State of AI: artificial intelligence needs more staff and better data

Artificial intelligence (AI) is promising, and it is increasingly ready for real world businesses.But there is a shortage of talents, a lack of diversity in this field and concerns about the processing of data which feeds increasingly sophisticated algorithms.

These are some of the observations of Nathan Benaich and Ian Hogarth, two leading investors in the field of artificial intelligence, who published their fourth annual report "State of AI", very dense, which reviews thedevelopments in this area during the past year.

Although the report focuses on academic AI and specific advances in medicine and in other fields, important developments have been raised for those who seek to take advantage of AI and machine learning to progress in creationsmart businesses."Insufficient efforts to align artificial intelligence on the part of key organizations that advance the general domain of AI, as well as concerns concerning the sets of data used to train AI models and partialityModel evaluation benchmarks, raise important questions about the best way to follow the progress of AI systems whose capacities are progressing quickly, ”say Nathan Benaich and Ian Hogarth.

Here are some notable AI developments during the past year:

Employment in AI

IA expertise becomes an increasing concern, as well as a field of opportunities."Computer researchers, software developers, mathematicians, statisticians and data scientists have seen an evolution of their employment much higher than that of the general active population," said Nathan Benaich and Ian Hogarth.“IT and engineering were the undergraduate diplomas that experienced the strongest growth between 2015 and 2018, representing 10.2 % of all four -year diplomas awarded in 2018.Their number increased by 34 % and 25 %, respectively, during this period, while the number of other diplomas awarded increased by 4.5 % on average.»»

Globally, Brazil and India are leading the growth of employment in AI, hiring more than three times more talents in AI today than in 2017, equaling or exceeding growthhiring in Canada and the United States, they add.

Data on diversity within American organizations differ radically between technical and non -technical teams, also affirm the authors. Il est intéressant de noter qu'au niveau mondial, « près de 30 % des articles de recherche scientifique de l'Inde comprennent des autrices, contre une moyenne de 15 % aux Etats-Unis et au Royaume-Uni, et bien plus de 4 % en Chine»», ajoutent-ils.

Projector on the data

Capital venture investors highlight the concerns related to the management of Big Data in the AI world."A careful selection of data saves time and money by mitigating the difficulties of the Big Data.Working with massive data sets is heavy and expensive.A careful selection of the examples attenuates the drawbacks of the big data by concentrating the resources on the most precious examples, but the conventional methods often become impractical on a scale.Recent approaches attack these calculation costs, allowing the selection of data on modern data sets.»»

Nathan Benaich and Ian Hogarth underline the need to improve data quality, in particular in real -time situations, such as detection or forecasting of events endangering the lives of people. Ils citent par exemple la menace des « cascades de données»», définies par les chercheurs de Google comme des « événements composés provoquant des effets négatifs en aval des problèmes de données»». Ces chercheurs avertissent « que les pratiques actuelles sous-estiment la qualité des données et entraînent des cascades de données»», pointant du doigt des facteurs tels que « le manque de reconnaissance du travail sur les données dans l'IA, le manque de formation adéquate, la difficulté d'accès aux données spécialisées pour la région/population étudiée»». Il est donc nécessaire de « développer des métriques pour évaluer la qualité des données, de meilleures incitations à l'excellence des données, une meilleure éducation aux données, de meilleures pratiques pour la détection précoce des cascades de données et un meilleur accès aux données»».

Les investisseurs prédisent également que l'année à venir pourrait voir le lancement d'une société de recherche axée sur l'intelligence artificielle générale, « formée avec un soutien important et une feuille de route axée sur un secteur vertical»», qui pourrait potentiellement impliquer des outils de développement ou une application des sciences de la vie.

Source: ZDNET.com