O'Reilly Bots Podcast - O'Reilly Media Podcast

The O’Reilly Bots Podcast: The technical and social dynamics of solving scheduling problems.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I talk to Jason Laska and Michael Akilian of Clara Labs, creator of a virtual assistant—Clara—that schedules meetings and interacts in natural language through email. E-mail is, to me, a highly promising (and somewhat underrated) venue for bots. Messaging is growing quickly, but e-mail is still the standard way to communicate within businesses and especially between businesses. E-mail conventions are somewhat standardized, and much of it is highly routinized—automatically generated reports, receipts, etc.—so it’s ripe for automation. Laska, who leads the machine learning efforts at Clara Labs, and Akilian, the company’s co-founder and CTO, talk about the reality of developing an AI-driven product, and explain Clara’s human-in-the-loop system. “People are still there to do some of the most challenging aspects of this work, and that’s exactly what you want to use people for,” says Laska.Discussion points: How both the Clara bot and its users deal with the often-complex social dynamics of scheduling, which is “fundamentally a negotiation,” says Akilian. How Clara parses and evaluates dates and times Email vs. messaging as a platform for AI bots Other Links Jerry Chen’s article The New Moats: Why Systems of Intelligence are the Next Defensible Business Model Microsoft’s paper on developing calendar help (PDF) Google’s multilingual neural machine translation system Facebook’s Poncho bot for weather forecasts

The O’Reilly Bots Podcast: The social impact of Facebook.In this episode of the Bots Podcast, Chris Messina and I reflect on what Facebook has become, the role that it now plays in our lives, and what it all means for developers. We recorded this discussion shortly after attending Facebook’s F8 Developer Conference in San Jose.When it opened to users in 2004, Facebook’s essential value was exclusion—it was available at first only to Harvard students, then to students at a handful of top-tier universities. Since then, it has grown to host two billion monthly active users, and along the way has come to feel like a utility—a simple reality of digital existence. Messina, an independent bot enthusiast and social media observer (and creator of the hashtag) calls Facebook “a state of mind, a belief system. It is a way of participating in the common discourse that reinforces your own perceptions.” Discussion points: Augmented Reality (AR) and Virtual Reality (VR) dominated the discussion (especially the keynotes) at F8.  In addition to reviewing the new product/project announcements, we talk about what is being done to cultivate the skills that the next generation of AR/VR content creators will need. In an age of fake news, we discuss Facebook’s responsibilities. Despite some calls for Facebook to become a gatekeeper, the company seems to recognize that doing so could alienate a large portion of its user base—and interfere with the safe-harbor protection that it enjoys as an unedited platform. We compare Facebook and Snapchat, and note that Snapchat will face mounting pressure to open its platform to developers. The new features announced for Facebook Messenger, including the Discover tab and parametric QR codes, provide some interesting avenues for bot discovery, one of the most formidable challenges that bot developers face. How much natural language understanding do bots really need? There are plenty of existing processes that can be valuably brought to messaging platforms without really engaging with NLU at all. As Messina says, “the best NLU is still done by humans.”

Messaging as the operating system for the enterprise.In this episode of the Bots Podcast, we peer into the giant companies that are beginning to adopt messaging and bots. My guest is Tom Hadfield, founder of Message.io, a service that syndicates bots across many different messaging platforms.Hadfield argues that messaging and bots are the latest in a long evolution of communications technologies that have revolutionized the workplace—from the telegraph through e-mail—and that they are about to become commonplace at very large firms. There, they’ll do everything from monitoring customer feedback to giving employees access to their HR records. “Messaging can be the operating system for the enterprise,” says Hadfield. Links: Botness Enterprise, the conference that Hadfield hosted a few weeks ago The Botness UI Primitives survey, which I wrote along with other members of the Botness steering committee. It aims to collect ecosystem preferences on user interface affordances so that platform managers can standardize their offerings.

