O slideshow foi denunciado.
Utilizamos seu perfil e dados de atividades no LinkedIn para personalizar e exibir anúncios mais relevantes. Altere suas preferências de anúncios quando desejar.

Applied AI/ML in the Workplace - Geek Food for Thought

56 visualizações

Publicada em

Overview to University of Texas at Austin on AI technologies and what Thomson Reuters Labs is about.

Publicada em: Negócios
  • Seja o primeiro a comentar

  • Seja a primeira pessoa a gostar disto

Applied AI/ML in the Workplace - Geek Food for Thought

  1. 1. Presented by: Tarek Hoteit, PhD, IT Director Katherine Li – Data Scientist Image courtesy of https://pixabay.com/en/caution-computer-desk-geek-humor-1295260/ Applied AI/ML in the Workplace Geek Food for Thought @ The University of Texas in Austin Computer Science Department Thomson Reuters Labs http://labs.tr.com
  2. 2. Agenda • About Thomson Reuters & Thomson Reuters Labs • Key AI projects @ Thomson Reuters • Geek talk on Applied AI for students • Code Walkthroughs
  3. 3. 3 Thomson Reuters • Leading source of news and information for the professional market • Business operation in 100 countries for over 100 years • Key businesses: legal, tax & accounting, financial & risk*, Reuter News • ~$5.3B revenue, 80% of 2017 Revenue cover legal professionals, tax professionals, and corporates. 20% includes Reuters News and global print, 4.4% growth annually (post sale of finance and risk) • Distinct customer group regions – 62% America, 27% EMEA, 11% Asia • 93% of product & services are online based • Ranked 66 Best Global Brand in 2017* Based on Investor Relations Summer 2018 Booklet & 2017 Fact Book *Global Brand 2017 link
  4. 4. 4 What makes Thomson Reuters, Thomson Reuters? Jim Smith, president & CEO explains
  5. 5. 5 Founding fathers of Thomson Reuters Paul Reuter (1821-1899) innovator and entrepreneur at young age • Started working as clerk at the age of 13 and accidently met Carl Friedrich Gauss, one of history’s most influential mathematicians • Passionate for speedy transmission of information since childhood • Age of 28 started a news-collecting agency in two cities between Germany and France and connected both through pigeon-posts • Expanded sharing of intelligence knowledge between Britain and rest of Europe at 30 and later with the rest of the world using special underground cable https://commons.wikimedia.org/wiki/File:PaulJuliusReuter-bust-v.jpg https://en.wikisource.org/wiki/1911_Encyclop%C3%A6dia_Britannica/Reuter,_Paul_Julius,_Baron_de
  6. 6. 6 Founding fathers of Thomson Reuters Paul Reuter (1821-1899) innovator and entrepreneur at young age
  7. 7. 7 Founding fathers of Thomson Reuters Roy Herbert Thomson (1894-1976), 1st Baron Thomson of Fleet, media entrepreneur • In his late twenties, started selling radios in Toronto but found it restricting for his audience so he founded a radio station with $201 in his pocket • Age of 40 acquired his first newspaper with a $200 down payment • Modest man who rode the London Underground each day to work https://commons.wikimedia.org/wiki/File:Roy_Thomson.jpg https://en.wikipedia.org/wiki/Roy_Thomson,_1st_Baron_Thomson_of_Fleet
  8. 8. 8 “With employees in more 100 countries serving across industries, we offer opportunities as boundless as the world.” visit our careers website http://tr.com/careers
  9. 9. 9 Thomson Reuters Technology & Careers • Diverse products and solutions in technology • Very active development community across the world • Research & development innovation center (labs) across the world - https://innovation.thomsonreuters.com/en/labs.html • Internship programs in the US include Ann Arbor, MI; Eagan, MN and Dallas, TX; U.S. – Ranked Top 100 internship program in 2018 • Developer Community for API access on financials, risk, legal, tax & accounting via • Open Source Community over 127 repos on GitHub https://github.com/thomsonreuters • Incubator program in Zurich for RegTech, LegalTech, TaxTech and NewsTech • Active career opportunities at http://tr.com/careers
