Read all of John’s articles here.
It was Alan Turing who first discussed “Machine Intelligence” in his seminal 1950 paper for Mind “Computer Machinery and Intelligence.” Although we freely interchange the terms “machine”, “artificial” and “computer”, it is the “Intelligence” that is significant. What did he mean by it?
Turing suggested that if, having removed the obvious machine characteristics, a human could be fooled into thinking that a machine was another person, then it exhibited intelligence. One way of removing machine characteristics is to only allow it to interact with another human via a computer terminal.
So imagine you were having a Skype conversation with someone you had never met. If at the end of the conversation you were convinced you were conversing with a human, then the computer could be said to be behaving intelligently.
For Turing’s definition to be true, it did not matter what was at the other end of the terminal – it could be a machine made of wheels and cogs. If it fooled you, it was intelligent.
Unfortunately, we seem to have lost sight of that original definition and are more concerned with how the machine works than what it does. You may have come across processes called neural networks which attempt to model how the human brain works in solving problems. These work by learning – rather than being taught. So, for example, a traditional approach to pronouncing Italian words might be to store the rules – which by and large (and unlike English) are consistent. Using a neural network approach you would verbally input a large number of Italian words to a computer, then provide the spelling, and, eventually, the neural network would work out the rules.
The benefit of neural networks is that the programmers don’t need to understand the rules. Indeed, in some cases, nobody even knows the rules so there would be no way to code them (for example identifying cancerous cells at early stages). However the downside is that they can require massive amounts of computing resources, and much training. This tends to restrict them to services provided by the international IT monoliths, which is possibly one reason why so many people believe they can only benefit from intelligent services provided by the likes of Google, Apple, Microsoft and Amazon.
In reality, the most basic service can be intelligent and benefit an SME. At one time, we had to control the temperature in our homes by putting more wood on the fire. A thermostat must once have seemed intelligent (if not magic) because it would have been difficult to believe a human was not controlling the temperature.
So, where does that leave the SME? Well, provided we base our choice of intelligent IT on how it can replace an employee and not on the logic by which s/he operates, then the SME can benefit.
The problem with describing neural networks as intelligent is that while they may provide a massive amount of computing power, they might not actually be able to fool us.
Let’s take automated telephone assistants for example. For sure, a computer which nominally performs all the tasks of a telephone operator will be far cheaper than a human telephone operator. And it will work 24/7, rarely be off sick and take no holidays. But if every caller has to go through an average of five options and repeat their account number six times before eventually completing their transaction, eventually they will lose customers. This takes time. In the case of the banks, it is the ones that are moving away from automated telephony that seem to be attracting customers.
Sadly, there are not many banks – even though new challenger banks are still entering the market – and the freedom to change is somewhat restricted. But this is not the case for SMEs.
Before we take the moral high ground about replacing people with machines, we should bear in mind that it is not always that easy to find good staff. It can be hard to find a telephone operator with an engaging personality. The wrong employee can give a worse first impression of a company than a good machine.
SMEs need to consider automation from a range of perspectives – including being able to record, file, and quickly receive, all messages (voice and text) relating to their business. This is already possible today through sophisticated but inexpensive technology, but despite its many uses – for record keeping, legal and other queries including GDPR and much else – many businesses are simply not yet doing it. And there can be no doubt, this technology is “intelligent” – it does things that humans would do if they had time.
So to summarise, how should an SME decide if “artificial intelligence” can help their business
● Why do I need artificial intelligence. What would Turing think?
● Ignore the definitions of artificial intelligence. Does the process provide what an employee does? Does the process provide what an employee cannot?
● Do I have the option of employing someone? If I had the exact job spec and budget, could I actually find anyone to fill it?
● What exactly is the “artificial intelligence” promised? Keep away from broad terms like “voice technology” and “neural networks” and establish in practical terms what the service provides.
● Don’t adopt AI because you think your competitors have it.
● Before I spend any money, can I talk to a real person about my needs? Large providers are fond of using great marketing speak, but are rarely interested in providing any human sales support to SMEs.
In any event, there can be no doubt that as we better understand what humans do, we can make machines to do it better. The important thing for an SME is to understand how that can be leveraged without costing more than it saves. Indeed, this is the same process for every technology required to run a business.