Rise of AI: Will there be space for humans in an automated world?


The history of human beings has always been entangled with technologies. The ability to create and use technology is a distinctive characteristic of our species. Humans are longing for new discoveries and human society is constantly shaped by the advancement of technologies. Are we living in the time in which technology will take over? In this world, where the life of billions of people is so perfectly integrated with computers, phones, processors… is there still space for humans? Are machines going to steal our jobs?


Is artificial intelligence cuddly or scary?

Artificial intelligence is a term that most of the people with an access to media and/or social networks have heard nowadays. The reaction when hearing this term is ranging from deep excitement to serious concern. As naturally happens when the majority of people has superficial access to a concept that only a minority have deeper knowledge of, a lot of confusion is generated.

My job is in the context of Artificial Intelligence and yet I don’t like to use this term. When talking of artificial intelligence most people think of an oracle, that is going to learn whatever it is to be learnt and give us solutions to problems that we didn’t even know we had. This type of intelligence is still relegated to Philip Dick’s novels. I very much prefer to use the term Machine Learning to explain what I do. Machine Learning is a series of mathematical algorithms, that allow our computers to see details and structures at a speed impossible for a human mind. We have a plethora of data and without the automation that only a machine is capable of we don’t have much hope to make sense of them. By using these algorithms, we are not having an infallible tool, that is solid under any point of view, but we get precious insights that critically used can help in finding solutions.

Machine learning is not magic

Machine learning algorithms have been around for decades, but they met a growth never seen before in the past few years. Higher computational speed, bigger financial investments and a higher number of people involved in the field are among the reason of this fast growth. The algorithms come from years of research, that is still ongoing, but they are applied to a very complex scenario. The data that Data Scientists study are often very noisy, too noisy to hope to analyse them deterministically. For this reason, we cannot put blind trust in machine learning. Machine learning can highlight a structure only if this structure exists and is visible in the available data. On the other end if the structure exist machine learning will reveal that in a faster and more efficient way than a human being. Machine learning cannot magically solve all our problems but can help us in addressing some of them, that would have required a lot of resources otherwise.

The world is integrated

Integrated systems, composed of many different machines and software working together and communicating, are everywhere nowadays. Even our mobile phone that is able to talk with our laptop, often through the cloud, constitute an example of an integrated system. For some companies the integration is extreme, counting thousands of nodes. In these cases, having a tool that monitors complex integrations becomes of crucial business importance. These systems have an enormous amount of parameters that need to be analysed simultaneously. Therefore, machine learning is a good candidate to make monitoring more efficient.

Machines vs. humans

It might sound surreal, but the most common approach to monitor an integrated system is still setting static thresholds by hand for each single parameter in your system (each of thousands!), and alerts of anomalies are sent when this, sometimes arbitrarily chosen, values are exceed (see also the nice blog post by Eva de Jong on nomeREva). This is not only extremely time-consuming but can also be extremely inefficient. Since human beings don’t deal efficiently with huge amounts of data simultaneously, only very few thresholds can be set reliably based on previous data. Machine learning can help us in going through all the historical data and highlight patterns. It learns what are the baselines for the normal behaviour of the system, in such a way that an alert can be issued on a dynamical threshold, that is learnt overtime.

Even when a drastic change happens, with the help of machine learning we can get new thresholds in a faster and more accurate way than an operator changing all the thresholds by hand, trying to guess the best solution. Moreover, machine learning can guide us to the identification of the root cause, by learning correlations among events.

These algorithms are not going to substitute the human operator, human feedback will always be very important. The feedback of a human will always be vital and will improve the learning of the machine. For example, the operator could confirm that an anomaly is really an anomaly and not the consequence of a man-made decision (such us doubling the memory or temporary switch off of a system) and decide if the machine has to use this new data to learn a new baseline. In some cases, it might even be useful to set some auxiliary threshold by hand, for few selected parameters, but this would mean to set just few of them instead of thousands.

It is worth to mention that if in the data that we feed into the machine learning there is no structure and/or too much noise, the algorithms are not going to learn, exactly in the same way as a human being would never be able to see any pattern in these data. Even with all the limitations, machine learning is a tool that speed up the monitoring and sees details that are sometimes impossible to catch for the human eye. Machine learning works in collaboration with humans, not in substitution.

But are the machines going to take over?

Machine learning is powerful whenever we have to deal with complex systems in which a lot of correlation is present. But let’s go back to the question that is haunting the dreams of humans: Are the machines going to take over? The present aim for which machine learning is developed is augmenting the human intelligence, for this reason this type of technology is not going to take over. Machines need the cooperation with humans in order to be intelligent! There is a threat in all this anyway. It is not machines overperforming humans, but humans underperforming with respect to their possibility. The problem would be if we stop being critical, thinking that machines could think in our place.



The jobs of the future

I would like now to spend few words on a very complex problem on which is important to keep the debate open: How will the concept of job evolve consequently to the huge automation of tasks that is taking place? This is a matter that requires serious consideration and for which a solution has not yet been found. Something that really needs to be thought through, without phobias towards ‘’evil technologies’’, that have as only result to delay the debate. New technologies keep happening, because they are the result of human curiosity and because they improve our life. Nowadays they evolve at a high speed, but they still need humans to reach full potential. We should understand and not fear them. Only understanding them we can reshape our society and the concept of job in order to integrate them.

Schedule a live AIMS demo from an integration expert