A few years ago, I was on a flight to attend a technology conference overseas. I had to fill up a passenger card on the plane, on a piece of paper, with a pen. Just let that sink in for a bit. Have you ever wondered why humans are still doing it this way to this day? It has been nearly a hundred years since the first government-issued travel identity papers were given to residents of late Ottoman Empire. By now, other travel technologies have reached a zenith level of mastery in physics – for example, propulsion engines to allow humans to fly like birds, but farther, faster, and safer.
A bubble popped up above my head.
All this information about me as a human has already been securely stored somewhere digitally. If we can make self-driving cars, and aspire to become spacefaring species, there has got to be a better way to do all this. Are we really still going to do this whole paper form and physical ID thing on Mars? Could this not be accurately and seamlessly sent via an API? Is the infrastructure insufficient? Are there no products that are good enough to solve this? Are we not suffering from enough pain to make it better right now, rather than a decade later? This should exist. In fact, some of these thoughts were what made us decide to start GTRIIP.
Are you who you say you are?
Just ask yourself this. How do you know that your spouse is really your spouse when you meet each other after work? I definitely remember how my wife looks like. If she calls me on the phone, I would recognize her voice too. If I am in doubt (which would probably land me in big trouble), I can always ask her about a secret piece of information, like a certain anniversary date that only the two of us would instantly know.
What is wrong with verifying humans manually?
It is just not very efficient or productive. We have been doing it manually since the stone age when we had to remember our own folks of the tribe, and prevent unauthorized intruders from entering the cave. We used techniques like common signs, speech, and passphrases. Time is finite. Our time could be better spent doing other important things — instead of highly repetitive tasks like verifying other humans to make sure they are who they claim to be, or painstakingly providing data and proof that you are who you say you are.
What are the most accurate ways to tell one human apart from others?
Human anatomy has many unique features that are specific to each individual. Argentina started using fingerprints for criminals identification in 1892. We have come very far from that discovery. Today, we have highly accurate and high throughput biometric authentication systems for various applications and industries. Artificial intelligence in hospitality has advanced so much in recent years that you can actually book a hotel room and complete the hotel check-in with a selfie. With robust methods to identify humans automatically, bad actors have also started using more and more sophisticated methods like deepfake videos to get past liveness checks of facial recognition systems. People are finding ways to counter such attacks too. A nonprofit scientific research institute known as SRI International (formerly Stanford Research Institute) has done a research project that can spot audio visual inconsistencies of a tampered video.
Imagine this. It took 90 seconds to process and a human face into a few landmark data values using a GRAFACON or RAND Tablet in 1964. It was faster and cheaper to hire a human to manually to check the person against her photo ID compared to the machine. Today the opposite is true. Three pretty important trends have happened in our lifetime. (1) Internet/cloud (2) deep learning AI, and (3) smartphones. With advancements in cloud technology accessible via high speed mobile networks and the prevalence of smartphones today, most humans finally have on their body an “always on, always connected” computing terminal. Anyone can now process complex biometric information on the go and accurately check the results against trusted sources, five decades after the technology to digitize human biometrics was first explored. This is why we have built our products to be cloud native, mobile first, and be powered by machine learning.
How can I use this in my day-to-day business?
If you are in the business of hotels, co-working spaces, or gym member clubs, we have combined hospitality and technology to create ready made hotel software with major system interfaces that you can use off the shelf. With this advanced artificial intelligence in hospitality, you can enjoy this next generation check-in experience. If you are in a different kind of business, we have cloud APIs with pre-trained machine learning models that you can plug-in to your existing app or website to perform a human identity verification or KYC (Know Your Customer) checks for a two-week free trial.
Why Narrow AI?
Of all the strides that have been achieved in AI, deep learning is one of the important breakthroughs that happened a few decades ago. Yet, we are nowhere close to creating a sentient general AI that can learn and do any task without prior-programming or sufficient training data. Self-driving cars – even with all that data, connectivity, and arrays of sensors for computer vision – still would not be very good at performing a slightly different task it was not designed for, such as helping to check you in at the hotel after it valet parks itself. It just is not part of its job scope, sorry.
One of the fastest ways to fruition for automation is to create a narrow, purpose-built AI that can do one task very well, and go through a lot of datasets to perfect that specific cognitive task over time. We gave GTRIIP’s AI for hotels more than just a digital identity, but a persona and a name. We call her Kelly. Kelly is meticulous, fast in getting things done, has a warm personality, and can definitely help to check you in at a hotel swiftly. We would know. To date, she has checked in hotel guests a total of 500,000 times across 10,000 hotel rooms for a joyful travel experience – and, she is getting better at it after each time too.
Kelly still does not know how to park a car even if her life depended on it. But we do not mind that at all!
This blog post was first published by GTRIIP Founder, Maxim Thaw Tint on LinkedIn.