Nexedi software stack covers all the needs of the enterprise: infrastructure, applications and productivity. It is licensed as Free Software with no US content restrictions and can thus be deployed anywhere in the world independently of political instability. Please refer to the Nexedi Free Software stack for more information on individual and complimentary solutions.

Political instability in the United States and the possible use of Artificial Intelligence (AI) for warfare have both raised concerns about the need for regulation of Big Data and AI. Internet pioneers such as Tariq Krim, who had always been strong proponents of latest digital innovations, have publicly expressed in French mainstream media their concern with the lack of regulation of AI and the damages of unregulated AI on democracy and social welfare. In particular, AI seems to increase social inequalities instead of reducing them.

Big Data and Artificial Intelligence are changing medicine from a science that can be taught to anyone at university using text books into cryptic machine learning models based on secret data owned by a few corporations. Use of AI in healthcare could even violate the Hippocratic Oath which commands to share knowledge between master and student.

Professor Feng Xiang of Tsinghua University recently published in the Washington Post an article that explains the contradiction between Artificial Intelligence and capitalism as we know it today. He suggests that planned economy based on AI could achieve better results than market. His views are not unrelated with those of Peter Thiel who considers that monopoly - a private equivalent of state planning - is the optimal form of industrial organisation in the digital age. 

AI has revived the 19th century debate about whether the invisible hand of market or planning can best maximise economic welfare .

Market approach consists mainly of investment in startup companies that create AI models for healthcare and later selling those companies to larger ones.

Cardiologs company was for example founded in France to provide a system able to recognise patterns in the cardiac signal in the same intuitive manner as human experts do. One of its key executives, Jia Li, is actually a Chinese engineer who studied at Ecole Polytechnique. He heads R&D and trained a neural network on over 500,000 ECGs. Sooner or later, Cardiologs company will be acquired by a larger company. Just like Cardiologs, startup companies focusing on AI for healthcare are popping up in Europe, in Americas and even more in China.

They will all eventually be acquired by what could become in the future a global corporation dominating the healthcare AI business by capitalising on as many datasets as possible.

Planning approach consists of trying to match the offer of healthcare data and the demand of healthcare data through so-called Big Data Challenges organised by public institutions.

The IMPAC big data challenge on Autism has for example been organised by INRIA Saclay to analyse data of patients suffering for autism and invent AI models. Some challenges may be restricted to public research institutions only whether other challenges are opened to private companies too.

So, which of market or planning optimises welfare best? 

The first theorem of welfare economics, which was demonstrated by Maurice Alais in the case of industrial goods, states that market price equilibrium optimises economic welfare. This theorem justifies for example that medicines can be produced for profit by private firms which compete each other and are free to set market price, because market will eventually optimise welfare and thus save lives.

However, this theorem does not apply to so-called non rival goods, such as information, software or AI models. Unlike medicine, an AI model can be copied at no cost. The only price that optimises welfare for AI models is zero, because it costs nothing to provide an AI model to every human being: it is just software.

Yet, most AI startups in the field healthcare intend to commercialise AI models either as a pay per use service or through proprietary licensing schemes, both of which depart from zero price. Market is therefore not able to optimise welfare. Instead, it leads to the formation of monopolies and eventually high prices.

Planning has other issues. It is not able to predict the future of demand and is thus weak at handling innovation. If we consider Big Data Challenges, not everyone is informed or invited to participate. Many opportunities of innovation are lost. 

Software industry teaches us a way to optimise welfare that is neither based on market nor on planning: Free Software.

One of the most popular software used for AI is called scikit-learn. It was created originally in France at INRIA and Telecom ParisTech. It is used by Google, Meituan, Criteo, Airbus, etc. This software does not belong to any company. It does not belong to any government. It is free: its price is zero. Anyone can contribute to it.

Many scientific research articles on AI for healthcare are based on scikit-learn:

• Machine learning for classification and quantification of monoclonal antibody preparations for cancer therapy (L., Marini, C., Gramfort, A., Nguyen, D., Cherti, M., Tfaili, S., Tfayli, A., Baillet-Guffroy, A., Prognon, P., Chaminade, P., Caudron, E. and Kégl, B.);
• Large scale screening of neural signatures of consciousness in patients in a vegetative or minimally conscious state (Sitt, J., King, J., El Karoui, I., Rohaut, B., Faugeras, F., Gramfort, A., Cohen, L., Sigman, M., Dehaene, S. and Naccache, L.); 
• Shared Endo-phenotypes of Default Mode Dysfunction in Attention Deficit/Hyperactivity Disorder and Autism Spectrum Disorder (Julius Kernbach, Theodore Satterthwaite, Danielle Bassett, Jonathan Smallwood, Daniel Margulies, Sarah Krall, Philip Shaw, Gaël Varoquaux, Bertrand Thirion, Kerstin Konrad, Danilo Bzdok);
• etc.

All those publications demonstrate the power of Free Software to capitalise and promote international innovation in AI for healthcare.

Free Software has played a key role from the end of the 90s to regulate software economy. At that time, Microsoft was becoming a monopoly. Its software was expensive, insecure and very buggy. Competition was disappearing. Apple nearly bankrupted.

Then came Linux, which powers nowadays most smartphones and Internet servers. Linux is a Free Software that anyone can copy and extend. The existence of Linux has played a key role for 20 years to temper Microsoft's monopolistic behaviours and ensure a perfect balance between innovation, competition, universal acess and economic welfare.

United States' NIH recently made a move that could change the way we think about AI for healthcare, just like Linux changed the way we thing about software. NIH Clinical Center provides one of the largest publicly available chest x-ray datasets to scientific community. The total size of the dataset is about 50 GB. 

It is a major news because it demonstrates that a different path is possible for innovation in healthcare. Instead of granting to a few selected startups or academic research centres some exclusive rights to access private patient data, NIH opens to the world anonymised patient data.

This way, everyone on earth can access this data without restrictions. International collaboration and innovation becomes fluid and is not hindered by any form of barrier to access data. Independent developers can research data and create innovative models without having to go through any type of procedure.

Free AI Software becomes a foreseeable possibility thanks to the existence of Free Big Data as it has been implemented by NIH. It represents a third way that can provide more innovation than economic planning while regulating the excesses of market.

A dataset consists of multiple individual files

Each individual file provides detailed metadata that is essential for data scientists and AI developers.