Slide 1

Slide 1 text

Aaron Li [email protected] AI & Blockchain

Slide 2

Slide 2 text

2 AI Blockchain Bitcoin: first crypto & blockchain GPU mining & first alt coin Ethereum: General-purpose platform First ICO with big success 1000+ new coins & tokens Deep learning on GPU General purpose platforms large scale adoption 1000+ AI startups Basic adoption & open source tools Towards applications 2009 2011 2015 2017

Slide 3

Slide 3 text

3 AI Blockchain Stage Early Adoption Experimental Adoption Government & Industry Mostly developers & enthusiasts Entry Barrier High Open Beneficiaries Mostly Institutions Mostly People

Slide 4

Slide 4 text

4 How AI and blockchain can benefit each other? Where are we going? What are the applications?

Slide 5

Slide 5 text

5 Challenges Privacy, Speed, Scale, Generality Projects Ekiden ($45M raised) Cortex (25 fund investors) And more Goal Perform training and inference on blockchain Benefits Transparency, trust, collaboration Machine Learning on Blockchain Blockchain Projects for AI

Slide 6

Slide 6 text

6 Ekiden v.s. Ethereum 600x throughput 1 / 400 latency 1000x less cost Credit Score Assessment Smart Building Thermal Modeling Multi-player Games General Purpose Computing And more Applications Machine Learning on Blockchain Blockchain Projects for AI

Slide 7

Slide 7 text

7 Data Sharing & Exchange Examples: Engima, Datum, BOTTOS Model Sharing & Evaluation Examples: OpenMinded, and many more Labels & Data Crowdsourcing Examples: Steem, Augur (alternative use case) Applications Algorithmic Trading, Data Trading Personal Data Collection, Personal Info Monetization Prediction & Model Market, Labeling Bot Machine Learning on Blockchain Blockchain Projects for AI

Slide 8

Slide 8 text

8 Information Summarization & Prediction Evaluation and Fraud Detection Similar to Chainalysis, TokenInsight, and others Parameter and Node Optimization Similar to what Google Brain did for Google infrastructure AI Projects for Blockchain

Slide 9

Slide 9 text

9 Information Summarization & Prediction Summarizing and predicting trends and sentiments for cryptocurrencies, dApps, and blockchain projects Crypto Qokka (cryptoqokka.com) Powered by Opinion Engine from AI Projects for Blockchain

Slide 10

Slide 10 text

10 It is nimble, fast, scalable. Get meaningful results even with small data e.g.: Learned 100M reviews in less than 3 days, visualized pros and cons of every Amazon product in knowledge maps High Efficiency Machine Learning Distributed Large-Scale Crawler Easily crawled hundreds of millions of reviews in less than a week. Take advantage of any public data to kick start your AI solution Scalable AI-as-a-service platform for unstructured text data APIs and solutions designed to make engineering and integration simple and easy. No AI engineers or PhDs required API Solutions Corpora Engine

Slide 11

Slide 11 text

11 Sentiment v.s. Price News Analysis Topics, Opinions, Highlights Market Overview Crypto Qokka cryptoqokka.com Corpora Engine for Cryptocurrency

Slide 12

Slide 12 text

12 Price: >20% drop in 1 day $283 on Sep 4 $225 on Sep 5 Negative Sentiment ~100% increase in 1 day 32% on Sep 1 62% on Sep 2 Positive Sentiment ~75% drop in a day 15% - 20% on Sep 1 5% - 10% on Sep 2 Ethereum Sentiment Spike Price Move Sentiment v.s. Price

Slide 13

Slide 13 text

13 Trading Volume Social Volume (# comments) Week ending in Feb 18, 2019 Sentiment Score higher +/- ratio 㱺 higher score less neutral sentiment 㱺 higher score

Slide 14

Slide 14 text

14 Bitcoin: Feb 12 - 18, 2019

Slide 15

Slide 15 text

15 Overall sentiment of the market How are blockchain projects perceived? Which project might “beat the market”? Market Overview

Slide 16

Slide 16 text

16 A quarter’s market sentiment in review Nov 19, 2018 - Feb 19, 2018 Relative to BTC Market Overview

Slide 17

Slide 17 text

17 Latest news for each crypto that people comment on Measures sentiment of people’s comments for each news News Analysis

Slide 18

Slide 18 text

18 Not all positive news leads to positive reaction Why sentiment analysis on comments?

Slide 19

Slide 19 text

19 News on Nov 18, 2018: Crypto ETF Approved in Switzerland A few days later… Bitcoin price crashed (~$4500 down to ~$3000) Comments are negative. Why? Because it is not really an ETF Not all positive news leads to positive reaction Why sentiment analysis on comments?

Slide 20

Slide 20 text

20 Trending topics for each crypto What do people talk about? Technical matters, or buzz words? What kind of technical matters? Topics, Opinions, Highlights

Slide 21

Slide 21 text

21 Visualization of topics and phrases for ETH (week ending in Feb 20, 2019) People still mostly discuss use cases and tech even in bear market Topics, Opinions, Highlights

Slide 22

Slide 22 text

22 Digging into topic “system” for Ethereum (week ending in Feb 20, 2019) What happened this week that really matters? How do people react and what do they think? Topics, Opinions, Highlights

Slide 23

Slide 23 text

23 Sudo Review Understand thousands of reviews in seconds Data 10 million Amazon products and 100 million reviews APIs Review Analysis, Summarization, Visualization, Aspect-based Search Crypto Qokka Decrypto crypto market sentiment and trends Data 10 million Reddit posts APIs Market Overview, Price and Sentiment, News Analysis, Topics and summaries Patent Immigration Legal Trading Asset Management Finance Trend Prediction Online Community Consumer Corpora Engine: Current Use Cases

Slide 24

Slide 24 text

24 Interested in other use cases? Let’s talk: [email protected] Scalable AI-as-a-service platform for unstructured text data Corpora Engine Scale up language understanding