Subscribe
Sign in
The Human Chatbot — Not all chatbots are created equal
This is an experiment from the VideoAsk team @typeform (www.videoask.com) to see how people would interact with video instead of a text chatbot.
Is this the future of chatbots? You tell us...
34
Replies
Best
@rrhoover Sure...would you mind applying for the pilot through the bot just so we can get all the info in the same place? Would be great to work together.
@rrhoover tell us a bit of background on how you might use this tech too...thanks
@david_okuniev1 you bet! I just applied.
I need to think more about this but I can see this being a useful tool for sharing some of the things we're building at Product Hunt/YourStack and gathering video responses/first impressions.
@rrhoover hey Ryan, just invited you to the pilot, did you receive my message?
I've been following the development of VideoAsk closely for some time. This is such a cool development of the tool itself. It felt like a bit like a mix of talking to a telephone robot and a real human, but in a good way.
The natural language responses worked great for me. I like seeing the voice input feedback as captions when I'm talking, indicates that the chatbot is listening and understanding what I'm saying. This is some really cool interactions, and I want to use this in at least a couple of my projects.
From my perspective there is huge value in starting customer/user/leads interactions like this, it's super personal, engaging and more fun than a standard chatbot. I honestly would not know why you would do a text chatbot anymore at this point?
The only con for me as a responder was that I felt a bit un-prepared for the last part of the conversation when it wanted me to record a response, given my current WFH situation, I'm not too happy recording a video of myself, would have loved if I could respond in text form.
@linusekenstam Great feedback! Good point on the live recording bit...maybe we could cushion that somehow, but switching to text although obvious, might defeat the object in this case ;)
@linusekenstam @david_okuniev1 can the response respond with an audio message, instead of recording an actual video of themselves?
@linusekenstam @david_okuniev1 @anna_0x I think this would be great, give the option to answer via audio message.
This is freaking me out, but in a good way :) As a non English speaker, the AI understood me perfectly. I'm wondering how far you can push this technology, but I truly believe it will be an incredible experience to the customer interacting with the Chatbot in the future.
P.S. Great UX with the last Chatbot ;)
@bujold19 thanks Gabriel, its early days but we’ve already got so much out of building this. Looking forward to see how people can use it.
I tried signing up for early access yesterday (I know I’m very late to the party) but it got stuck at the uploading stage... 30 minutes later I just closed the tab.
Hey Peeps,
At Typeform we're constantly looking to push the boundaries to help brands build better relationships with their customers.
A few months ago we officially launched www.videoask.com, a new product that helps businesses get 10x more personal with their customers.
We recently started playing around with the idea of building a video chatbot and today we want to share progress. It's an experiment which we will hopefully fold it into the VideoAsk product in due course.
If you want to use this beforehand you can apply for the limited pilot.
For now...go ahead have fun with it (you can even curse at it).
David Okuniev (co-founder @typeform)
Top Hunter
@david_okuniev1 this is a clever idea, and fairly well implemented, but the impedance in the walkie-talk-like back and forth makes it feel SO cumbersome. It demonstrates how much subtlety is conveyed in face-to-face synchronous human interactions so, for example, my interlocutor would know to move ahead to the next option or another topic if I grimace or make a face that indicates displeasure. With this system, I have to wait until the video clip finishes playing, which increases my annoyance with each successive turn.
@chrismessina That's great feedback!
But perhaps you are coming in too strong with the expectation that this will feel like a face to face conversation, with constant interruptions and "chit/chat".
It's an experiment so it's all about learning. Depending on the nature of the conversation it will work better/worse. For example imagine "speaking" to a doctor using this tech; you might be more inclined to patiently listen to the advice (I think the same thing would happen in real life).
I guess it all depends on who is really driving the conversation and how much authority the "bot" has... Anyway, glad this is opening a good conversation and one of the reasons why we built this.
Thanks for checking it out.
Top Hunter
@david_okuniev1 totally — but I'm also comparing this "enactment" techniques against text-based chatbots to see if the human element really enhances things, or just gets in the way.
If you were to use Maslo's emotional detection technology, for example, then maybe you could detect when a user is frustrated/annoyed/wants to move on and could respond. That would make the use of video more effective.
Otherwise, this feels like a combination of Walmart and Eko's interactive retail initiative (example) with a speech-to-text interface — a kind of bounded choose-your-own-adventure where the form might be interfering with the content.
I will say that after a few turns, the video crashed/got stuck "thinking", so that may have colored my experience.
@chrismessina Agree. And I think it would not be a huge leap to provide some way of just triggering the video to go 1.5/2x faster ("Get on with it/get to the point). Don't even think it necessarily needs to be detected, could be just a standard feature of the interaction.
This is super cool @david_okuniev1 –– just applied for the pilot. My team is exploring the development of a chatbot for queer youth to support their mental health. We just finished up a chatbot for young cancer survivors that was clinically shown to decrease anxiety; hoping this might be a way for us to take that work even further for a new group of young people!
Great concept there!
Could you tell us a little more how is it build?
@benj_malartic thanks Benjamin. We will probably release more info / write-ups soon :)
I love it but it is just the beginning. The path is made by walking -- With Deepfaking and artifical technology becoming better and better every day, in a couple of years, videoask will be the perfect tool to bots answer everything u want with a human physiology and gestures. And if u are the first to implement this like videoask is doing, you will be top of mind (always be remembered) This means, less effort to marketing your tool in the future
Loved using this, felt so 'Typeform-y' , very natural addition to the platform !
So frustrating! It can't parse my english accent - 'tell me about the chatbot' is understood as 'tell me about the chapel' 3 times - gave up
@safeasmilk Maybe it thinks you should go to church 😁(sorry could not resist) ;)
@safeasmilk But in all seriousness... NLP is never going to be 100%.
But what you can do is teach the bot to listen out for keywords that may sounds the same to bring detection accuracy up.
Atm this is an experiment, so we have not gone that deep yet.
Product Hunt