Has the importance and impact of artificial intelligence (AI) been overhyped?
What is meant by an AI winter?
The benefits and reach of developing technologies are often oversold. The same can be said for AI, which has had many cycles of boom and bust. Huge promises are made, and we are told AI is going to impact every part of our lives. This leads to a surge in investments and many new companies with AI in their name. Almost inevitably, when it is realised that promises can’t be delivered and things take longer than expected, the investment and enthusiasm reduce significantly, leading to an AI winter. There have been three major AI winters previously, 1966, 1974 and 1987.
What is happening now?
If you believed everything in the media about AI, you would be forgiven for thinking that machines are about to take over. However, this is not the case. As with most new technology, AI has been the victim of being overhyped, often by those who do not have the full picture, or by those looking to grow their own business. This has led to disillusionment and we are seeing funding being reduced in some areas, and more carefully applied in others.
AI: What does this mean for the average business?
Contrary to popular belief, most companies do not use AI, and only a small number have AI projects underway. Here at FLOvate, we periodically host workshops on the benefits of machine learning. These events are attended by technicians or managers from a wide range of companies, from SMEs to large corporations. Based on data from the people we have spoken to, we estimate that only 5% of UK companies are actively pursuing AI projects.
Focusing attention and investment
At FLOvate, we believe that for AI to be worth the investment, it must improve one of three key areas. For example:
Will it lower operational costs?
Will it improve process outcomes?
Will it improve customer experience?
Machine learning can improve all three of these, which is why it is one area of AI that continues to attract attention and investment. However, it can only make improvements if the environment is right. So, it is not the solution in all cases, as it can be costly and relies of large amounts of data.
One example of AI working well is your phone letting you know expected journey times for your commute, just before it expects you to be leaving work. It is very clever and is based on identifying any commonly made journey. However, it was probably not cheap to implement, is based on the large amount of existing data your phone holds and is not something that can be easily applied for other uses.
AI has the potential to make big differences to businesses, but it can be expensive to do well. This will limit its adoption to particular type of business and specific problems. The reach, impact and timescale have been overhyped, but we will continue to see an increase in AI based technologies, such as:
Machine learning uses statistical routines to discover design logic based on multi-dimensional grouping of outcomes in a parameter matrix. What this means in simple terms is that a system will monitor repeated patterns and will eventually be able to learn these steps and take next steps itself.
AI routines can also be developed based on explicit rules, that is, a system can make certain decisions based on a set of configured parameters. This works particularly well in compliance environments and is much easier and cheaper to implement than true machine learning. FLOvate has implemented a script/GUI based decision engine that can work based on explicit or implied rules. For 2019, the FLOvate team is working on a very exciting visual decision designer that will further allow end users to include decision making in their systems.
Natural language interpretation
Progress is also being made in natural language interpretation, and we seem to be on the verge of being able to ask questions of a system and receive text answers back through a messaging service (SMS, WhatsApp etc.). The answers would have to be one of many preconfigured replies, but the system would have to decide which answer was appropriate. It could even enter into a digital conversation to determine exactly what kind of solution was required.
Low-code, in which FLOvate specialises, is a type of AI. It gathers information based on the configuration of existing elements and from this, decides on the type of system that is required. Subsequent triggers activate specific routines that automatically generate the appropriate code to deliver a fully working system. The benefit: the system is built on configuration rather than manual development.
Is winter coming?
AI cannot deliver everything that has been promised. It will not, as data scientists predicted decades ago, replace the jobs of humans. AI can do a lot of what we do, far more quickly but it does it with brute force and by rote, rather than intelligently. As a result, it will take some seriously big developments for it to reach anywhere near the level of hype.
So what’s different this time?
Undoubtedly, AI today is different from AI in its previous forms. Low code, machine learning and natural language interpretation: these are all examples of AI in its latest incarnation. These won’t mitigate the gap between the hype and the investment to date, but something is different this time.
Simply, AI is now consumer focused. Whether chatting with a virtual assistant, being advised on a movie or the best route home: we all get to experience the product of AI, not just the data scientists. It is this key difference that could set AI onto its next cycle and finally get closer to delivering everything that it promised decades ago.
Maybe we are looking at an AI autumn rather than a winter?
Find out more about FLOvate and LEAP at flovate.com