The O’Reilly Bots Podcast: Solutions from big data sets.In this episode of the O’Reilly Bots Podcast, I talk about deep learning at the extremes of scale and computing power with Prabhat, who leads the data and analytics group at Lawrence Berkeley National Laboratory’s supercomputing center. If you’re working on commercial AI, it’s worth glancing across the divide at scientific AI.Prabhat talks about his work at the the National Energy Research Scientific Computing Center (NERSC), including a project that aims to locate and quantify extreme weather events. He explains how this moves climate data analysis from a focus on core statistics—especially the change in the average mean temperature of the Earth in any given year—to analyzing the impact of extreme events.  He’s also working on the Celeste project, which uses telescope data to create a unified catalog of all objects in the visible universe. Looking ahead, Prabhat sees broad applications for deep learning in scientific research beyond climate science—especially in astronomy, cosmology, neuroscience, material science, and physics. Links: Prabhat’s new O’Reilly article, "A look at deep learning for science" Prabhat’s 2015 O’Reilly article "Big science problems, big data solutions" Prabhat’s presentation at Strata + Hadoop World 2016

The O’Reilly Bots Podcast: Conversational interfaces for the Internet of Things.In this episode of the O’Reilly Bots Podcast, I speak with Tom Coates, co-founder of Thington, a service layer for the Internet of Things. Thington provides a conversational, messaging-like interface for controlling devices like lights and thermostats, but it’s also conversational at a deeper level: its very architecture treats the interactions between different devices like a conversation, allowing devices to make announcements to any other device that cares to listen.Coates explains how Thington operates in a way analogous to social media; in fact, he calls it “a Twitter for devices.” Just as people engage with each other in a commons, devices chat with each other in Thington’s messaging commons. He also discusses the value of human-readable output and the challenges involved in writing human-understandable scripts. Other links: Coates’ blog post “The Shape of Things,” an overview of how connected devices will communicate with humans Google Translate’s interlingua The O’Reilly Artificial Intelligence conference, June 27-29, 2017, in New York  

The O’Reilly Bots Podcast: Automating “psyops” with AI-driven bots.In this episode of the O’Reilly Bots Podcast, I speak with Tim Hwang, an affiliated researcher at the Oxford Internet Institute, about AI-driven psyops bots and their capacity for social destabilization.Until recently, the psychological operations (psyops) conducted by governments and political organizations were mostly analog: dropping leaflets from airplanes, blasting radio messages across frontiers, planting stories with journalists, and dragging loudspeakers through city streets. Now, like some other forms of publishing, the practice of psyops is contemplating an online, AI-driven future in which swarms of carefully targeted bots disseminate information instantly. Compared to traditional psyops, AI-driven bots are highly scalable, offer sophisticated targeting capabilities, and are cheap to deploy—accessible to one-person organizations as well as great-power governments. Hwang is the author, with Lea Rosen, of “Harder, Better, Faster, Stronger: International Law and the Future of Online PsyOps (PDF),” published recently by the Oxford Internet Institute. He outlines a handful of conceptual “future scenarios” in which hostile actors might use bots to sow chaos—for instance, to find people who might be open to radicalization, or to misdirect crowds of bystanders during terrorist attacks. Hwang says existing legal frameworks aren’t sufficient to manage these threats, but we talk about three possible ways to address them: Governments come together to form an international body that brings transparency to the field by cataloging attacks and publicizing methods (a parallel to the INTERPOL approach for policing international crime) Governments pressure social media platforms to regulate and stop hostile psyops campaigns A social approach that emphasizes “media literacy” among the public Other Links: Harder, Better, Faster, Stronger: International Law and the Future of Online PsyOps—Tim Hwang's recent talk at Oxford on the topic of his new paper Bots spread misinformation during the Columbian Chemicals explosion hoax in 2014 The so-called “50 Cent Army,” which the Chinese government uses to discourage political activity O’Reilly’s Artificial Intelligence Conference, June 26-29, 2017

