AI as a service?
A few years ago I had the amazing experience of meeting Kevin Kelly, the founder of Wired Magazine. Towards the end of our conversation, he asked me about if I had heard about AI as a service. The question baffled me a bit. This was late 2013 and my only immediate response was that services like ASR or NLU could fall into that category. Or maybe Watson’s API?
However, after I left that meeting, the question stuck in my head like a thorn. AI as a service? What would it do? Why would we use it.
However, I started to break down our own intelligence and started to think about how certain technologies could augment our abilities.
Just a few weeks after that meeting, while at CES 2014, I came across a company called Beyond Verbal that has a service to identify emotional traits in speech, regardless of the language being spoken. The service could do things like tell an inexperienced phone operated how a caller was feeling while they were on the phone with them. This was an augmentation of the operator’s emotional intelligence.
The light bulb went off and all of sudden, it was clear that there were going to be dozens or even hundreds of different services coming out that could augment our own intelligence in some way.
At that time, we had been exposed to IBM Watson as part of their first global developer contest. We found the early API clunky but it was good at one thing — very quickly sorting through through text via natural language query. In that respect, it was similar to MindMeld.
Now, the AIs that are available come in many forms:
- Speech recognition (likely exceed human ability on a trained model this year)
- Natural language understanding (gathering intent out of a phrase)
- Language translation
- Emotion detection
- Writing tone analysis
- Personality type
- Text recognition
- Face recognition
- Image feature recognition (e.g. is it a boat, a bird, a landmark…)
- “Inappropriate content” visual recognition
- Facial expression detection
- Visual age estimation
…and many more.
This doesn’t include all of the tools that we can use to augment our own performance, such as email reminders through Boomerang (takes a huge cognitive load off your plate by bringing an email back to your inbox if the recipient doesn’t respond after a specific period of time) or booking time through Calendly.
IBM, Microsoft, and Google have all launched pay-as-you-go services. Amazon and Crowdflower have had for years a “human-in-the-loop” services like Mechanical Turk that could be used to train such systems.
Developers have just started to scratch the surface of the possibilities of using these different services. We’re going to see some very spooky-cool apps come out that will start to take advantage of these new abilities.
