20 March 2017
Artificial intelligence (AI) is a broad term that incorporates everything from image recognition software to robotics. The maturity level of each of these technologies strongly varies. Nevertheless, the number of innovations and breakthroughs that have brought the power and efficiency of AI into various fields including medicine, shopping, finance, news, and advertising is only growing. All of the companies undertaking such initiatives had to undergo a number of changes. Introducing any technological change into an organization presents a different set of challenges.
Here are the most common problems of artificial intelligence implementation:
1. Different development approach.
Most development in a traditional systems environment follows the usual phases such as plan, analyze, design, build, test, and deploy. The AI environment is quite different. Most of the time, development is about identifying data sources and then gathering content, cleansing it and curating it. Such an approach requires different skills and mindsets, as well as different methodologies. In addition, AI-powered intellectual systems have to be trained in a particular domain.
In case we compare conventional (regular) and AI programming, the differences will look the following way:
Generally speaking, with AI we are not developing a system but training, giving feedback and supervising an AI-powered solution.
2. A system is only as good as the data it learns from.
Everyone already knows that AI needs data to learn about things. AI and machine learning rely on enormous amounts of high-quality data from which to observe trends and behavior patterns, as well as being able to quickly adapt to improve the accuracy of the conclusions derived from the analysis of that data. Basically, first you get the data then you get the AI. Such systems don’t just require more information than humans to understand concepts or recognize features, they require hundreds of thousands times more. Another important thing is the quality of data used to train predictive models. The data sets need to be extremely representative and balanced, otherwise, the system will eventually adopt bias that those data sets contain.
3. No clear view on how insight is generated.
Another of the problems of artificial intelligence is hiding in its experimental nature. It is difficult to say how much of an improvement it may bring to a project. Therefore it is almost impossible to predict ROI. This makes it really hard to get everyone to understand the whole concept. One thing that is necessary to optimize the result is a skilled team that can write or adapt publicly available algorithms, select the right algorithm for the desired result and combine algorithms as needed to optimize the result.
There are benefits and dark sides to every disruptive technology, and AI is no exception to this rule. The important thing for every company is to identify the challenges that lay before them and acknowledge the responsibility to make sure that they can take full advantage of the benefits while minimizing the tradeoffs that problems of artificial intelligence may impose.
Already have a project in mind? Let’s talk.