‘Imagine 10 years ago people would say 'if only I knew the internet was kicking off I would've bought nike.com' - the same thing is happening with AI right now. If you have ideas, you should start developing them right now.' Adam McMurchie, DevOps Solutions Manager, RBS
Over the last 5 years since our inaugural event in San Francisco, RE•WORK has hosted countless summits across the globe in AI, Machine Learning and Deep Learning, covering topics from healthcare to finance, and from retail to robotics. As the events grow, we receive an increasing amount of feedback from our attendees, which we carefully consider to curate the next generation of RE•WORK summits. This year alone, we have held events covering a number of new topics such as AI in Industrial Automation and Deep Learning for Robotics. This week in Houston, we have been excited to launch two new events with world-renowned speakers at the Machine Learning for DevOps Summit and the Applied AI Summit.
‘What is DevOps? I’m frustrated hearing it’s continuous integration and that’s it! It’s much more than that. It can be a process, but it’s more importantly a culture!’ Marios Fokaefs, Polytechnique Montréal
'The main pain points in AI adoption are caused by barriers between AI and the rest of the business. It's hard for us to progress without interaction between AI and the rest of the business. To make the ML code work, you need a huge structure. We can use business insights to improve AI through better problem definition.' Biao (Bill) Chang, Sr. Data Scientist, eBay
Across the two days as well as learning from over 60 speakers, there were panel discussions, workshops, interviews and podcast recordings, and an exhibition area. Attendees travelled from as far as South Africa and Australia to learn about some of the latest applications and progressions in ML and AI. It’s always fantastic to hear what our attendees have to say, and here are some of our highlights over the two days:
- “Love the diversity at the event, not just in people’s backgrounds, but also the wide range of Industries” - Cheyenne Cazaubon, Northwestern University
- “Great to learn about technologies that can help developing nations, especially in Africa. I hope RE•WORK will be there soon!” - Uboho Victor, Learn for Life
- “I think this a great event for professionals to learn from people in other fields and apply knowledge to their own focus. Really great event.” - Washington Mashingaidze, GCC
Both summits got stuck into the technical discussions straight away with industry and research examples being shared by speakers, with some great Q&A engagement as well as comments on social media. In the morning’s sessions, we also kicked off the workshops, starting with an open floor Q&A for those new to Machine Learning with this morning’s speakers answering those pressing questions. Some of the key topics were covered such as ethics, sustainability, accessibility and diversity, and in the exhibition area we were asking attendees some of the most pressing questions on everyone’s minds:
56% of people think that their companies are diverse, whilst only 6% think not.
50% of people think healthcare is the area that will be positively transformed by AI the most - other popular areas were education, safety and accessibility.
In the Applied AI Summit track room we were exploring AI transformation in business, with General Electric, GM Financial and eBay beginning the day with some case studies and applications. Joe Michaels from eBay explained how they’re ‘trying to avoid using user-generated data in our machine learning models to increase accuracy.’ Similarly in the DevOps room attendees were exploring the current landscape of ML and DevOps, and looked at the interaction between the two. Koshsuke Kawaguchi, CTO of CloudBees and the creator of Jenkins explained that his unique position means that ‘I get to see lots of real-world software development, so helping people become more productive is my number one motivation.’
What did we learn?
Applied AI Summit:
Hao Yi Ong, Lyft
Hao Yi Ong shared some interesting insights into the applications of AI in their business: ‘A small team that constantly develops better features and maintains a good ML model retraining pipeline will beat any army of analysts that manually handcrafts and manages hundreds of business rules. Most of the work you’re doing isn’t model development. We need to avoid building ad-hoc solutions to serve ML models. Real world ML systems require a really complicated non-modelling infrastructure that can be shared across teams.’
Coursera provides education to the masses and their goal is to make education fair and accessible globally. Vinod explained that ‘When we first tried to map content to skills we looked at mapping skills to content based on occurrences and embedding model. Then we tag them with supervised labels and trained the model from this data. It turns out there were a couple of challenges here e.g. limited coverage or skills and courses, and instructors having different notions of skills - and we realised we could do better!! We decided to crowdsource with our learners and now we train on real data.’
Kyle Tate, Shopify
Shopify is using machine learning to help in lowering the entry level into business. With Shopify capital, you join Shopify, start building your business, and you find out how much money you’re eligible for then you accept and it’s straight in your account. ‘We eliminate all the barriers that small businesses start in getting off the ground. This is built entirely by machine learning.’
Prakash Mall, Target
Counting people is a big use case for Target. Think about how many people walk out without any sales. AI is important in helping us assess this. 'Also we can use AI for out of stock detection - a camera can automatically alert someone in the store that an item is out of stock.'
Sunanda Parthasarathy, Wayfair
"My journey has been from a PhD in physics to a postdoc in quantum computers, into data science. Now I'm leading a team that creates and writes algorithms and solutions.' 'We are a tech company operating in the home furnishing space. It was founded by 2 engineers and we have 127 data scientists and 120 business analysts.”
Machine Learning for DevOps Summit:
Marios Fokaefs, Poly Mont
“It is my opinion that DevOps is a vision yet to be fully realised. It's becoming more and more valuable as a business driver, and we can address change in real time. AI can help us complete this vision and can help us explain, reason and trace this in real time.”
Kohsuke Kawaguchi, CloudBees
“The reality is that no person has any idea about the full extent of the possibilities of automation. We need to start trying to make sense of this in DevOps.”
Nicolas Brousse, Adobe
“Some of the findings that we have had, are that humans are not good at evaluating risk. Often they will bypass the risk and keep going.”
David Pierce, USAA
“AI for DevOps means digging into the meat of your pipelines and not throwing away data and using automated systems. The opposite is taking a systems approach and building it into ML.”
Diego Oppenheimer, Algorithmia
“What we do is very relevant in how DevOps teams organise themselves. If you have an ML model, Algorithmia can run it and scale it for you.”
Back for the second day, we were excited to dive deeper into the next generation of these technologies and learn about emerging applications in both AI and ML. We began the morning in the Applied AI Summit hearing about up and coming startups. Startups are an important part of AI progressions, and the session showcased exciting breakthroughs as well as some applications of AI and ML being used for social good. We heard from Stephen Odaibo, CEO & Founder of RETINA-AI who spoke about The Intersection of AI, Mobile Devices, and Healthcare who explained that ‘There is a huge shortage of retinal specialists, across the US which is the richest country in the world. AI can fix this.’ Additionally working in the healthcare system, Droice Labs shared how they ‘made a system that could predict when a patient could get sicker, and it was adopted by hundreds of doctors.’
The morning in DevOps looked at Operationalizing Data with Faiyadh Shahid, Research Engineer at EmbodyVR talking us through the current state of AI infrastructure for HRTF prediction. ‘Currently, in this process these are isolated silos and no-one has thought to bring them together. We are now trying to bring them together for the next-day station with the help of cloud computing. DevOps to the rescue! It has helped Research Engineers and Software Engineers come together as both code very differently.’
Running alongside the presentations were more workshop sessions, and Accenture ran an interactive session on designing ethical AI solutions. Accenture expert and Responsible AI lead, Greg Adams, led the design thinking and ideation workshop to illustrate how AI solutions can imbue ethics and responsibility. He asked ‘How do we proceed to create a machine learning algorithm, that can better improve the hiring process, and not make broad sweeping mistakes that eliminate strong unconventional candidates?’ Attendees helped help design an AI solution and provide guidance for the right kinds of ethical considerations. Each participant will leave with an understanding of applied ethical design.
What else did we learn today?
Applied AI Summit
Daniel Ellis, Reddit
Daniel spoke about how Reddit are using Machine Learning for better user experience and said that ‘at the end of the day, we need to remember that there are humans at the end of our product. Forgetting about the intricacies of the technology, simply our goal is to build something good.’
Julia Badger, NASA
One of the highlights of today was learning about Robotics and AI and exploring the future of people in space from Julia. She shared some of the exciting progressions in making autonomous robots capable of manning spaceships without human interaction. ‘The purpose of our project at the moment is to enable astronauts to move further out in space.’ At the end of the presentation, there was a really interesting Q&A session with one of the key questions being: How do you give some of what you’re doing back to be used in other industries? Julia explained that ‘we try to open source stuff, and we have something called tech transfer and we transfer it to US companies and licensing is available for lots of things we do!’
PANEL: Predictions for The Next Decade of AI Applications
Wrapping up the day we were joined by Amy Gross, Founder & CEO of VineSleuth, Sandeep Golkanda, Sr Spl Data Scientist at AT&T, Seshu Yalamanchili, Director for Artificial Intelligence Application Strategy at Visa, moderated by Brandon Ellett, Global Director at Hypergiant. The panellists enjoyed a glass of wine on stage, courtesy of Amy, who shared her exciting work in personalizing wine recommendations. Brandon led the discussion by asking the panellists where they see AI heading in the future:
Amy, VineSleuth: we need to move away from the dream phase of what AI can do. We'll start to see companies collect better data which is absolutely critical for getting good results.
