GDPR & Neuromorphic Computing
GDPR 2 Years Later,
Data protection is one of the main points identified by the European Commission to ensure its digital sovereignty. Thus 2 years ago, it implemented the General Data Protection Regulation (GDPR), designed to “harmonise” data privacy laws across all of its members countries as well as providing greater protection and rights to individuals, which came as a landmark in terms of data protection regulation.
The 99 articles long text details the rights individuals have over their private data (data that can identifies a subject) and states new rules about data collecting, processing and storing for businesses, that if not applied can lead to important fines. Has it really been the game changer expected by the European Commission?
2 Years later
The European commission firmly believes the regulation has been a success. All European Union states, except Slovenia, and the United Kingdom -which signed up to GDPR pre-Brexit- have adopted it or adapted it into national data protection laws. European citizens are more than ever aware of their rights -there are 30% more data related complaints coming from individuals each year. Almost €400 million fine were given, with a record of €50 million given by the French CNIL in January 2020.
However, there are still many efforts to make. First EU states must work on harmonizing their regulations. Indeed, implementation of GDPR across member states is far from being consistent, which creates fragmentation that impacts cross-border business, particularly when it comes to new technological and cybersecurity developments. Secondly, while most of big businesses are now compliant, SMEs struggle to understand and implement the GDPR — in January 2020 more than half of small businesses weren’t compliant.
Overall, it did have the impact wanted by the European Commission: it has changed people’s minds over privacy and has established the European Union as the vanguard of data protection, as an example for countries to follow…
Neuromorphic Computing, The Next Generation of AI?
Current computers mostly use the “von Neumann” architecture, in which data is transferred between a central processing unit and memory chips in linear sequences of calculations. An instruction fetch and a data operation cannot occur at the same time because they share a common bus. This is referred as the von Neumann bottleneck and often limits the performance of the system.
While this architecture works well to do calculations on numbers and with accurate algorithms, when it comes to processing sensorial information such as sounds or images and make sense out of them, it lags far behind our brain and its billions of neurons.
Even if the incredible development of machine learning, and especially deep learning, this past decade has brought computers a bit closer to apprehending and understanding the environment, a small feat like learning to distinguish an apple from a banana still requires a massive amount of data, training and energy while it would take one look for a human.
One of the options that seems promising to allow computers to process massive amount of information with a minimal energy consumption and extend AI into areas that correspond to human cognition is neuromorphic computing. The concept was developed by Carver Mead, in the late 80s, and it describes the use of very-large-scale integration systems containing electronic analog circuits to mimic neuro-biological architectures present in the nervous system, i.e brain-like chips and algorithms. It introduces a level of parallelism (in gathering and processing information) that doesn’t exist in today’s hardware. A speech recognition algorithm working on a neuromorphic chip would be smarter and far more energy-efficient than current ones running on emulated neural networks.
Research and development efforts are beginning to produce tangible results with chip such as Loihi, developed by Intel Labs, which contains 128 neuromorphic cores, three Lakemont CPU cores and off-chip communications network. In March 2020 they combined 728 Loihi chips in the Pohoiki Springs system for a total of over 100 million neurons (the brain of a small mammal).
Neuromorphic chips won’t replace today’s CPUs and GPUs as they will remain much more efficient in “traditional computing”; they are more likely to be embedded next to them as separate cores
