Human-Robot, Avatar, Virtual Assistant, Chatbot: Open platform as the best business model
We are publishing a series of articles on Emotion AI in human-computer conversational interfaces. This is the second article in the series.
Last time we talked about the key role of multimodality and the types of interface solutions based on different modalities in Humanoid Companions for Customer Services.
Today we’ll discuss why it is so important to think about open platform and its basic functionality for effective development and delivery of human-computer interface solutions such as Human Robot, Avatar, Virtual Assistant, Chatbot.
These solutions may seem very interesting from a sci-fi perspective and sometimes can be created “just for fun”, but still, all companies will always be oriented towards business KPIs and sales growth goals. To meet the goal and achieve KPI targets there are two important tasks among other things:
• First — proper customization, the task of the ultimate match between functionality and the market’s demand (let’s be frank, when for example you look at some young robotics startups, you inevitably ask yourself: does the product manager really understand market/niche demands and what is the plan of addressing them in the product features?)
• Second — geographic coverage, delivery network
Of course, customization and geographic coverage are not all the tasks which should be solved.
Why do we highlight these two?
- AI-based solutions are sensitive to the data they use. Successfully entering the market depends on customizing a solution for a particular niche. Customization in its turn is always effected directly by the specifics of the original dataset. You cannot train the machine on Asian movies and then use it in US retail business-cases, or visa versa. A great example here is Norman — the world’s first psychopath AI from MIT.
“Norman is born from the fact that the data that is used to teach a machine learning algorithm can significantly influence its behavior. <…> Norman has been specifically trained as a psychopath. In Norman’s case, the data-set included extended exposure to the darkest corners of Reddit, with violent image captions and thus it explains the psychopathic tendencies of the AI”.
Obviously, no client would enjoy working with such «issues of non-customized product»…
2. This data-sensitivity is geographically-related because data is geographically-related. It is impossible to make a solution once and use it for every language, or use the audiovisual content from any country the same way without an understanding of possible local specifics. Basically, what is needed is a localized solution for each market you plan to enter, adapted to the specifics in the area. Moreover, the demands of diversity for multinational countries should be considered. The way people interact is geographically specific, therefore the solutions should be as well.
3. Solution functions and features should be customized properly. In the task of making a thorough list of all potential functions that the client might need, the list would be enormous! Let’s consider just one example — robots for customer service: they should speak (and sometimes we expect them to sing cute songs), count the price and collect customer satisfaction data, in some cases indicate direction to the exit or gate, or provide information about flight, give an opportunity to pay for delivery and finally… hug and give compliments — everyone loves a robot with warm hugs! Sure, it’s a joke, yet this variety of functions demonstrates how hard it is to understand what features should be integrated into the solution — what functions and for whom this humanoid companion should carry out. Look at the CES demonstrations — you might be mistakenly looking for a credit card slot on a ‘hugging’ robot.
4. Last but absolutely not least. Any interface solution should be able to process natural data. People communicate in natural language, which is (including speech, sign and body languages) largely geographically dependent. (We are preparing another article on Natural Data Processing and NLP).
It is hard to overestimate the vital role of geographical and industrial customization: the use of customized datasets, geographical specifics and functionality that will be relevant for different industries.
Which business model takes into account all of these factors? Look at the picture below.
First of all, the business model should include a geographically distributed network of frontend (local) partners — this network provides geographical and industrial customization. What is more, local partners assure high speed and efficient supply chain — selling the final solution to specific customers. The reason is simple: local partners (good local partners) know local clients and have direct access to them. They know their problems and what is required to solve them, and often they have the necessary client data to include in the “correct” dataset. The development of such a partner network in strategic regions and a competent partner capacity planning based on sales targets and sales strategy is equally important for the business success of a company that develops human-computer interface solutions.
Further, it becomes obvious that a company with such a business model (with the focus on “geographical and industrial customization” through a regionally distributed network of frontend partners), has the necessity to create partner-ready platform (hardware and/or software) that will become the basis for the development of a final customized solution.
And one of the most important factors here is how functional and convenient this platform will be for the partners who use it and how correct the set of opportunities in it will be.
Finally, one more very important unit in this chain is a collaboration with a range of highly specialized technological partners. Let us remind you that AI-based human-computer interface solutions are extremely cross-disciplinary and complicated. They require:
- Careful research in several scientific fields
- Highly qualified machine learners
- Business-people who understand final business tasks
Part of these can be managed by the company itself, but another part of them can only be solved by external technological partners that have expertise in this specific area, experience of deep experiments in collaboration with global scientific partners, and the results of scientists and AI-specialists’ joint work. These partners have a deep understanding of the topic and know how best to steer the client from the creation of Norman the psychopathic AI, towards a product that more closely resembles C3PO or R2D2, or something else, depending on what exactly the customer wants.
Neurodata Lab is such a technological partner in the Emotion AI industry that provides advanced tech in the sphere of Affective Computing for interface solutions (chatbots, virtual assistants, avatars, robots).
In the next articles, we’ll discuss how Emotion AI helps solve the NLP challenge and what is required to work with emotions ‘in the wild’.
If you want to receive our publications directly or you’re interested in the topic and want to get access to the demo portal with our Emotion AI products, contact the author directly:
Partner Development Director
Neurodata Lab LLC