For a Human-Centered AI

From research to business, Mauro Cettolo’s story

April 17, 2018

After 26 years in the Foundation, the HLT-MT researcher began a period of secondment in the company. "A decision not taken lightly, but it seemed like a good opportunity to try something else"

In April, Mauro Cettolo started a secondment period in Pervoice, company leader in the field of speech recognition technologies stemmed in 2007 precisely from FBK technologies and skills.

A two-year period made possible by Provincial Law No. 6 effective December 13, 1999, art. 5, art. 19 and art. 19 bis (Aid for applied research, for the dissemination of scientific research and for the integration of researchers into companies).

Mauro, leaving the Foundation after so many years must be a bit like leaving home.
Oh yes, it feels a little bit like that. It’s a new opportunity that I was presented almost by accident and that I decided to take due to a series of random situations and meetings that occurred last summer. The conditions arose in a very short time for me to say “why not?”.

So?
So, after 26 years of work at the Foundation. I am going to live an experience in a company. Obviously it was not a decision I took lightly.I was always impressed by a sentence that Nadio Delai, former president FBK, once said: a person should change jobs every ten years.
It’s a take that I share, and that I tried to make my own in my professional career, first changing my research interest from speech recognition to automatic translation after thirteen years and now, after another thirteen years, going to a company.

How did this decision come about?
The reasons that led me to consider the possibility of changing and accepting PerVoice’s offer are many. On the one hand, after so many years, there is certainly the need to try something new, to get involved. And then, as mentioned, some coincidences both of an exogenous nature and of an endogenous nature. The first concerns the change of paradigm that has occurred in the field of language processing and machine translation, with the establishment of deep neural networks based on algorithms that are more performing and winning than those with a statistical approach. The latter, the endogenous ones, have to do with some factors that have changed the structuring of our research group.

Your secondment is made possible by Law 6 and is also in line with the provisions of Article 55 of the Provincial Collective Labor Contract for Foundations staff.
The possibility that legislation offers me is important, it is an ideal instrument especially for those like me who are no longer so young. Before now I have never worked in a company and I think it can be an excellent opportunity – I don’t know whether there would have been other – to try and experience.

Two years.
That’s right. I think it’s a reasonable length, not too short not to not fully get into the life of the company, not too long. Although two years of “detachment” from life and work in the Foundation are many. Then as said, we will see what happens in this time frame.

What do you expect from this experience?
Certainly a clear change in priorities and in the way of working. In addition to the actual research and its “reporting” through publications, a significant part of the work in the Foundation was “community service”: reviewing papers, organizing conferences and seminars, participating, supporting students, collaborating with journals. See, all this activities will gradually diminish until they cease altogether. I cannot say now if I’ll miss them.
From a practical point of view, in the past 10 years there have been significant innovations, as I said. Web giants have developed and made available software packages for machine translation. The research work involves improving some specific aspects of this software. In the company, improving can translate into cutting costs, making this software more performing, adapting them to the specific needs of clients.

For example, not all companies are inclined to use the services offered by these giants because they are not willing to sell their data and those of their customers. And it is here that companies such as PerVoice can be competitive on the market, guaranteeing adequate performance and at the same time safeguarding data privacy.

And what when you come back?
I believe I will have a much broader vision of the area in which I work, precisely because I will have acquired a perspective that is different and complementary with respect to the one experienced so far.
In the company the model, the best solution are developed, based on the hardware and technology conditions. In research, the contrary is true: the optimal model is developed and later on we think about how to implement the infrastructure that makes it work better. We often lack the awareness that some aspects that are considered irrelevant for publication can instead be considered of value for companies. More generally I think that those who transit between business and research can bring new life and help to better examine these aspects.


The author/s