The O’Reilly Bots Podcast: Slack’s head of developer relations talks about what bots can bring to Slack channels.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Amir Shevat, head of developer relations at Slack and the author of the forthcoming O’Reilly book Designing Bots: Creating Conversational Experiences.We often talk about consumer bots on the podcast, but workplace bots are arguably a more attractive market for the time being. Companies are able to drive adoption by fiat (“all employees are now required to file TPS reports through the bot”), and bots can draw on large volumes of well-linked internal data in ERP systems, calendars, and so on. Slack is principally a workplace messaging platform, so we kick off our conversation with Shevat by talking about design considerations for workplace bots and bots that can work with groups of human users. We also cover the recent release of Slack Enterprise Grid, a new Slack offering for very large companies with up to half a million users. Discussion points: Developing bots for very large installations How bot developers can test bots for the enterprise Slack’s January release of threaded conversations and its impact on bot development (see Shevat’s VentureBeat post “Building better bots with threads” for more details) The state of conversational AI: Shevat describes two types of conversations—“topical” (for which a great deal of AI is necessary) and “task-led” (which needs less AI) Other Links: Eric Stromberg’s “Startup Idea Matrix,” which outlines markets and ways to create new products for them The O’Reilly Artificial Intelligence conference, June 27-29, 2017, in New York Videos of Siberian Huskies saying “I love you,” the inspiration for Shevat’s choice of an animal for the front cover of his book

The O’Reilly Bots Podcast: The 2017 bot outlook with one of the field’s early adopters.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Chris Messina, bot evangelist, creator of the hashtag, and, until recently, developer experience lead at Uber. We talk about the origins of MessinaBot, ruminate on the need for bots that truly exploit their medium rather than imitating older apps, and take a look at what’s ahead for bots in 2017.Discussion points: Traditional résumés provide the same comprehensive overview regardless of who’s reading; MessinaBot customizes the experience of learning about its creator, and draws together content that’s ordinarily spread across many channels. “Conversational commerce,” the messaging trend that Messina named in 2015 Messina’s work with Esther Crawford, who launched her résumé bot EstherBot in 2016 The state of bot skeuomorphism Product Hunt’s role in the popular emergence of bots For small businesses, setting up a bot interface could be easier than creating a traditional website. Until recently, Path’s Talk app even had a call center to route messages to businesses that didn’t have messaging capabilities.

The O’Reilly Bots Podcast: A universal bot for messaging, mobile voice, and the home.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Brad Abrams, group product manager of Google Assistant, the company’s new AI-driven bot that lives in many different contexts, including the Pixel phone, the Allo messaging app, and the Google Home voice-controlled speaker.Discussion points: “Actions,” Google’s API for Assistant plug-ins. These are available for Google Home now, and will be rolled out for other instances of Assistant later. The relationship between Assistant and Google’s Pixel phone. Google’s plans for the recently acquired API.ai, and where it fits in with Assistant. Google’s WaveNet technology, a text-to-voice engine that uses neural networks. Google’s voice user interface design guidelines, and how Google uses different voices in different settings.

The O’Reilly Bots Podcast: Recapping a revolutionary year in AI and bots, and looking ahead to 2017.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I look back at 2016, a big year for bots that saw important developments in platforms, tools, and underlying AI.We recap some of the biggest bot-related stories of 2016, including: People want to use messaging interfaces! Facebook Messenger, Kik, Slack, and Microsoft all made it easier for developers to create bots The field of conversational interfaces exploded Google’s open source TensorFlow library was a major accelerant for the field of deep learning Developments in workflow bots from Smooch to PagerDuty The emergence of god bots—Amazon Alexa, Google Assistant, Facebook M—and what they mean for bot discovery Microsoft’s new Zo bot, which uses “guide rails” to steer conversations away from some of the controversies that plagued Microsoft’s Tay bot The December introduction of Microsoft Teams We also present our predictions for 2017. Pete foresees progress in more seamless payment methods, ways to monetize bots, discovery and installation, and bot-to-bot communication. I expect further development in the micro-features of bots and improvements in NLU. I also think we’ll see more large organizations deploying their first bots next year.

