Facebook's 'Rosetta' AI Can Extract Text From a Billion Images Daily

Wow, cool to Facebook take such a major step towards dealing with some of its more unsavory challenges. Specifically “to give screen readers a way to read what's written on them and to make sure they don't contain hate speech and other words that violate the website's content policy.” Also notable are Facebook’s foray into language translation expansion. Could this be an indicator of bringing Wit.ai’s language library out of Beta?

Victoria Beckham Embarrassed After her Facebook Chatbot says 'I love Becks' Money And I'll be Happy to Divorce'

I don’t know what’s cringey-er, the poorly cobbled together pun in this article title, or the actual subject matter. This is an example of bad bots happening to good people and quite frankly, this is not ok in Q3 2018. Sure in the early days of 2016, we’d let this slide, but we know better now. People are mean to bots, they like to push limits, find the seams, “trick” it. You have to be prepared for the worst. The internet can be a scary place. That is not a variable. The way you program your bot  to preempt that behavior, and quickly iterate and optimize the stuff you missed, is. That’s what all bot makers should want for their bots, what they really, really want. 

Google Introduces New Search Engine for Finding Datasets - Search Engine Journal

Did you hear that? That’s the sound of the world getting a bit more magical. Love this huge step forward in making data more accessible to those who want to innovate. Now if only this could apply to NLP train

Hype Kills Value, and Other Hard Lessons from Veteran Voice App Developers

Great rundown of one of the best sessions from VentureBeat’s Transform AI conference a couple weeks ago. With insights from Google and a couple startups focused on creating voice AI, it’s a great, candid look into what’s real, what’s not, what’s hype, and what real value actually looks like (and how long it takes to appear.