At the end of 2019, Forecasting AI has spelled out a lot of articles on AI to offer you the most likely predictions for Artificial Intelligence in 2020. And here is a non-exhaustive list of what could await us in the field of Artificial Intelligence!
1. The Rise of AI as a Service
As soon as companies try to use the benefits of AI on a large scale, its democratization will increase considerably. However, few people will have direct access to this technology, which requires non-standard computing power. This is why AI as a service should take on a new dimension in 2020. AIaaS, an acronym comparable to SaaS for Software as a Service, will allow companies to use AI as a tool in the cloud without needing all the hardware in-house.
IBM Watson, Microsoft Azure and Google Cloud are just a few examples of how the technology is being used in service-oriented environments. Servers support most of the processing power, allowing clients to access a solution remotely. Quantum computing will continue to drive the growth of the remote access model. With a third party taking care of everything, enterprises have access to a cost-effective, no-risk, no-obligation AI solution.
2. More and more human beings
This is not anecdotal, nor is it new in 2019: AI scares a lot of people. However, it is also very useful and invites itself as soon as it can in our daily life to help us and assist us in our tasks. Therefore, before possibly replacing your colleagues at work or yourself, Artificial Intelligence should still provide a good number of solutions to our problems until we, perhaps, become more people in 2020.
So, before replacing you, AI should therefore save you time in your work or also allow you to be more efficient or productive. A French startup called Q°emotion, for example, is already using AI as a tool for emotional and semantic analysis in order to identify the blocking factors of certain companies in relation to their customers. A customer service department that uses AI for these purposes could undoubtedly respond to a greater number of people or requests, which are always more complex. The customer experience could then be even faster, more accurate and more qualitative thanks to an AI capable of analyzing a greater number of variables than a human being can before providing conclusions. It will only be up to the employee to decide on the strategy to follow based on these conclusions. Banks, manufacturers, transport companies will thus be able to improve and optimize their relationship with their customers.
3. An even more ethical AI
The trend at the end of 2019 leaves little room for doubt. AI in 2020 will be ethical or it will not be. Finally, it will be, it's obvious, but to avoid breaking the trust that some have placed in it or to try to win new followers, it will have to show a white paw.
Thus, the European Commission has already published seven guidelines to be followed for the development of ethical AI. Companies such as Google, IBM or Microsoft have even embarked on the creation of ethics committees in charge of monitoring the various AI projects in order, for example, to avoid bias; and we are not even talking here about the Cambridge Analytica scandal or about those Google employees who demonstrated after their company had decided to develop projects in favour of the American army...
4. AI with less data is possible in 2020
It is often said and repeated on Forecasting AI: no data, no AI. Indeed, data, in great quantity and quality please, are nowadays essential for the development of deep learning projects. And yet, in 2020, these Artificial Intelligence may be less data intensive. Make no mistake, however, they will still need as much data to train and develop their "skills".
Nevertheless, this data could be derived from data synthesis. A concrete example? You need portrait photos of people to work on facial recognition and train your model. Why try to appropriate photos that could be subject to image rights when generative adversarial networks - hyper-realistic image generating algorithms - can allow you to have millions of different faces to train your models?
It is therefore especially in data collection, an often long and time-consuming task, that Artificial Intelligence could well evolve very quickly.
5. Vocal First!
It's an expression that comes from our imaginations but which finally sums up quite well what is expected of vocal assistants in the near future. According to Gartner, 30% of research will be vocal by 2020 and 7 billion devices, yes 7 billion you don't dream, will be equipped with voice assistants. The technologies around Siri, Google Home or Amazon Echo are constantly evolving for searches that are increasingly accurate, fast and personalized according to the user.
Finally, Google has not been left out in the cold either, thanks to BERT. BERT, acronym for Bidirectional Encoder Representations from Transformers, is an extraordinary advance in the Mountain View firm's algorithm for better understanding natural language. And consequently to offer increasingly relevant results to those who use the search engine as a tool to answer their questions.
In conclusion, AI is still in its infancy, despite some great advances in 2019. And we could have talked about quantum computing expected at the turn of the century, GPUs still indispensable in computing power, new applications in marketing or even autonomous vehicles and convolutional neural networks. Technologies or applications, some of which are still in the innovation phase in the Hype Cycle curve of emerging technologies, again according to Gartner. You will have understood it, hurray 2020!