Onboarding is the process of integrating a new employee and getting them the tools and information they need to become a productive member of the team. Given that AI, and in particular its large subset, machine learning, will be joining your company soon, if it hasn't already, successfully onboarding AI will be critical to many firms.
CES 2020 featured lots of smartwashed products: "smart" coffeemakers, washing machines, etc. Really? Not much different than the promise of "fuzzy logic" a decade ago, and about as important.
The nascent trend: figuring out how to build machine learning and AI into our daily lives without forcing average people to learn the AI equivalent of a command line interface. AI doesn't replace people - not yet anyway - it augments them. The challenge is getting the proper fit.
Large scale AI today resides in the cloud, powering apps on smartphones. But that's the mass market. We have to learn how to integrate AI into ever-smaller niches over the next several decades, much as we've integrated computers into pretty much everything since 1950.
William Gibson says the future is already here, it just isn't evenly distributed. My corollary: the future is already here, it just isn't integrated.
What I saw at CES 2020 were all the elements needed to onboard AI. But no one has put all the pieces together. Those pieces include IoT data sources, edge devices and apps, business tools, and the UI.
IoT, the feedstock for AI
http://www.murata.com was showing off its range of low-cost edge sensors, along with the networking and software, required to instrument any physical plant - buildings, ports, hospitals, offices, retail - to feed needed data to a range of AI subsystems. USIS was highlighting its drone surveillance systems, which include remote, automated landing and charging stations for unmanned operations.
The most tantalizing IoT tech was Smart Dust, from Epic Semiconductors, which claimed "A Battery-Free, Energy Harvesting Sensor with Integrated Pre-Quantum Superposition Processor and Artificial Intelligence that Communicates Wirelessly and Bidirectionally to the Cloud (RF-Free & Zero Power)." The sensors are tiny cubes, 0.3mm on a side.
I'm not sure Smart Dust is real - the marketing hype is over the top - but there's no reason such a device couldn't be built. Maybe CES 2021 will offer more proof.
Edge devices and software
On the hardware front, Google and Kneron were promoting their low-power AI engines for edge computing. Google was pushing its Coral tensor processing unit (TPU) ASIC for machine vision.
Kneron offers an AI co-processor system on a chip that includes a Dual ARM CPU and a neural processor. They offer visual recognition SDKs for recognizing faces, objects, scenes, and gestures.
On the facial recognition front, CyberLink offers their FaceMe cross platform facial recognition engine. They claim a better than 97 percent accuracy rate and it runs on iOS, Android, Linux, and Windows.
Prevision.io, whose motto is AI-first, promises to automate your data science with an end-to-end platform, on cloud or on-premise. Enables you to create and analyze predictive models and deploy them in real business applications dashboards or smartapps.
Commerce.ai offers a "Deep Product Learning" platform that can read, see, hear, and understand product feedback from websites, product reviews, customer reviews and videos, and more. The platform digests the data to help product teams discover new markets and document product requirements.
Virtual humans: the UI for AI
Our "social" networks seem to have left us more isolated than ever. Amazing how shifting our gaze from a face to a screen can do that.
Technology's answer: a screen with a fake human on it. Samsung made a stir with Neon, a virtual human intended as an interface to AI systems. I don't know if it will work, but it's clearly an idea worth exploring as machine learning systems grow in expertise and usefulness.
Gatebox offers a personal version of Neon, intended as a companion. Well done, but I'm left feeling sad for the people who don't have real humans to to talk to. But as birth rates drop, populations age, and helping millions of older people manage on their own becomes a real and pressing challenge.
The Storage Bits take
I'm a professional sceptic, and I'm sure some of these companies are claiming more than they can deliver. But the needs they are responding to are real. If AI is going to radically improve human productivity, hundreds of millions of people are going to have to get comfortable with using it every day.
Today AI is where computers were in the 60s, when you needed to know FORTRAN or COBOL to get work done. In the coming decades we need to go from that to conversational integration that non-specialists can use, much as billions of people use smartphones to get real work done.
That's the challenge. And the opportunity.
Comments welcome. How would YOU onboard and AI?