Seshu, Visa: with NLP we're able to come up with solutions to deal with short tail responses quickly to apply in business.
Sandeep, AT&T: from an end to end perspective, the opportunities cover a huge scope, there are still so many challenges we're facing, and we're going to see machine learning integrated into everything.
Machine Learning for DevOps Summit
Chandni Sharma, Cloud Engineer at Google
Everyone was really interested to learn about how Google are using Kubeflow for Kubernetes, and Chandni explained her unique position: ‘my specialist is around AI and big data, and I’ve now grown into DevOps. I’m the person who’s always been on a laptop writing your AI, and now I’ve finally moved to the cloud! I’m now right in the middle of ML and DevOps.’ She asked, who can actually use AI today? ‘Few users can create a custom ML model so we need to make AI accessible to millions more. Our aim is to help as many people as possible.’ We heard how Google are using Kubernetes for machine learning, packaging it into Kubeflow, ‘Since the launch the momentum was unexpectedly high. We have 70+ community contributors over 17+ countries. From November 2017 to November 2018 there are continuous contributors way larger.’
@ablythe TIL: about Kubeflow https://www.kubeflow.org/ - a Cloud Native platform for machine learning based on Google’s internal machine learning pipelines from Chandni Sharma at @googlecloud at the @reworkDEVOPS conference #reworkAI #reworkDEVOPS
Praveen Hirsave, HomeAway.com
Praveen started off also by asking ‘who is here for DevOps for Machine Learning, and who is here for Machine Learning for DevOps?’ There was a 50/50 show of hands which started the discussion nicely. He explained that ‘home away have graduated from agile to DevOps and it this journey has given us an understanding of how to build an assembly line much better.’
PANEL: Challenges & Opportunities of Investing in Machine Learning
Moderating the session, Clint Wheelock, Founder & Managing Director at Tractica began by introducing the participants and explaining that ‘data scientists are a tiny slither of all developers so finding talent is difficult.’ Lynn Calvo, AVP of Emerging Data Technology at GM Financial, Aaron Blythe, Enterprise Architect at NAIC and Juni Mukherjee, Product Marketer at CloudBees all shared some of the most pressing challenges in the mass roll out and investment of ML.
What did some of our attendees have to say about the presentations?
@watersgisele @DroiceLabs Helps hospitals make sense out of disintegrated patient records to improve #healthcare At #ReworkAI they shared how to cement adoption of #AI 4 doctors to consider AI treatment recommendations long term. Understanding #clinical #workflow was essential! @bihgieee_isto
@ablythe Fascinating to see how engineers at @Spotify are approaching using CI/CD from Haresh Chudgar at @reworkDEVOPS #reworkDEVOPS #reworkAI basically what I gathered is train and deploy, train and deploy... machine learning engineers are on-call, not just creating models.
@AJStroud_APQC@SathRao, Director of Digitsl Solutions at Hitachi - Only 3% of data generated by machines (shop floor data) is currently being utilized. Data from shop floors tends to be generated in silos. You have connect employees with that data. #reworkai
@TBM4W, Watching Daniel Ellis, tech lead with @reddit speak at @teamrework #appliedai #reworkai in #houston #tx. @ JW Marriott Houston Downtown https://www.instagram.com/p/Bq0FOtsApey/?utm_source=ig_twitter_share&igshid=clmdky3cpgl8 …
What else did you miss?
Over the two days, we recorded interviews with several speakers for our digital content platform, including our Women in AI Podcast. Eunice Chendjou, Founder at DataGig live streamed her recording where she spoke about her journey founding a startup, as well as the digital apprenticeship marketplace for the nextGen data professionals. We were also lucky enough to get behind the scenes updates from Lyft, NASA, eBay, Shopify, Target and more.
Margaret Mayer from Capital One joined us on the podcast and shed some light on supporting women and encouraging diversity in AI:
“I’ve had situations where I’ve been mistaken for an event organiser rather than a speaker, just because peoples biases take over. At this Summit it’s great to have a balance of genders both speaking and attending. It’s great to create a community where women can support each other, and also encouraging the women in tech to go to the conferences where they can feel that there’s a group of these women who all showed up!”
As the last summit for 2018, the RE•WORK team are now looking forward to the world’s biggest Deep Learning Summit in San Francisco in the new year on January 24 - 25. The event will feature 10 stages and over 90 expert speakers covering topics such as sustainability, ethics, AI assistants, reinforcement learning, AI for social good, NLP and many more.