Giving the WSJ a lot of love this week, but this really struck us because it’s something anyone who’s working on or touches NLP has probably endured: training data. This article takes a look at how AI companies out of China, for instance, have the benefit of “access to Chinese security databases comprising millions of people that it could use to train and refine its algorithms.” Something States-side we simply don’t have in abundance. Anyone aspiring entrepreneurs out there want to take on the challenge of training data access??
We’ve written about this before--Do Not Pay bot which files lawsuits for you--and were really amused/intrigued by it then. Now, 6 months later, this bot is proving that it’s more than a flash in the pan, but actually working and delivering payouts from the lawsuits filed! We love seeing ideas like this maturing into actual, tangible value. What other painful processes can be made less painful through a bot?
Apple edging its way into the B2C communication business. Seems interesting and compelling enough for the North American market, but if you’re interested in targeting outside of that, your business might find this limiting. It remains to be seen whether this will cause any user confusion--between FB Messenger, Twitter, Whatsapp, and Google--how will consumers keep track of which apps support their business’s bots?
Based on a Pew Research study (not pay walled, thankfully), this article digs into the general sentiment around AI and automation. Lots of good, meaty stats in there, but our favorite is quote from MIT professor, Sandy Pentland, who perfectly articulates basically what we’ve been saying all long“AI systems are often not used because they are built as black boxes and do not provide transparency into what assumptions and decisions the underlying algorithms are making on the user’s behalf. For humans to effectively team with automation, displays are needed that can explain and visualize what decision criteria the system uses.” The biggest challenge--and opportunity--of AI is that it shouldn’t be a developer-only environment. It’s not a tech problem, it’s a human one.