Before getting into discussions involving Artificial Intelligence (AI), I think it’s (always) best to clarify what we refer to as AI, for everyone has a somewhat different understanding. I will try to do so in the following paragraphs.
In the sense that we will be using, AI refers to the intelligence of machine and robots, from a computer Science standpoint. And from that same standpoint, AI is classified in two ways – weak artificial intelligence (weak AI, or narrow AI), that is focused on a narrow task or specific problem, and strong AI or artificial general intelligence, which outlines a machine or software with the ability to apply intelligence to any problem, and to learn to solve general problems. It’s important to understand that all systems that claim to use AI nowadays are operating a weak AI focused on narrowly defined specific problems, and changing the problem in any way, or it’s input data, will, in most cases, render the solution useless or non-performant. As a result, if a specific problem can be solved acceptably well with AI, like, for example, Siri performing tasks of a virtual assistant via voice interface in English, that does not mean that changing to a language with a different structure, like Chinese, would perform as well. Not to mention that even if voice commands are perceived correctly and consistently, the technology is nowhere close to actual ‘understanding’ or interpreting natural language on a general level.
In order to achieve Artificial General Intelligence, or to solve hard specific problems like text understanding or interpretation, a lot of research and innovation is still required, with the timespan of achieving this, real experts in this field agree, can range from 10 to 100 years, even though it’s plausible it will happen sometime in the next 20.
AI, independent of the techniques used, be it machine learning or something else, is an intelligent adaptive pattern matching solution based on statistics and probabilities. Meaning all solutions solve a problem of finding specific patterns in a given set of information. Depending on the problem and the data set some solutions are better than others.
With that in mind, let’s have a look at what AI is useful and how it can be used in improving day to day businesses.
AI can be applied successfully in any field where there is enough structured or semi-structured data available and a problem can be formulated in a way where finding patterns can provide a solution. For example, in voice recognition, the same word is pronounced in a similar manner by most people, even if it may have a different pitch, tonality, volume. In photography or object recognition there are sets of pixels that determine features that determine patterns of a specific object or digital creation. In games certain sets of actions lead to a better performance in the end, and given enough game data these actions can be determined, connected and improved – see AlphaGo or OpenAI.
In any business, AI can help improve performance, but it’s always a matter of effort and reward. When it comes to effort, AI requires consistent data, sometimes manually annotated by people – and this can make a potential AI solution very expensive to implement or scale efficiently. Nevertheless, in today’s world, it is definitely worth trying to figure out if an AI solution could give your business a competitive advantage in the market and what options you have to implement it.
Businesses, both small and large, should be looking at AI as a tool to create value for their customers, their employees, the bottom-line or to improve a critical internal process. Often time it involves all four of them. An easy way to assess this is to assess the problems and potential of improvement – here’s how to start:
What’s the problem?
– Identify problem areas in your business (e.g. high rates of return, errors on the production line, ineffective marketing campaigns, misplaced products, stock management issues…)
– Identify possibilities for automation (e.g. correcting errors in financial transactions, using a chatbot for support …)
– Identify frustrations from your consumers (e.g. long lines, technical or customer support…)
– Identify employee irritations, dissatisfactions (e.g. inconsistent schedules, doing repetitive tasks…)
– Determine which one or two would make the biggest impact on your business
– Consult with experts in AI to determine which solutions are viable. Implement the plan created either in-house if you have an R&D team, through companies who specialise in AI or through 3rd party SAAS services.
So how to go about it? If your core business is not AI, unless you already have a decent sized R&D department, hiring an AI data scientist that knows what he/she’s doing may be quite expensive and it would not make sense from a business standpoint. Relying on a good developer but with no data science, background or real expertise to build the AI solution your company needs is in most cases doomed to failure. Understanding what techniques and architecture is needed or might work, as well as knowing how to prepare the data is crucial for a successful application. So that leaves you with either hiring a third-party specialised team – an agency that has actual data scientists – or using an existing service that solves your problem. Our advice is to contract someone to figure out what the best solution would be for your business, not to build it, but to consult. This way the specialist gets paid to make the best recommendation and would not have any incentive to upsell you on custom development that may not be needed.
AI is changing the way we do business. It might not (yet) be in the way we see in sci-fi flicks, but its creative use is helping to create more loyal customers, happier employees, optimal processes and better bottom-lines. Give AI a try – figure out your business improvements and get in touch with a specialist.