26 May Twilio Signal Summary
Who said telephony is boring? The Twilio Signal Conference was this week in San Francisco, and it was quite interesting for a company whose primary purpose is to give developers an entryway into the sometimes dull world of VOIP and telephony.
Twilio’s strengths are how well they communicate with developers, the strength of their API’s for SMS messaging and programmable telephone networks, and how they manage to make all of this glamorous and exciting.
That was certainly true in the two mornings of keynotes, which have already been posted to Youtube. Those two long videos are full of marketing and interesting announcements, and below them I am also including some of my top takeaways from attending the Signal Conference.
Day 1 Keynotes (Includes announcements of Programmable wireless SIM cards for IoT, and a nice IoT demo at the end with disco lights)
Day 2 Keynotes (Includes some interesting use cases including how SMS is being used to save lives on a suicide hotline for youth)
For more information on those large announcements, you can check out these links on Twilio:
Twilio Programmable Wireless (still in Preview):
Cellular data, SMS, and voice communications for connected devices. Provisioned and managed via a developer-friendly API.
Twilio Add-Ons: Access pre-integrated partner technologies through the Twilio API. You can also submit your own products to be vetted as an add-on.
Twilio Notify: Orchestrate notifications over SMS, push, messaging apps, and more with a user-centric API.
Here are some of the highlights from the conference for me (Note: I will be updating these and adding more over the next couple of days).
Conversational Bots are not easy, but they are real
I attended a number of conversations about bots. Since this is a Twilio conference, most of the discussions assumed a text based bot over SMS, but the principles are the same in a variety of settings. These talks got me thinking about what the intersection of bots and WebRTC will be, which I blogged about on WebRTC.ventures.
James Thomas of IBM demo’d a conversational movie service demonstrating how to build a cognitive bot with IBM’s Watson platform. You can see that demo here.
James made the point that building bots is not that hard – several speakers mentioned that the basic idea of simple bots has been around for a while and it’s not programmatically that hard to reply with something useful based on keywords a customer uses. The hard part is understanding natural language, and that’s what the newest generation of Natural Language Understanding and Artificial Intelligence tools from IBM Watson and others are trying to solve.
The basic way to make a bot understand natural language is to use a tool like Dialog from IBM Watson, which will extract the key entities of data from a conversation and put it into a conversation profile. Then you train the bot to match those entities to a class of actions in your application so that the bot can call pieces of code to get key information the customer wants. This training process will require thousands of data examples to teach the bot sufficient understanding of what your users need.
Here are the references from James’ presentation on Building Cognitive Bots.
Ben Brown from Howdy talked about building an omni-channel bot. Their botkit for Slack, Messenger and Twilio IP messaging lets you build the brain of your bot without worrying about which channel you are on or how the messages are being received.
Matt Makai from Twilio also did a demo of phone bots in Slack. Along with Don Goodman-Wilson from Slack, they showed how a Slackbot could send out SMS messages and make automated phone calls to a Slack channel in order to coordinate an event or product launch. This Twilio blog post shows their code on how to Add Phone Calling Bots to Slack with Python.
Alex Godin from slash-hyphen gave a very entertaining talk on 22 rules for Bots, which gave a bunch of tips for building successful bots. Alex sarcastically joked that just like every startup used to need a mobile strategy, “and now every fucking startup needs a bot strategy!”
Some of my favorite tips from Alex were to build bots that serve a very specific niche in order to make it most useful. Make it clear to users what a bot can and cannot do. Take advantage of the contextual data around the bot to make your bot more intelligent (ie, you’re in a chat room so you should know something about the people in the room). Bots should have simple interactions, and sometimes you should even consider a button-based interface for simple functionality instead of the complexity of natural language understanding.
Talks on WebRTC and Video
The Signal conference was great, however as a WebRTC fanboy I was disappointed not to hear any big announcements about Twilio’s Video product in the keynotes. There were smaller announcements in some of the talks, but the big announcements were all about Twilio’s core products of SMS/VOIP, and their play into IoT with Programmable Wireless.
Considering that Twilio announced an IPO the day after the conference, that shows where they see the most investor value in their business plan. Which makes sense, Video can be a powerful offering from Twilio but it doesn’t appear to be the most strategically important to them.
Nonetheless, there were some interesting sessions on WebRTC and Twilio Video.
Al Cook from Twilio talked about Contextual Communications and made an excellent statement about what WebRTC really is. “WebRTC is just a printer driver,” said Al. This means that WebRTC is ultimately a transport layer, and what you do with it will create value.
That’s a great attitude that has paid off for Twilio in the past. SMS text messaging is really just a transport protocol, but Twilio has done an excellent job over the years making it glamorous and showing developers how easy it can be to add strategic value to a business by adding in SMS. A similar attitude will benefit those of us building WebRTC based video applications.
Al’s larger point about contextual communications is “Don’t Eject Users” – take the contextual data about where a user is in your web or mobile app, and use that to speed up the Video conversation. Calls should not start with – “How can I help you? Can I have your name and account number?” Instead, they should start with “Hi Arin, I see that you are trying to dispute a charge on your credit card, how can I help you?”
Whether it’s SMS or WebRTC Video, in the future Al pointed out that all apps will have communications built in.
Dan Jenkins from Nimblea.pe gave a nice overview of WebRTC, and also pitched that if you’ve tried WebRTC in the past and failed, now is the time to give it another try. The community is stronger, the standard itself as well as the browser implementations are more stable, documentation and examples are better, and it’s clear that WebRTC is here to stay.
Chris Eagleston and Iñaqui Delgado showed off some of the latest additions to Twilio Video SDK’s for mobile on iOS and Android. Chris gets the Iron Man award for doing two presentations on mobile WebRTC back to back.
Iñaqui and Chris showed how you can do screen capture on Android with Twilio Video, and Iñaqui played a level of Angry Birds for us mid-presentation, which Chris streamed from his laptop via Twilio Video. In case you’re wondering, Iñaqui did complete the level successfully despite the pressure of an audience watching. That man is cool as a cucumber.
Chris pointed out that due to limitations of iOS, you cannot screen share your full desktop or other apps from iOS, you can only stream the view of your app.
They also showed off improvements to how Twilio Video handles rotation of the phone, with more seamless adjustment of the video orientation. The Camera Capturer features that they showed are available now, and the Screen Capturer functionality is expected to be released in June 2016.
WebRTC and video in general have advanced beyond basic video chat. There were a number of vendors who showcased how their technology lets you detect the sentiment of someone’s text messages or their face. Boisy from Affectiva gave a nice demo of their API with Twilio Video during a session on Emotionally Aware Video.
The Twilio code is written to capture a frame of the video as an image on a regular basis (they recommended every 5 seconds), and send that to the Affectiva API for analysis. Affectiva quickly sends back a Valence measurement, which is basically how happy you are versus how sad your face looks.
In the Twilio code, they matched this to emoji’s and made a nice live demo of Boisy making faces at the camera, with pretty accurate emoji representations following quickly thereafter.
It’s about solutions, not technology
One of Twilio’s strong suits is communicating well with developers about their API offerings. They also did a good job at the conference of showcasing solutions, because while developers care about technology, our customers only care about solutions.
My favorite “use case” talk was by Dr Precious Lunga, who talked about SMS in Africa. In partnership with local telco’s, she has built services for sending SMS health tips to their customers, and relayed stories of how those tips have both improved and saved lives.
With over 90% mobile penetration in Africa, and smart phone usage growing rapidly, this presentation was a powerful reminder of what communications developers can do to improve the world around them.
I would definitely recommend the Signal conference to others, or check out more of their videos on YouTube as the talks are released publicly.