Conversations with AI

Experiment with smart assistants

Intelligent assistants serve us daily as a humble weather alerts, radio and music players, alarm setters, smart home hubs and yet can do a few other things (check out Google’s and Alexa’s cheat sheet). Nonetheless, those skills are only simple question/answer operations, with little or no context at all (place/time/date is the most common added variable for their QA operations).

Google introduced smarter assistant features a few of years ago during Google I/O, capable of maintaining longer conversations where the follow-up questions narrow down a field of search, and the assistant was able to lead the negotiation over a phone call and obtain a desired result. That sounded like a real break-through. The Touring test would definitely have been defeated with the verbal agility and an added non-verbal “mhm…” between the phrases. 

Alexa doesn’t shock here, but also gained a skill of “conversation mode”, and opened an API that enables developers to build additional skill sets that bring new functionality to its end users. Not that ground-breaking yet, but with the potential Amazon has, we can see more in the nearby future.

These assistants may not be the only manifestations of conversational AI in our daily life. According to Deloitte’s report the conversational AI has been gaining a lot of traction in the recent years. The number of patents related to context-based conversations and subject personalization, profiling has been growing. This indicates conversational AI could be out there in the field already.

I’ve personally have been receiving calls from companies advertising products, but not just automated calls. The speaker on the other side was responding to my answers in a very smooth, however not a spontaneous way. After getting suspicions that it may actually be an “AI” I asked her if she’s a person or a robot, she responded “oh, it may be because I’ve been working so many hours” (neither yes or no?).

So to check if the house AI assistants are actually getting closer to being our dear “Samanthas”, (like the voice of Scarlett Johansson in the “Her” movie), I’ve enabled the “Continued conversation” and “Conversation mode” for Google Home and Alexa respectively, and started asking some questions, hoping to get to a more sophisticated answers.

First Google Home (Nest). While I know Google Assistant from the mobile phone, I know that it sometimes falls back too easily to show search results when I want it to take some action, like navigate or open and play something in an app, I expected to see something more in the stand-alone assistant. Well, it certainly didn’t call anyone on my behalf (the functionality is only available in selected countries for now, not in Spain where I live), and also disappointed me with a very little capacity to retain the context of the previous questions/answers. The best I have discovered so far outside of this experiment is the continuous translation that allows two people talk almost like with the “babelfish” inside the ear. For the most time Google’s AI resulted in being repetitive or not responding at all. Still does the job every day, but doesn’t exceed the expectations.

Second, Alexa. Amazon’s AI surprised me with having a bit more initiative to provide answers even if there was some lack of understanding. I have discovered that with the added skills, even if they are pretty basic for the moment, you can get a bit further, but still not close to the real human skills. I got some unexpected answers though. Made me laugh at first but then again that was also better than a plain “Google tell me a joke”. What I can definitely say is that this second assistant showed much more personality (even if accidental). 

And I’ve left some cherry topping for the end, and tried to pair the two assistants, to see if they would be able to collaborate, either by asking each other to complete the tasks or even start a conversation. How rich would that be, not only to have a pet assistant, but the whole community at home. The result a little unexpected, but very positive gives a faith in the AI, which probably won’t try to eradicate us  from planet Earth (i.e. “Skynet”).

Leave a Reply

Your email address will not be published.