�ݺ�ߣ

HOW WE REINVENTED MACHINE
LEARNING
SOLVED NP-COMPLETE ML
PROBLEMS & CRACKED RSA 2048
TOO!!!
Automatski Solutions
http://www.automatski.com
E: Aditya@automatski.com , Founder & CEO
M:+91-9986574181
© Automatski Solutions 2017. All Rights Reserved.

AUTOMATSKI’S EFFORTS
(LAST 25 YEARS)
1. 25+ years of Fundamental Research
2. Solved 50-100 of the Toughest
Problems on the Planet considered
unsolvable in a 1000 years given the
current state of Human Capability
and Technology
3. Including 7 NP-Complete + 4 NP-
Hard Problems
4. Broke RSA 2048
5. … in 1990’s

AUTOMATSKI HISTORY &
TIMELINE

WHAT WAS THE PROBLEM?
 We Solved Multiple NP-Complete Problems like…
 The N-Queens Completion Puzzle…
 With that it was proven that RSA 2048 is broken
 The Problem?
 For us to be able to actually demonstrate doing it
 Someone would have to encode the RSA 2048 problem in the N-Queens Completion form or one of the forms
of the NP-Complete Problems we had solved
 Question: Why would anyone/researcher put in a lifetime of effort to workout encoding to an NP-Complete
form(s) which is known to be unsolvable?
 Due to extreme Apathy and Misfortune,our inability to raise funds, we weren’t able to do it and
demonstrate it. So we did what we could do best…
 We solved more and more NP-Complete Problems one after the other, reinforcing that if we weren’t able to
break RSA 2048 in one way we would be able to do so in 10-13 of other ways.
 Those who offered to help simply wanted the complete source code to the Algorithms
 Hence progress has been slow…

BACK TO OUR
DISCUSSION…
 What is machine learning?
1. Regression
2. Clustering
3. Classification
 For one concept why do we have 50 even 100’s of Flavors of Algorithms?

ADITYA’S FIRST NP-
COMPLETE PROBLEM
SOLVING CONJECTUREconjecture
/kənˈdʒɛktʃə/
noun
an opinion or conclusion formed on the basis of incomplete
information.
1. All NP Complete Problems can mostly be solved
by Classical Computing in O(N^3)<= Order <=
O(N^5)
2. It is near impossible to do so in O(N) and rarely
requires Order >= O(N^7)
3. Sometimes can be done in O(N^2) and needs
O(N^6)
*** Empirically Based on solving 7 NP-Complete & 4
NP-Hard problems in Polynomial Time

K-MEANS IS
NP-HARD
 K-Means is NP-Hard even on a plane because, if you solve it in P, you can also
solve planar 3-SAT in P. Planar 3-SAT has already been proved to be NP-Hard, and
thus K-Means is NP-Hard.
 Or in technical terms, you can reduce planar 3-SAT to k-means...
 And by Induction if you solve K-Means in ‘N’ Dimensions in Polynomial Time You
Solve “General” K-SAT
 And you crack RSA 2048

HOW IS K-MEANS DONE
CURRENTLY
 (Think of items as points in an n-dimensional
space). The algorithm will categorize the items into k
groups of similarity. To calculate that similarity, we will
use the Euclidean distance as measurement.
 The algorithm works as follows:
 First we initialize k points, called means, randomly.
 We categorize each item to its closest mean and we
update the mean’s coordinates, which are the averages of
the items categorized in that mean so far.
 We repeat the process for a given number of iterations
and at the end, we have our clusters.
 The convergence criterion is met when the maximum
number of iterations specified by the user is exceeded
or when the cluster centers did not change between
two iterations.

MACHINES USED
 Development of Solution
 Machine #1
 AMD FX-6300 (6-Core), 32GB RAM
 Eight Year Old Machine
 Physical
 Testing of Parallelization & Concurrency
 Machine #2
 AWS EC2 m4.16xlarge ($3.2/hr)
 64 Core, 256 GB RAM
 Virtual Machine
 Machine #3
 AWS EC2 x1.32xlarge ($13.4/hr)
 128 Core, 1952 GB RAM
 Virtual Machine

BEFORE THE
A Message from our CFO!!!

MESSAGE FROM THE CFO!!!
 We are looking for Research & Development Grants
 Between
 $1m to $100m
 $1m will help us with the Research
 $100m will help us develop Path Breaking Solutions for over 10+ Domains using the
Breakthrough with a Big Bang.

AUTOMATSKI FUNDAMENTAL
RESEARCH
 Fundamental Research at Automatski has been working for the last 20-25+ years on
solving the toughest problems on the Planet.
 We have solved 7 NP-Complete Problems and 4 NP-Hard Problems, including the N-
Queens Completion Millennium Problem.
 We are applying them towards breakthroughs in 50+ Technology Domains
 These problems are considered unsolvable in a 1000 years given the current state of
Human Technology and Capability

