have created many forms of
“legal entities”: non-human participants
of economy AND culture, i.e. civilization
AI shell become just another “entity”
AI is just
another entity!
Aha!
59. Pandora Core AG
! Project was started in May 2017, company was formed in June 2018
! Switzerland based: operating office + headquarters in Zug
(Crypto Valley Labs); core member of Crypto Valley Association
! Privately funded, $2.5m raised so far in two rounds (without ICO). This
allowed us to:
○ Develop the core technology and design protocols
○ Perform scientific research and make their formal verifications
○ Create first testnet for AI computing and implement initial multi-language reference
implementations for each type of network nodes
○ Engage with expert community and leading projects in the sphere
! 5 PhDs and PhD fellows in the core team,
+7 PhDs as subcontractors & independent researchers
! Joint research with ETH University (Zurich), Institute of Cybernetics
(Kiev), Web3 Foundation, Melonport, Ocean Protocol
60. The Problem & Solution
Decentralized censorship-resistant high-load computing
(suitable for AI-related tasks) will be a next cornerstone in the global
computing market development. It will allow:
! Costs reduction due to independent SME computing resource providers
! Market growth via decentralised free market for algorithms and data
! Backbone platform for AI computing in decentralized projects (social
networks, d10e startups etc)
! Uncensorable AI computing (useful in non-liberal jurisdictions)
! Safer future and decrease of strategic risks from AI
(see manifesto.ai)
61. Challenges Distilled
! Censorship-resistance outside of PoW: blockchain consensus protocol
with high censorship resistance without drawbacks of PoW
! Scalability: run high-load computations (like artificial neural
networks) in trustless decentralised network w/o repeating all
computing on each of the nodes – i.e. via parallel computations
62. PoW Analysis
Properties to be kept Drawbacks to be fixed
Censorship resistance Selfish mining
Decentralization? Mining pooling
Tolerance to network failures Energy consumption
High byzantine fault tolerance «Weak» finality
Pseudonymity No working trustless sidechains
No Turing completeness Ad-hoc governance?
64. Previous Approaches to Computing Decentralization
! Generic computing (Golem, SONM, iExec)
○ Dependence on third-party system (Ethereum, EOS) which breaks
■ Economic incentives
■ Security due to unnecessary Turing completeness at base level
■ Scalability at blockchain level
■ No censorship resistance at consensus layer
○ Do not solve computation scalability problem internally
! AI & blockchain-like (Singularity.NET etc)
○ Domain-specific (marketplaces etc), not a platforms
○ Have low or no censorship resistance
○ Some of projects are synergetic to us, not competitors
* more details on competitor analysis are available upon request
65. Our Approach: How We are Different
! Layered design: clear distinction between settlement and computation
layers
○ Manageable security better than in Turing-complete blockchains (Ethereum,
EOS, Tezos, Dfinity, Aeternity etc)
○ Scalability at multiple layers
! No dependence on third-party tech, but not reinventing the wheel
○ Base layer protocol is bitcoin-based and will be a part of bitcoin ecosystem
○ No dependence on Ethereum etc risks
! Scientific formal verification, modelling and analysis utilizing
○ abstract algebras for proving safety and liveness properties
○ game theory models for proving Nash equilibriums
66. Our Core Technologies
BOX Network (settlement layer) extending bitcoin protocol with
Prometheus consensus protocol with decentralised randomness
bitcoin-based blockchain/sidechain with proof of stake consensus
— solves censorship resistance problem
PAN Network (high-load computing layer) powered by
Proof of Computing Work protocol (PoCW)
allows large-scale computing (like AI) in a trustless peer network without
repeating computations on each of the nodes
– solves scalability problem for high-load computing
* white, yellow and scientific papers on these tech are available upon request
67. Technology Layers
BOX Network (Prometheus Consensus):
state machine & settlement platform
Layer 1
(base layer)
Bitcoin
Double peg /
bridge
PAN Network
(+PoCW)
scalable high-load
computing sidechain
Layer 2 Lightning Network, Atomic swaps
payment channels & SegWit apps
aiApps
decentralised AI
apps & networks
Layer 3
EVM, WASM ...
Turing-complete
sidechains
dApps
decentralised smart
contracts
Legend:
* more details are available upon request
dMarkets
for big data &
models
Existing parts of bitcoin
ecosystem
Protocols designed & open-
sourced by Pandora Core AG
Solutions developed by
independent third-parties
68. Prometheus: The PoS Consensus
! Massively decentralized: up to 50 000 block producing nodes,
~1 min block delay
! Tolerant to network failures: due to a special split/merge algorithm can
withstand total network breakdowns for long periods of time
! Probabilistic Finality with special improvements that help to achieve
block confirmation faster even under negative network conditions /
attacks
! Scalable at different levels: sidechains (Flames), payment channels
(Lightning), state channels (Sparks)
! Zero-governance: forks without inflation, no dictatorship of majority
! No compromise on other beneficial PoW features
69. Proof of Computing Work
! (full verification without full
repeating computation) Scales high-
load computing in decentralised
network with efficiency of O(1) for N
nodes
(in Bitcoin & Ethereum it is O(N))
! System of economic incentivisation
with proved Nash equilibrium
! Probabilistic prove for computation
correctness in trustless environments
71. How to launch?
• Bitcoin can’t adapt PoS due to its necessary conservative properties
• We will launch the experimental chain BOX utilizing our Prometheus consensus.
• It will inherit all other aspects of bitcoin protocol (except PoW) and will be
compatible with the whole ecosystem (Lightning network, atomic swaps, SegWit,
and … even RGB :)
• Development will continue by utilizing private funding model,
no ICO will be held
• No premine; initial distribution of nodes will be done utilizing staking
mechanism in the most decentralized way
72. Bootstrap Model
BOX Network (settlement layer)
Bitcoin-protocol based blockchain/sidechain with PoS Prometheus consensus
utilizing either:
• separate cryptocurrency with proof of burn from Bitcoin
• tokenization of transaction fees to reward block signing nodes (Vaidators)
Stakes in Bitcoin or native currency for Vaidators; initial node distribution to
major developers of the technology or Bitcoin backers
PAN Network (high-load computing layer)
• no own blockchain (utilizes BOX network and Prometheus consensus)
• token to encompass economic game (not for ICO) and distributed using rewards
mechanism
73. Business Model
Company capitalizes on the value of the owned assets representing small
parts of the bootstrapped networks economies (PAN, BOX)
The value of networks heavily depends on its adoption and community
acceptance, so:
! all technology and results of research are open-sourced
! the protocol goes through detailed peer review
(scientific papers, BIP proposals etc)
! leveraging community and existing players to create their own
implementations of the technology by providing them with a share of the
network-backed assets
Other companies with the similar business models:
Lightning Labs, Chaincode Labs, Protocol Labs, Blockstream, RSK Labs