The O’Reilly Bots Podcast: X.ai founder on personal assistant agents that schedule your meetings.In this episode of the O’Reilly Bots Podcast, I speak with Dennis Mortensen, founder and CEO of x.ai, a personal assistant bot that handles meeting scheduling through email. Discussion points: Social considerations for personal assistants and the bots that stand in for them x.ai users can call their virtual assistant either “Amy” or “Andrew.” Mortensen says users overwhelmingly choose the name of the opposite gender. (But for the most part, he says, it’s only male users who flirt with the bot.) The goal: agents that carry over conversations from one channel or platform to another (like continuing an email conversation in Slack) Humans in the loop: how x.ai’s 40 human “trainers” review scheduling-related data and train the response algorithm How the scheduling bot attempts to guide conversations in order to improve outcomes Other Links The Call for Speakers for the O'Reilly Artificial Intelligence Conference, June 26-29, 2017 After Dennis and I spoke, Microsoft launched a similar service that uses Cortana to schedule meetings.

The O’Reilly Bots Podcast: Making neural networks more accessible.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I talk with Richard Socher, chief scientist at Salesforce.  He was previously the founder and CEO of MetaMind, a deep learning startup that Salesforce acquired in 2016.  Socher also teaches the “Deep Learning for Natural Language Processing” course at Stanford University. Our conversation focuses on where deep learning and NLP are headed, and interesting current and near-future applications.Discussion points: Accessibility, in a couple of senses: making deep learning easier for computer scientists to implement, and making the power of deep learning available through intuitive applications AI-enabled question answering systems and dynamic co-attention networks The issue of interpretability, and progress in creating more interpretable models Why Socher believes that human-in-the-loop is the best solution for the current “fake news” controversy, the hottest topic in NLP now Why Quasi-Recurrent Neural Networks (QRNNs) are an advancement over Long Short Term Memory networks (LSTMs), the subject of a recent paper co-authored by Socher Other links: The Stanford Question Answering Dataset TensorFlow and Chainer, two frameworks for working with neural networks Summaries of recent papers by the Salesforce research team  

The O’Reilly Bots Podcast: An optimistic look at the future of bots.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Ben Brown, co-founder and CEO of Howdy.ai, the bot toolmaker behind the Botkit framework. Brown also runs the Talkabot conference, which was held in Austin this past September. Discussion points: The types of bots that people are building and experimenting with on Botkit How the evolution of bots compares to the evolution of mobile apps The strikingly collegial bots community: Brown cites the contributions that IBM and Microsoft have made to Botkit Praise for new types of bots, including notification bots (such as Poncho), what Brown calls “unfurling content” bots (such as Purple), and bots that open channels to discussions with humans Thington, an Internet of Things platform that functions like a chatroom for micro-bots connected to different home devices Other links: The Harvard Business Review bot, which sends out advice articles Bloomberg article on Slack’s partnership with IBM to build chatbots IBM’s “World of Watson” event, held in October

The O’Reilly Bots Podcast: Using data science to allocate campaign resources.The problem confronting a modern campaign manager is similar to the problem that any marketer encounters: how to spend finite resources to reach the right people and convince them to act. In this episode of the O’Reilly Bots Podcast, I talk data strategy with Andrew Therriault, chief data officer for the City of Boston. He was previously director of data science for the Democratic National Committee and is the editor of a free O’Reilly ebook called “Data and Democracy: How Political Data Science is Shaping the 2016 Elections.” We talk about how data is changing today’s political campaigns—particularly the way that campaigns now determine which potential supporters to target with phone calls, mailings, and door-to-door contact.Discussion points: How campaigns are combining different sources of data (including census and consumer data) to augment traditional voter registration information The similarities and differences between political and commercial data science The importance of forecasting and poll aggregation sites (such as FiveThirtyEight.com) in this era when “the polling industry is in a crisis,” says Therriault. How online polling will become increasingly relevant The increasing use of social media data in campaigns How deep learning and neural networks might be used in future campaigns