SOLVED PROBLEMS
 N-Queens Completion
(Millennium Puzzle) Clay
Math Institute
 3-Sat/k-Sat
 Knapsack***
 Longest Common
Subsequence
 Travelling Salesman
Problem***
 3DM/nDM
 Graph Coloring -
Chromatic Number
 Linear Programming
 Integer Programming
 Mixed Integer
Programming
 Quadratic Programming
 Universal Expression
Programming
 Global Optimum in Hyper
Dimensional Space
 K-Means Clustering
 Universal Clustering
Algorithm
 Universal Constraint
Programming/Scheduling
 Integer Factorization***
 Prime Number Test***
 Universal Regression
 Non-Linear Random
Number Generation
 Automatic Theorem Proving
 Universal Experience
 Universal Heuristics
 Consciousness, Mind, Brain
 Genomics
 Billion & Trillion Actor Nano
Second Framework
 Universal Multi-Scale Simulations
 Internet Scale Rule Engine
 Internet Scale Workflow Engine
 Perfect Finance/Markets
 Perfect Environment
 Compromised All Cryptography
(RSA-2048, Elliptic Curve etc.)
 Post Quantum Cryptography
 Logarithmic Gradient Descent
Convergence
 Blackbox Function
Cracking/Reversal
 Hash Reversal (Incl. SHA-
256/512, LanMan etc.)
 NP-Complete Machine Learning
Algorithms (Clustering,
Regression, Classification)
 NP-Complete Deep Learning
Algorithms (ALL)
 Artificial General Intelligence
 Robotics (Simulations + RAD)***
 Universal Emotions
 Universal IQ
 Universal Creativity

FUND OUR RESEARCH
Together we can build the foundations of a better world

THE DEMO!

WHAT JUST HAPPENED?
 Did I use Quantum Annealing?
 Was that Outcome Probabilistic?
 It seemed like a Linear Order O(N) Algorithm
 How is that Possible? It can’t be true.
 No, we didn’t use Quantum Annealing
 And it was a Completely Deterministic Algorithm.
 And it is in Polynomial Time
 And its Massively Parallelizable.
 And we haven’t even done it on GPGPU’s, ASIC’s/FPGA’s, HPC, Super Computers yet
 And it can solve problems in a Billion Dimensions with Billions of Data Points.
 Run once and get a solution in 100% certainty.
 Run Just Once and RSA 2048 will Definitely be cracked! Eventually in Polynomial Time.
*** It was pausing and massively slow due to Garbage Collection

HENCE…
 We have reinvented Machine Learning
 RSA 2048 is Cracked!
 Yet Again!
 For the Nth time using N different schemes. Recall…
 P = NP !!!
 7 NP-Complete & 4 NP-Hard Problems Solved in Polynomial time
 N-Queens Completion etc. Including K-Means
 Quantum Computing
 Other Schemes…
 Integer Factorization
 K-SAT
 …
 You know the best part?
 Out of all the Polynomial Algorithms we have to crack RSA 2048. This is not the fastest.

ORDER OF ALGORITHMS
 Exponential => O( 2 ^ N ), O ( 3 ^ N ), O ( a ^ N) …
 Polynomial => O ( N ^ 2 ), O ( N ^ 3), O ( N ^ 5 ) …
 Logarithmic => O ( Log( N ) )…

BUT YOU DIDN’T SHOW US …
1. Regression, and
2. Classification
We debated it for years before this presentation. And concluded if we demonstrated all
three, then a statistical reconstruction of our Algorithms would be imminent. And
demo’ing/proving one was enough to prove RSA had fallen. Hence we didn’t.

FROM HERE…
 Fact: We have Already proved feasibility of cracking RSA 2048 in 10+ Proven ways. We
now have to crack it.
 We will incorporate our ML Algorithms and release a SaaS based Cloud ML Platform.
Which will be one of the finest in the world.
 In the future video’s you will see
 More Polynomial Time Solutions to NP-Complete Problems with accompanying proof.
 Logarithmic Time Algorithms (Our best kept secrets)
 E.g. General Purpose Constrained Optimization Solvers
 Also a Deterministic Algorithm to Find the Global Optima in a Billion Dimensions/Variables.
 I’m sure you will find it very interesting!!!

AVAILABLE
CONFIGURATIONS
 1k Dimensions/Data Points (Entry Level)
 1m Dimensions/Data Points
 1bn Dimensions/Data Points
 1tn Dimensions/Data Points

DO YOU HAVE ANY IDEA?
 What we can do with all the things
Automatski is demo’ing?
 Even though we have just shown
everyone 3-4 out of a 100+
Inventions and Innovations?
 Can you put a $ value on it? How
much will it be? Millions? Billions?
Trillions???
 Any guesses? Wanna try???

NEXT STEPS FOR
PROSPECTIVE CUSTOMERS
 Please contact sales at info@automatski.com
 Send an email with the following information…
 Areas & Applications of “Deterministic” Machine Learning
Needed
 Or Kind of Cryptography that needs to be Cracked.
 We will take it from there…
*** No Free Trials/Pilots/POCs Offered

WARNING!!!
 Don’t contact us asking for The Source Code
 Don’t contact us asking us to
 File Patents
 Make Public Disclosures of our Algorithm(s)
 Publish Academic Papers

SAMPLE PROBLEMS &
RESULTS
 The .Zip File Contains
1. Sample Problems
2. And their Solutions
 Download .Zip File here
 http://bit.ly/2DMHf0a
 Download this Presentation here
 http://bit.ly/2ItyRl0

�ݺ�ߣ

Automatski - How We Reinvented Machine Learning, Solved NP-Complete ML Problems & Cracked RSA 2048 Too!!!

More Related Content

Automatski - How We Reinvented Machine Learning, Solved NP-Complete ML Problems & Cracked RSA 2048 Too!!!