327: Data Science Trends for 2020

Episode Summary

Executive Summary: The episode reviews five major AI trends for 2020: robotic process automation, natural language processing, reinforcement learning, edge computing, and open-source AI frameworks. Kirill and Adlan explain each trend, why it matters now, and cite real-world corporate examples showing how these technologies cut costs, improve speed, and unlock new products and services.

Main Topics: Robotic Process Automation (RPA): Software bots automate repetitive, rules-based office work such as invoicing, onboarding, billing, payroll, reporting, and logistics. The hosts emphasize that RPA is one of the simplest AI technologies but can produce major operational impact. Natural Language Processing (NLP): NLP is presented as a fast-growing AI area for understanding and generating human language, enabled by breakthroughs like BERT and driven by demand for chatbots, sentiment analysis, translation, summarization, and captioning. Reinforcement Learning: The episode highlights reinforcement learning as a reward-based approach that learns through trial and error, with strong applications in games, self-driving cars, ad optimization, recommendation systems, and customer value optimization. Edge Computing: Edge computing moves AI and data processing closer to devices and local nodes instead of distant cloud servers, reducing latency, improving privacy, and enabling real-time applications like voice assistants, IoT, and AR. Open Source AI Frameworks: TensorFlow, Keras, PyTorch, and related open-source frameworks are framed as a key enabler of rapid AI research and deployment because they lower barriers to building, training, and evaluating models. AI Workforce and Business Impact: Across all five trends, the hosts argue AI will displace some tasks but create more jobs overall, while improving efficiency, savings, and new capabilities for businesses and workers.

Key Arguments: RPA automates high-volume, repetitive manual work and frees humans for higher-value tasks. AI adoption should be seen less as a job destroyer and more as an industrial transition that creates new roles; the hosts cite the World Economic Forum view that more jobs will be created than displaced. NLP is accelerating because of research breakthroughs like BERT and because consumers increasingly prefer natural language interfaces over typing. Reinforcement learning is valuable because it can optimize decisions under a reward system and has already produced major business gains in ads, energy, and personalization. Edge computing will become increasingly important as devices, IoT, and real-time systems need local processing for speed, reliability, and privacy. Open-source frameworks are central to AI progress because they allow broad participation, faster experimentation, and easier model development. The five trends were selected because they are already influencing industry and are likely to matter most in 2020 and beyond.

Data Points: Jobs automatable by RPA: 230 million - Estimated number of jobs worldwide that can be automated, cited as about 9% of the global workforce. Global workforce share automatable: 9% - Approximate share of the global workforce that can be automated. Jobs created per job displaced: 1.7 new jobs per 1 job displaced - World Economic Forum 2018 estimate used to argue AI creates more jobs than it removes. Tableau revenue in 2018: $1.2 billion - Mentioned to illustrate the growth of visualization technology after the first Tableau release. TensorFlow initial release date: November 9, 2015 - Used as a milestone for the AI boom in machine learning frameworks. Synergy error reduction: 99% reduction - RPA use case in billing/invoicing at an energy company in Western Australia. Synergy speed improvement: 40% faster - Outcome from deploying 280 bots for billing automation. Synergy annual savings: $2.3 million - Savings reported from the RPA implementation. Synergy ROI: 163% - Return on investment from the automation project. Deutsche Bank task automation range: 30% to 70% - Range of tasks they can automate in experiments with RPA. AT&T software robots: Over 1,000 - Scale of AT&T's RPA deployment. AT&T training reach: 2,000 employees in 100 organizations - AT&T has trained others to deploy RPA. Vanguard assets under management: $3.2 trillion - Used as a large-scale example of RPA in financial services. Alibaba ad ROI improvement: 240% - Reinforcement learning increased return on investment in online advertising without increasing budget. Google data center energy reduction: More than 30% - DeepMind reinforcement learning reduced energy consumption in data centers. StarCraft result: 5-0 - DeepMind/AlphaGo Zero reportedly beat top StarCraft professionals by five games to zero. Manual work eliminated at Diehan: 5,000 hours - Edge computing reduced manual entry by detecting product defects faster. Amazon Alexa edge chip RAM: 1 MB - A low-memory chip that can listen for the wake word locally as a first step toward on-device processing. Commercial loan agreements interpreted by COIN: 360,000 hours - JP Morgan Chase’s NLP system reportedly replaced lawyer time needed to interpret contracts.

Pivotal Quotes: "for every one job that AI displaces in the coming years, 1.7 new jobs will be created" — Kirill / World Economic Forum reference: Used to reassure listeners that automation creates more opportunities than it removes. "Cloud computing is big data, but edge computing is instant data." — Kirill: A concise framing of why edge computing matters for latency-sensitive AI applications. "we see artificial intelligence as adding value to the world" — Kirill: Summarizes the episode’s overall pro-AI stance and business-focused perspective.

Implications: Listeners should expect AI adoption to accelerate through automation, language interfaces, smarter optimization, local processing, and open tooling. Businesses that experiment early can cut costs, improve service, and gain competitive advantage.