MEETING RECAP: IoT Forum on Smart Home
WIth thanks to our host, Nokia, the IoT Forum meeting this month focused on the subject of Smart Home. We had a room-wide UNpanel discussion, followed by fast pitches from a bevy of startups.
The UNpanel discussed the issues of importance to the audience members, including:
- What are the killer apps?
- UI, ease of use
- Protocols, connection technology
Among the information that emerged from the discussion, were answers to the killer app question, with Voice Assistants being the obvious success so far. But panelists Tali Chen, CMO at DSP Group, Arsham Hatambeiki, SVP Product & Technology at Universal Electronics Inc, and Daniel Zhao, Director at China Mobile took it further, discussing ideas such as Home Security, Energy Efficiency, Aging in Place, Insurance, Content Discovery, Ambient Computing, and even Biophilia – the connection between humans and natural environments.
We seldom arrive at concrete answers, but the room seemed fairly confident that the question of what is the human interface for the Smarthome most likely falls into one of these four categories:
- Voice Assistant, natural language
- phone or tablet
- Wall Panel, for legacy AND fallback
And one of the most interesting take-aways was the notion that we currently have “obedient homes” and “connected homes”, meaning they will do what we tell them to do, fairly well, if we’ve set that up. But to be truly “smart” we must wait for the era of automation, which we agreed was a few years away before it’s truly done well. However, a there are two paths to the “automated, smart home”, and one is bad while the other is good.
The bad path is automation, where the automation makes frequent errors in what you might want, playing the wrong music, setting the wrong temperature, or switching the wrong lights. This is not unlike the bad speech recognition we all had in our cars in the recent past.
Instead of bad automation, the good path is to evolve to automation via suggestions rather than automations. Instead of just dropping your thermostat because it “thinks” you’re not home, if not > 99% certain that’s the right choice, a home could offer that suggestion, making it easy for the human to accept or decline. This suggestion/acceptance model could deliver value in machine learning, as the AI slowly learns what was right and what was wrong. This is a lot like how Tesla cars learned to AutoPilot by “shadowing” human Tesla drivers for a year of Machine Learning before launching the trained AI.
Thank you to everyone who contributed to the discussion, to our expert panel, and to the startups who introduced their propositions during the fast pitches. As always, presentations are available to our members via the Member Library, with our demo table elevator pitches available on our YouTube channel.
Meeting Website Links: