More Than Just AI: How To Successfully Develop AI Products
Successful AI products and services are the result of a combined effort of complete technological teams working to solve actual human problems.
The artificial intelligence (AI) talent race is in full swing as companies race to hire AI specialists. But if you venture into the creation of an AI product or service, it soon becomes evident that hiring AI engineers is only one piece of the puzzle. The centerpiece is, unequivocally, AI technology itself. Directing your efforts and allocating your resources primarily toward the development and enhancement of your AI capabilities should be your paramount focus. But for an AI product to live up to expectations (in the lab and in real life) requires an entire engineering team of various disciplines that consider the technological and the human dimension.
The Human Element in Action
Imagine you access an AI-based application that lets you enter different criteria and search for vacation offers. Perhaps you’re looking for a “romantic holiday in the Caribbean at the end of November, at the best price possible” with no specific destination. The tool takes into account the objective criteria such as region, time period, and price, but it also might suggest typical “party” destinations like Cancun or Fort Lauderdale rather than more romantic options. When an AI tool fails to fulfill your needs like this, it’s easy to walk away from it and take on the task yourself.
When you’re shopping for clothes, for example, instead of searching for a dress with concrete characteristics, you can type: “allow me to select a little black dress like the one Audrey Hepburn had in “Breakfast at Tiffany’s””. This type of search resembles the dialog you’d have with a sales assistant. It can act as a counselor.
So what’s the key to building an AI tool that not only works and generates results but also gives people what they really want? The answer is pretty straightforward — you must first have a very good idea of what users need. In the above cases, how they choose their vacations and how they pick their clothes.
Bringing a group of engineers into an AI project is integral to its success, but it solves only one part of the equation. AI specialists are perfectly capable of developing a service that gives results based on strictly defined search criteria, but this doesn’t necessarily solve the actual human problems. If you want to have a successful AI product, you must add a human element to the equation.
The User Experience
Making your AI product relevant to real life involves also paying a great deal of attention to the user experience (UX).
You may recall the Tay chatbot scandal from 2016. Famously dubbed “Microsoft’s Racist AI Chatbot”, the project was shut down less than a day after its launch. It was designed to be part of a research in conversational understanding and engage Twitter users in casual talks. In a matter of hours, Tay began replicating users’ most offensive behavior. The bot was trained using Twitter data, but it wasn’t given an understanding of inappropriate language. In other words, the issue didn’t stem from the algorithms per se but rather necessitated enhancements in the training process to elevate the user experience. In this context, the UX primarily revolves around the content of the responses themselves.
Some of the most remarkable AI products witnessed recently, such as ChatGPT, may appear deceptively simple in their interface design. In fact, ChatGPT’s UX design extends far beyond its minimalist text prompt. The user experience has been meticulously crafted through substantial and costly endeavors aimed at finely training these algorithms to enhance response quality. It accounts for data curation, fine-tuning, guidelines and reviewer feedback, bias mitigation techniques, user feedback, handling of controversial and sensitive topics, abusive content filtering, and public input and transparency.
Complete Engineering Teams
There’s a common misconception that attributes the creation of brilliant AI products and services solely to a group of AI prodigies. But the most successful tools are the result of the joint effort of an interdisciplinary team. A complete engineering team sees AI specialists collaborating with business strategists, business analysts, designers, software developers, infrastructure specialists, security experts, and quality engineers.
As you can see, AI engineers are not a majority in the AI team, and this has a very specific business intention. Every team member brings their solid understanding of AI’s application on board and ensures their expertise is seamlessly integrated into the development of the product. At the same time, AI specialists maintain ongoing communication with their colleagues from the other disciplines.
A MentorMate client that effectively implemented AI in a mission-critical domain shared its experience with us recently: “We faced significant challenges in adopting AI. Some specialists in this field focus solely on solving problems in controlled lab settings and lack the expertise to seamlessly transition their models into production.” Production means, among other things, deploying these algorithms so that they work in real time, securely, and at scale. If they don’t, your product will either crash, in case of shortfalls in the engineering, or not do anything useful for users, in case of a poor UX approach.
If you aim to build a commercially successful AI product, it’s imperative to foster collaboration among diverse disciplines right from the project’s inception.
Final Thoughts
Over the last two decades, significant investments have flowed into applications and services harnessing the power of the Internet and mobile technologies. One key takeaway from this period is the realization that digital technology, on its own, represents just a component in crafting impactful products and services that have the power to positively transform our world.
Now, these investments are being redirected towards AI.
Cutting-edge AI technology holds incredible potential. To unlock and wield this potential for the greater good, a comprehensive strategy is essential. It involves not just AI experts but multidisciplinary teams of technologists who can safely deploy these powerful tools securely, in real-time, and at scale. Moreover, realizing its full capabilities requires expertise spanning far beyond technology to help effectively connect it to the humans who use it.
Original post here.
Authored by Sebastian Ortiz-Chamorro.