The O’Reilly Bots Podcast: Bots are the new web.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I recap O’Reilly Bot Day, held October 19, 2016, in San Francisco. The event gave us a good picture of what the bot community—and bot landscape—looks like, and the diverse group of attendees conveyed a strong sense of optimism about bots. Slides from Bot Day presentations are available here.We then speak with Shivon Zilis, partner at Bloomberg Beta, who has written extensively about artificial intelligence and its impact on a variety of industries. She wrote influential surveys of what she calls machine intelligence in 2014 and 2015, and she’s planning to publish her 2016 update at the end of October. Discussion points: Where “agents” (which Zilis calls “things we can trust that act as an extension of ourselves”) fit into the landscape of machine intelligence The impact that bots and AI agents will have on the future of work How the current AI hype cycle is creating, in Zilis’ words, “a tendency to throw AI at every problem” Can startups even compete in the AI landscape against large players like Google, Amazon, Microsoft, and Facebook? The impact of open source machine learning libraries like Google’s TensorFlow

The O’Reilly Bots Podcast: How bots are transforming the way companies interact with their customers.Customer service is a key application for bots—one of the first that we think of when we imagine a world full of AI-enabled conversational interfaces. In this episode of the O’Reilly Bots podcast, Pete Skomoroch and I talk with the founders of two companies that have developed bots to help consumers and companies talk to each other.We begin by hearing from Michael Schneider, founder and CEO of Service, which offers a bot that mediates disputes between customers and large companies. The bot interacts with the customer to gather basic data about the issue and the company, and then a human representative tries to solve the issue by contacting the company. Service is positioning itself as a neutral third party to, in Schneider’s words, “negotiate a fair resolution between consumers and business.” Service also has an intriguing idea for solving the discovery problem that many bot creators are facing: instead of waiting for users to find it on a bot directory somewhere, Service scans Twitter to find people who are complaining about customer service but haven’t been helped yet. Service then reaches out on Twitter with an offer to contact the problem company on the customer’s behalf. For our Bot of the Week segment, we speak with Rachel Law and Alyx Baldwin, co-founders of Kip, a bot that makes it easier for groups, especially those in the workplace, to make purchases through Slack. Kip assists personnel who have to manage group-wide orders (for office supplies, food, etc.) by automating the process of collecting requests from multiple people. Links: The recent Botness survey of bot builders See highlights from O’Reilly Bot Day, which Pete and I co-hosted on October 19, 2016 in San Francisco.

The O’Reilly Bots Podcast: Bots that can respond to groups of users.The workplace is a rich venue for bots that help with productivity and collaboration. In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I focus on workplace bots. We begin by talking with Jassim Latif, head of partnerships at Slack, about bots written by outside developers for scheduling, organizing meetings, and managing human resources. “We’re building toward a future where the value of these apps and services that people are building on top of Slack far outweigh the value of Slack on its own,” Latif says.You might know that Skomoroch is CEO of a bot startup called Skipflag, but he hasn’t given any public demonstrations of the product he’s building—until now. Rover, Skipflag’s bot, scans messages from Slack and “turns conversations into knowledge,” Skomoroch says. In this episode of the podcast, Skomoroch shows off Rover’s integration with the Amazon Echo. Join us at O’Reilly’s upcoming Bot Day on October 19, 2016, in San Francisco for insight on messaging bots, business bots, and more. Bot of the week We discuss Allo, Google’s new messaging application that includes Google’s AI assistant—appropriately called Assistant. The bot provides structured Google search results in a chat interface, and its capabilities blur the line between voice bots and text-based bots.  Skomoroch calls it “the best bot I’ve seen so far.” Links Api.ai, recently acquired by Google Apple’s iOS 10, which includes new bot-friendly features in iMessage and Siri. Siri is also now available in macOS (Apple’s desktop operating system formerly known as OS X).