  10. 10. 10 Thomson Reuters Labs – “Innovation is at the core of everything we do” Toronto: + Waterloo • Partnerships with CommuniTech & University of Waterloo. London: • Partnerships with the Open Data Institute and Imperial College Cape Town: Partnered with the Cape Innovation and Technology Initiative (CiTi) Singapore: Exploring economic development programs and university partners. Feb 2017 Boston: • Partnership with MIT, • FinTech Sandbox Zurich: • Expanding our footprint & establishing new partnerships. Blockchain incubator. 2017 San Francisco: • Starmine Dallas: • We are just getting started • Planned partnership internal and external (universities, community) labs.tr.com
  11. 11. 11 • Global team of data scientists, research scientists, full stack developers, and designers; specializing in data science & analytics, data visualization, artificial intelligence and blockchain. • Major products recently rolled out • West Law Edge – the most intelligence legal research platform ever • World Check– risk intelligence for compliance and fight against crime • Data Privacy Advisor – content intelligence with automated question and answering capabilities • TR Knowledge Graph Feed API & Thomson Reuters Intelligent Tagging – entity relationship access to 2 finance billion records to incorporate in big data solutions • Reuters News Tracer - separate fact from rumor on Twitter in real-time Thomson Reuters Labs – “Innovation is at the core of everything we do” labs.tr.com
  12. 12. 12 About WestLaw Edge Team of 60+ AI research scientists, software engineers, and designers delivered the product in partnership with other parts of the business. Key AI Features: • Identify overruling risk using machine learning and nlp • Are cases that were decided on the same legal issue as red-flagged cases still good law? • NLP for issue identification and machine learning for risk classification • Question Answering, using advanced NLP & machine learning • Litigation Analytics • Statistics by judge, court, attorney, law firm, or case type • NLP for parsing dockets; machine learning for classifying outcomes and chaining motions with outcomes
  13. 13. 13 About WestLaw Edge http://www.westlawedge.com
  14. 14. 14 Mona Vernon with the Economist on the World in 2018 What’s driving the revolution of AI and machine learning? “There is a combination of things that are coming together. A massive increase of compute powers, algorithms are getting better, and, more importantly, there is this rise of big data” – MonaVernon ,TR Labs CTO
  15. 15. 15 Industry AI innovation includes you! The University of Texas at Austin is at the hub of AI research • Great clubs in AI development : UTCS Artificial Intelligence • More 100 projects and majority in neural networks (project list) • Mostly PhD students and alumni (people-list) • Numerous publications (list) • Wide variation of labs such as automatic programming, computer vision, deep learning, neural networks, and more (labs list) • Also lots more clubs for computer science students, grads and under- grads. (CS Students organizations) Computer Science program ranked 6th in the world in 2014 (article)
  16. 16. 16 Don’t just stop there… Consider applying AI now and all the time as a potential startup, side projects, open source contributions, social change, or even at home !
  17. 17. 17 Publishing is great but consider more applied AI, open source code contribution, makers projects, and solutions that have a societal impact! We have enough papers. Stop publishing, and start transforming people’s lives with technology!” - Andrew Ng 2017 Andrew Karpathy (2017) “a peek at trends in machine learning”
  18. 18. 18 Consider automated machine learning services such as AutoML for your next project • Cloud-based services for ML/AI accessible for everyone • Small-business like Sittercity automated baby-sitter review process with no ML experience using Google Cloud AutoML (article) • Feedzai Genome currently using AutoML for fraud dection (article) • Traditional ML vs. AutoML SOURCE: JANAKIRAM MSV
