I work for GYANT[0], and we're building a chatbot for doing medical triage, patient follow-up, and answering health questions. We're young (launching this Thursday actually), but it's already been pretty successful in a number of countries (like Brazil) with a few hundred thousand users with minimal marketing. My girlfriend was able to use it without issues to go from "my throat hurts" to visiting a doctor in 24 hours because she likely has tonsillitis. The interest and response from the medical, the pharmaceutical, and the insurance communities have been very positive and they're approaching us left and right to bring this out to more users.
Chatbots were probably overhyped and oversold to people. Not every platform needs to explode overnight and produce SV unicorns. I think there is immense value in chatbots as a platform, but expectations should be tempered. I've seen little to no overt marketing to regular users that they can ask a bot rather than looking for an app or website. Most users don't really want to download and learn another app and create another account for every little thing. Additionally, many budget phones in emerging markets have limited data, limited storage space, and minimal specs for downloading and running apps for every little thing.
That said, this article is pretty terrible. It almost seems like they chose the worst possible examples to fit a given narrative.
How has your company's interaction with regulators been so far? From your description and the animation on the top page of the Gyant website, it looks like the chatbot is claiming to diagnose medical conditions, which in most cases would put it under FDA oversight.
We honestly haven't faced much scrutiny from regulators yet (although I expect that may change once we become more popular). We fall pretty squarely into the existing rules on symptom checkers which have become pretty prolific on the web.
Not really. We have a legal agreement, and then the users state their chief complaint. We record their symptoms as we ask them questions and produce a list of possible conditions along with their disposition (e.g., Go see a doctor in 24 hours, visit an emergency room, home care, etc.) based partially on the Schmitt-Thompson protocol. Schmitt-Thompson is the current battle-tested standard for medical triage.
Why did you decide to build a chatbot, rather than a website/app? It seems like it should always be possible to provide a better experience in an app.
So far I've heard only three compelling reasons: to support users without smart phones, to support voice interfaces or to be accessible through Slack etc.
We support voice interfaces and we can support other platforms such as Slack, Telegram (we used to support it actually), and more. Like I said earlier, most users simply Google their chief complaint (e.g., "my throat hurts") which lead them to sites of varying quality such as homeopathic sites and WebMD. These sites do more harm than good. Users are also reluctant to download apps when they just want answers such as health answers.
Bots allow us to reach people all over the world no matter whether they are on their desktop, tablet, smartphone, or budget phone on every platform (iOS, Android, etc.). When I look at my girlfriend's phone, she has 95+ updates waiting and she's not looking to download more apps when a Google search would suffice.
We've discussed building an app in the future to enable even more features, but it's not a high priority right now. We're competing against Googling what's wrong with you, and we have to be just as easy to get started with. The priority for us is answering people's health questions.
You don't really need "AI" for this. For this "triage" use case a good expert system/decision tree is more than enough. Think of it like akinator for a given domain.
Hand-building the tree to be efficient is actually the hardest part.
Edit: yes, this is probably "AI" to some people, but AI only exists as long as we don't understand it. This kind of tech has been well-understood since the 70s.
There's a fair bit of AI, NLP, etc. The system starts out with decision trees built by/evaluated by medical professionals and learns what conversational elements are required to reach a diagnosis. This isn't just a trivial set of hand-built trees.
Chatbots were probably overhyped and oversold to people. Not every platform needs to explode overnight and produce SV unicorns. I think there is immense value in chatbots as a platform, but expectations should be tempered. I've seen little to no overt marketing to regular users that they can ask a bot rather than looking for an app or website. Most users don't really want to download and learn another app and create another account for every little thing. Additionally, many budget phones in emerging markets have limited data, limited storage space, and minimal specs for downloading and running apps for every little thing.
That said, this article is pretty terrible. It almost seems like they chose the worst possible examples to fit a given narrative.
0: http://gyant.com/english/