The O’Reilly Bots Podcast: Hilary Mason, Jimi Smoot, and Roger Chen on what AI means now.Something remarkable is happening in the world of artificial intelligence. At the O’Reilly AI Conference in New York, people weren’t just talking about AI as a far-off dream; they were talking about AI as something that exists in real products today. In this episode of the O’Reilly Bots podcast, I talk with three artificial-intelligence practitioners about the real practice of AI: Hilary Mason, Jimi Smoot, and Roger Chen. Hilary Mason, founder and CEO of Fast Forward Labs, a startup that conducts research on machine intelligence, says that today’s AI “gives us a capability that would have seemed like magic even five years ago, and yet that capability is not nearly as interesting as the fact that the app is actually useful.” We also talk about the potential of AI-generated content, and some products that provide a glimpse of what a new AI-written world might look like. My second conversation is with Jimi Smoot, founder and CEO of Vesper, a hybrid AI and human assistant that helps executives with tasks like scheduling and travel arrangements. AI that augments human functions is likely to be a facet of the next economy. Smoot says that “early in the process with a new user, having a human touch is critical to developing trust.” Finally, Roger Chen, co-chair of the O’Reilly AI Conference, talks about what the term AI really means, the origins of the AI Conference, and how companies can implement AI now.  “I think a lot of these interfaces that we call AI and bots are just going to be known as seamless, great experiences and interfaces,” he says. O’Reilly’s upcoming Bot Day on October 19, 2016, in San Francisco, will provide more insight on AI for bots. Other links: Google’s "Deep Dream" paper, illustrating how a neural network can be used to turn an ordinary photograph into a dream-like composite Composing classical music using neural networks Video highlights from the O’Reilly AI Conference Brief, from Fast Forward Labs, a summarization engine that uses AI to extract the most interesting sentences from long passages of text

The O’Reilly Bots Podcast: Group interaction through social computing.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Lili Cheng, general manager of FUSE Labs at Microsoft Research. Cheng’s team is responsible for the Microsoft Bot Framework. She’s also a speaker at O’Reilly’s upcoming Bot Day on October 19, 2016, in San Francisco. Cheng talks about Microsoft’s experimental bots and their goal of making conversations playful and engaging. We also discuss the importance of designing good dialog; the potential of workplace bots; and Xiaoice, Microsoft’s popular Chinese chatbot. We also reflect on the fate and significance of Microsoft’s Tay bot.Bots of the week: This week, I discuss three bots that either augment or replace customer-service representatives. KLM Royal Dutch Airlines’ bot that gives you updates on your reservation through Facebook Messenger and provides a way to chat with a human representative “Macy’s on Call,” a web-based bot that provides in-store users with information on where store departments -- and the restrooms -- are located. Whole Foods Market Recipe bot, Facebook Messenger bot that gives shoppers an easy way to look up recipes on a mobile device while they are in the store. Other links: My Infographic on the Bot Platform Ecosystem WolframAlpha, the online computational knowledge engine that answers questions

The O’Reilly Bots Podcast: A look at some of the technologies behind the chatbot boom.In this episode of the O’Reilly Bots Podcast, Pete Skomoroch and I speak with Andy Mauro, co-founder and CEO of Automat, a startup whose tools make it easy to build AI-powered bots. (Disclosure: Automat is a portfolio company of O’Reilly AlphaTech Ventures, a VC firm affiliated with O’Reilly Media.) Mauro will be speaking at O’Reilly Bot Day on October 19, 2016, in San Francisco.“Anyone looking to get into this space should be looking at the emerging new platforms,” advises Mauro. “Pick the right channel, pick the vertical problem, and pick a platform that fits your strategic goals.” We discuss various bot platforms, including those developed by Facebook, Slack, and Microsoft, and Andy’s “sleeper pick,” Kik. Other discussion points: The development of conversational interfaces over the last 15 years The importance of language understanding “AI should be invisible to the end user,” says Mauro Why big companies are now talking about their bot and messaging strategies Other links: Apple’s 1987 “Knowledge Navigator” concept video KLM’s messenger bot for passengers