  19. 19. 19 Play around with gadgets and DIY tools Raspberry PI Zero Or Zero W $5-$10 Amazon DeepLens $249 Google AIY ~$50 for Voice, ~$90 for Vision Pocket Chip $50 remaining stock IBM TJBot -A Watson Maker Kit
  20. 20. 20 Consider multilingual projects for your hometown Field of Study Percent Internation al Number of Full-Time Internation al Graduate Students in 2015 Number of Full-Time U.S. Graduate Students in 2015 Electrical Engineering 81% 32,736 7,783 Petroleum Engineering 81% 1,258 302 Computer Science 79% 45,790 12,539 Industrial Engineering 75% 7,676 2,539 Statistics 69% 4,321 1,966 Economics 63% 7,770 4,492 Mechanical Engineering 62% 12,676 7,644 Civil Engineering 59% 9,159 6,284 Chemical Engineering 57% 5,001 3,834 Pharmaceuti cal Sciences 56% 1,931 1,502 Metallurgic al/Material s Engineering 55% 3,723 3,103 Agricultura l Engineering 53% 726 654 Agricultura 53% 881 796 79% of computer science students in the USA are international students but most benefit will go to China and USA. Great but how about other nations or in your own hometown? https://www.insidehighered.com/quicktakes/2017/10/11/foreign- students-and-graduate-stem-enrollment PWC 2017
  21. 21. 21 Be practical in your solutions - think future generation (your generation) “Gen Z Is Set to Outnumber Millennials Within a Year” “Gen Z will comprise 32 percent of the global population of 7.7 billion in 2019, nudging ahead of millennials, who will account for a 31.5 percent share” - Bloomberg https://www.fusionnetworks.co.nz/news/2018/09/generation-z https://www.bloomberg.com/news/articles/2018-08-20/gen-z-to-outnumber-millennials-within-a- year-demographic-trends “Gen Z Is Coming to Your Office. Get Ready to Adapt.” “The generation now entering the workforce is sober, industrious and driven by money. They are also socially awkward and timid about taking the reins” - WSJ
  22. 22. 22 Edit presentation title on Slide Master using Insert > Header & Footer Code walkthrough
  23. 23. 23 • We saw BeVote app launched by UT Austin and found it awesome! • Brainstormed in few minutes how to use Applied AI for a chatbot version • We then took some content examples from the website and developed chatbot for Facebook Messenger Took some text And added to an AWS Lambda function in Python Used NLP TFIDF algorithm and Facebook Developer toolkit Code Walkthrough 1 – BeVote app-inspired Chatbot – 2 hours to complete
  24. 24. 24 Code Walkthrough 2 – courses recommendation engine • We built different types of recommendation engine for courses, similar to Netflix or Amazon: 1) similar courses 2) people who took this also took that 3) personalized recommendation
  25. 25. Built in few hours a Sentiment Analysis on Twitter using Django, Docker containers, Python & Google NLP Twitter API using Tweepy Python Library & Twitter Dev Account Local Docker running PostGresql databaseDjango & Python to run and manage the code and data Google Cloud Natural Language Processing SDK to run sentiment analysis GITHUB Source Code: https://github.com/hoteit/sentiment- tweets Code Walkthrough 3 –Sentiment Analysis using pretrained Google Cloud NLP
  26. 26. Training Google AutoML for 4 hours & $5 to categorize Coursera students reviews Searched for a dataset on https://toolbox.google.com/da tasetsearch Found “100K+ Scraped Course Reviews from the Coursera Website (As of May 2017) Analyzed the data, cleaned when necessary (pretraining step) Created Google Cloud AutoML project & activated NLP APIs, uploaded data No AI expertise needed! Dataset import Train/Evaluate/Predict model GitHub Source Code https://github.com/hoteit/coursereviews-automl Code Walkthrough 4 – Course Reviews using AutoML
  27. 27. 27 Contact us Tarek Hoteit - @hoteit LinkedIn https://www.linkedin.com/in/hoteit Homepage http://tarek.computer Tarek.Hoteit@tr.com Katherine Li LinkedIn https://www.linkedin.com/in/shiqi-dartmouth/ Homepage https://katherine-shiqi.github.io katherine.li@tr.com