For some time now, people have been talking about the coming technologies that will revolutionize society and leave no line of work unaffected. Yet, so far, disappointingly few applications have actually seen the light of day. And those who have might seem rather underwhelming in scope. Where did the future go?

It is not for lack of trying, as many AI companies with noble intentions find themselves stranded mid-development. There are just so many things to factor in when converting a bright idea to fully-formed products and services.

To the world, AI might seem like a superior technology, leapfrogging the sluggish, incremental ways of traditional innovation. But the truth is rather that, without the blueprints of predecessors’ learnings and a holistic approach to value creation, AI companies have to overcome all odds to reach wide societal implementation, as it inherently struggles with fundamental challenges.

For machine learning to work, you need data to feed its insatiable hunger — and of the right kind and quantity. Then enter the design phase, in which you must find the right way to productize the offering through some type of software or hardware. And even then, you still need to convince puzzled customers and reluctant users before you can roll out plans for outreach and on-boarding. Any step along the way might prove fatal.

Here’s a sobering case in point: As the future continues to deny us flying cars, we can at least pin our hopes on autonomous vehicles. Google has invested billions of dollars into developing such machines, so why are we not already emancipated in the former driving seats of stoic rail carts traveling in orderly rows along invisible vectors?

Partly because accidents during testing have grown into major obstacles, as the imagery of rogue killer machines eclipses the fact that statistically, self-driving cars are still much safer than their human-driven counterparts. The status quo wins, which is why it is so hard for AI companies to pivot out of the research phase and into reality.

At Corti, we realized that we were facing a comparable challenge. is overflowing with edge cases — too many for technology to flawlessly predict. But flawless is what’s required, so we set out to design a process, which would allow the AI to flourish alongside its users.

Though the Corti AI has no billion-dollar wheels to move around in society, it is deployed in emergency medical dispatch and thus involved in the space where life and death is leveraged on an hourly basis. In Copenhagen’s emergency department — already one of the best at detecting out-of-hospital cardiac arrests — Corti’s automatic speech recognition and live detection algorithms have been proven to increase detection rates by 28%.

This suggests that implementation should be a no-brainer, but introducing new technology in the healthcare sector is never straightforward. We were still faced with a reality riddled by edge cases, ambiguous data, and difficulty discerning what was actually a value add. To face these challenges, Corti’s team designed a three-fold process to cross the chasm between research and reality, consisting of:

1. Capturing data

2. Interpreting data

3. Enriching data

We started with Copenhagen Emergency Services, where technical limitations in their existing infrastructure restricted our ability to capture live data for processing. So Corti needed to design around this obstacle, eventually inventing a new piece of edge computing called the Orb, whose design encases a compact supercomputer within its small and durable exterior. The department needed only to connect the Orbs to their workstations, and they were up and running, detecting cardiac arrests on live calls. The Orb’s deliberately ambient exterior achieved a very smooth landing with the dispatchers, who quickly took a liking to their new AI companions.

Having obtained access to real-world events through the Orb, interpreting the data was the next big challenge. That’s why Corti’s product team developed a new triaging software with Seattle Fire Department, one of the world’s best emergency services. The finished product’s light, almost ethereal interface organizes and presents expert information in a way that saves crucial time by skipping unnecessary protocol steps that are not relevant to the conversation at hand.

Every step the dispatcher takes in diagnosing patients is annotated directly on top of the live audio stream, providing a human interpretation for the AI to learn from. Triage was built in collaboration with senior paramedics, who were offered the opportunity to alleviate decade-long frustrations by describing ideal workflows for the Corti product team.

To complete the cycle, Orb and Triage had to be complemented by a third Corti product in order to enrich the data sets. This gave rise to Review, a quality assurance platform in which dispatchers and managers are motivated to go back and analyze past calls to further their own learnings as well as their department’s learnings relative to others.

While the Review software was built to drive quality assurance for emergency departments, it simultaneously provided crucial data enrichment for Corti’s models, allowing the human learnings to continuously feed the AI.

The wheel formed by this trinity creates a self-improving process for capturing the necessary real-world data (the Orb), auto-collecting, annotating, and interpreting the messy information to become machine learning-ready (Triage), and finally letting users clean and enrich the data themselves (Review) to enable the AI to start learning beyond the imagination of any research team.

Bringing bleeding-edge research to societal implementation demands a product team capable of negotiating the transition between yesterday and tomorrow, while concurrently providing instant and tangible value.

A product can hardly be considered optimal if nobody cares to use it. And as with many other disruptive efforts, AI’s intelligence must be questioned if it proves tone-deaf to the behavioral patterns of end users as well as the requirements for live data to flow back from sea to source.

This product strategy has taken Corti out of the laboratory and into reality, signifying a new stage in Corti’s journey, in which our researchers now can take machine learning to the next level and realize AI’s true potential. All this while their experiments will run along a product , enriching dispatchers and data alike to change healthcare forever.

Source link—-819cc2aaeee0—4


Please enter your comment!
Please enter your name here