AI is making its way into our lives more every day and the acceleration of its development should contribute to a 14% increase in the World GDP ($ 15,700 billion …) by 2030. But behind these astronomical gains are also emerging pressing ethical issues, because we must ensure that these technologies actually benefit society as a whole. If only a few catastrophic futurologists venture to evoke the imminent danger of a takeover by artificial intelligence, the reality is that automation is increasing at exponential speed and, with it, the risks of drift…
For European Commissioner Margrethe Vestager, Executive Vice-President for A Europe Fit for the Digital Age, ” When it comes to artificial intelligence, trust is not a luxury but an absolute necessity. Preventing the risks of AI requires making strategic technological choices now, choices that ultimately guarantee that humans collaborate optimally with machines while maintaining control. This condition is essential in all critical areas impacting the lives of individuals and organizations: defense and cybersecurity, but also transport, finance, health or education.
This is why humans and businesses are increasingly uncomfortable with “black box” type AI algorithms that give no indication of how decisions were made. We need an AI to dialogue with in order to follow these decisions with confidence. This assumes that the recommendations of the machine are accompanied by explanations. The challenge for this next-generation AI is therefore to be a “white box” (and not a “black box”), capable of enabling a relationship of trust with humans.
Data-driven AIs, such as deep learning (“data driven”), will not be able to meet these challenges. Neural networks are very powerful, but they are black boxes with indecipherable decisions, even for their own designers. The future will belong to those who know how to combine the strengths of the different approaches to artificial intelligence: AI must become hybrid, that is to say capable of combining the two main currents of artificial intelligence, intelligence “ symbolic” which is based on the consideration of logical rules, and “digital” intelligence based on the processing of quantities of data by neural networks. Finally, it must be energy efficient: While DeepMind, during the famous game which opposed it to the world go champion, used a computing power of 1202 CPUs and 176 GPUs – the new generation AI which has just victoriously faced 8 bridge world champions has consumed 200,000 times less energy to achieve its ends.
This hybrid AI also has geopolitical advantages that should not be overlooked: if the GAFAM explain to us at length in tweets that outside the deep learning there is no salvation, it is because these mastodons have gigantic databases. While the data volume available in the world will increase tenfold between 2020 and 2025, they want to convince us that it is useless to compete with them because they have already won the Big Data battle. Perhaps. But they are still far from having won the war of artificial intelligence.
The prospects opened up by hybrid AI represent a fabulous opportunity for Europe, and more particularly for France. Our country has high quality researchers belonging to the different AI communities and, according to NatureFrance ranks 5th in the world for research in this field.
Large French companies and the French State must play their role as catalysts to promote the emergence of global players in hybrid AI. They must trust start-ups working on these technologies and entrust them with large-scale projects that guarantee them long-term funding and early positioning on the international scene.
Our country must capitalize on all of its talents to embark on this path of transparent, sovereign and low-energy AI. This is how we can contribute to the emergence of an AI that really takes advantage of intelligence.
We would love to say thanks to the writer of this short article for this awesome material
How to keep control of the machine? – AI news
Explore our social media profiles along with other pages related to themhttps://www.ai-magazine.com/related-pages/