How Good Is AI Today?
I am sure that quite many people today – and I am by no mean different – hear and probably think that Artificial Intelligence (AI) is something fundamentally difficult and sophisticated and lies beyond their understanding and all measures of normality. However, even today AI is not as developed as one may think. In this article, I would be happy to introduce you to up-to-date information on AI limitations and its main struggles.
?AI, at least as most of us may think of it, is certain algorithms that are set up to perform certain functions to meet some objectives. This implies, that those algorithms first need to be trained on large sets of data so that they grow from primitive not yet work-ready codes to somewhat sophisticated algorithms capable of solving articulated pre-established objects. Now, AI is only as good as the data it is trained on. This might effectively mean that it is too premature to speculate about AI as a self-sufficient and self-managing system that collects data on everything to exercise control.
?Some experts familiar with the subject say that now – and most probably some considerable time into the future – AI and Machine Learning (ML) will remain impossible to imagine without some form of human contribution (e.g., human insights and human-established goals, directives, etc). This sounds very appealing for people who are afraid of that AI stuff, the rise of machines, etc.?
?Chris Bergh, CEO of?DataKitchen, points on a fundamental issue regarding current AI settings and the complete integration of AI into the current infrastructure of business and economy. The root of many current AI and ML shortcomings may lie in the fact that today the whole system is constructed in a way that AI is using databases to train its algorithms. Bergh voices an opinion that AI should not be the consumer of databases; AI itself should be the administrator of databases so that the entire system works in unison. He stresses the importance of AI being able to properly manage the data to avoid even greater struggles and to retain what AI systems have achieved so far. For me, the most important insight by Bergh is that there probably is a quite pronounced knowledge gap between those people doing data management, Machine Learning, and AI which renders a negative impact on the development of the entire AI thing.
?Erik Brown, a senior director in the technology practice of?West Monroe Partners, says that human energy should be spent wiser in areas of AI and ML than it is currently spent. He is sure that human involvement should be solely strategic and not operational as it is currently happening. For human presence in AI to become strategic, the data environment should become very autonomous – i.e., databases should somehow be managed without a profound human involvement. Strategic involvement may mean contextualizing AI findings, interpreting them, figuring more interesting ways of AI usage, etc.
领英推荐
?A senior architect in West Monroe's technology practice, Jeremy Wortz, mentions that completely autonomous data environments will take many years to develop and that there are very few data environments apart from Amazon and Google that are ready for this new way of doing AI and ML things. He also thinks that this is very early to push yourself into thinking that AI can be used to solve some big and critically important questions as of today. Instead, AI systems should be polished and perfected when it comes to dealing with relatively small and narrow-focused tasks such as assistance during surgery, etc. Focusing on and achieving success in such narrow AI applications can bring many fruits in the future. At the same time, people from the industry should pay great attention to creating self-managing database systems and environments.
?Important voices from the industry claim that in the future analytical skills and business knowledge – in the context of corporate usage of AI – will be almost as important as technical and computer skills when it comes to AI since the technical core becomes increasingly user-friendly while the importance of result interpretation is crucial.
?
?