Hi! It’s been a busy few weeks for us, presenting at 4YFN in Barcelona, at the Smart Cities expo in Madrid, and a panel at Social Media Week in NY--thanks for your patience as we were literally in 3 places at once. Back again and excited to talk bots. Let’s dive right in!
And, if you’ll be SXSW, let’s meet up! Drop us a line at SXSW@Reply.ai
Messenger announced platform updates and everyone has something to say about it. Lots to unpack here, worth a read, and more importantly, to take a moment to formulate opinions. Our thought is these changes were necessary for a number of reasons--users stil need to be trained in how to use a bot (that’s why we now include a brief tutorial at the beginning of all ours--see for yourself with our Kia Niro bot). But it also comes with drawbacks, that might lead to the proliferation of very simple subscription-based bots based on RSS, Youtube, etc. that may damage the concept of bots as just a dummy automated marketing tool. The writer of this article seems perplexed by the option to remove the text input field for certain questions, but we think this is actually pretty useful (and something we’d already implemented with our own web widget months ago). Little steps like this nudge users in the direction they should be going vs. confusing them with too many possibilities. Now if only Messenger could do more to highlight its code scanner, that would be great. More thoughts on this soon.
NLP is getting better, but still far from real natural language understanding. The article gives this example of this sentence: “The city councilmen refused the demonstrators a permit because they xxx violence.” If `xxx = feared`, it's clear that `they = the demonstrators`. However, if `xxx = advocated`, then any human would understand that `they = the city councilmen`. Current tech is pretty far from getting this one right.
Number one question we get from clients is around discoverability. So we appreciate this article examining advertising options for bot creators. Specifically the power of being able to direct a user from Facebook Ad into chat with the bot. This is significant for many reasons: it’s native, easy, and you’re able to leverage all the highly sophisticated targeting FB ads have to offer. This article gives an example from a bot they made for UFC--with $0.07 per click, then they had a 70% conversion of click to actual engagement. Which is surprisingly good.
A lovely article (full of fun graphics and literary references) written by one of Typeform’s Storytellers. A great retrospective of A.I. and its intersections with art. Great quotes in here, and stats like “in China, people send voice messages to each other rather than text. Why? According to this Quartz article, “…because the [Chinese] language is notoriously hard to type.” Important to look back not just on this last year of Chatbot learnings, but on the topic as a whole as humanity has discussed and played with it.