Neuroscience to be the key enabler of better artificial intelligence

Neuroscience to be the key enabler of better artificial intelligence

Dr Shyam Diwakar*

Alan Turing's 1950s paper (Computing Machinery and Intelligence, Mind, Volume LIX, Issue 236, 1 October 1950, Pages 433–460) and his question "can machines think?" has been the basis of the then artificial intelligence. In March 2019, a paper in frontiers journal points on the consensus on a brain interface to a cloud through the collage of nano-manufacturing methods, computations, nanotechnology and artificial intelligence leading to internet of thoughts. Reconnecting the dots from the 50's to last week and extrapolating them to the future, it is not uncommon for scientists to turn to brain circuits and function to unravel modern and futuristic neurosciences. That's exactly what many groups of academicians and researchers from Google, Amazon, Facebook, IBM have started to do.

However, is India ready for it? The answer is indeed, yes and now. India is a land of IT giants, some powerful mathematicians and applied statisticians and more importantly, educationists. With its 907 Universities and several thousands of colleges, India can boast of a strong young talent pool. There are some shortcomings for the country to hit global star status using AI as potential. From 2005-2015, one can see trends among top companies moving to computing and IT related such as Apple, Facebook, Amazon and Google more than just oil and natural gas or automobiles. One may need to invest early to ride the wave.

India's neuroscience funding and resources are not comparable in number to many of its well established counterparts. Even today, countably few neuroscientists are working on producing relevant models or data. There are several academic authors generating research publications (people like me) but the question is - will that be sufficient to generate powerful findings, methods and algorithms. Computational neuroscience is a field of major thrust connecting the dots from experiments to hypothesis testing and from data to behaviour.

At Amrita Vishwa Vidyapeetham, a realistic cerebellum model is being developed and is being tested by our computational neuroscience laboratory. The cerebellum, a well known part of the brain that was known for over 2000 years, was recently incriminated as involved in major pathologies like Parkinson's, Alzheimer's movement disorders, autism, depression and more. Colleagues at National Center for Biological Sciences, modeller Prof. Upinder Bhalla have been using data to predict new memory processes. At IIT Madras, Prof. Srinivas Chakravarthy models movement initiation, planning in a brain structure called basal ganglia. The basal ganglia circuit is known to control voluntary motor function and other roles. At IISER Pune, efforts are ongoing to understand computations in the hippocampus, a memory and emotion center in the brain. There are some others at IITs and other institutions.

Modelling in AI has many more players trying algorithms on data, building new methods and more. So here in India today, informatics for AI is more popular than neuro-inspired AI, both of which can have joint or independent roles in transforming India.

Modern neuroscience and neural processes and patterns will lead to better robotic controllers and transformation devices; neural circuits showcase how sparsely connected structures process multimodal information better than many existing devices; challenges in local and long-term plasticity address how devices can fine-tune to get optimal results. Yet, to relate neuroscience data and models to AI is another looped process.

A crucial reason to pursue such models is to reverse-engineer brain function. In that context, we may need a close collaboration with physiologists, data scientists, behavioural scientists, statisticians, algorithm developers and AI colleagues to use physiology, fMRI (functional magnetic resonance imaging usually done before surgeries or tests in hospitals or research centers) and neuro-imaging data and more crucially, improve current methods of computing and processing of computations. This leads to a new dimension where biologists, psychologists, physiologists, computer scientists and engineers may need to level out on data from the same dimension and this may be where Indian thinking is really good at.

To access cumulative human knowledge, there is a need to explore neurosciences from the perspective of an engineer and reconnect with the doctors and physiology researchers who relate brain function and dysfunction. With new explorations tying neuroscience to AI, some researchers may choose to answer how autism is perceived, or how Parkinson's affects motor control changes or how dementia can be reversed or whether the sleep problems could be rectified. Using the same or similar data, some will choose to build new technologies, methods, and algorithms. The challenges here are data exchange among entities like devices, people and data processing of large amounts of data.

With ideas like cortex in the cloud (Front. Neurosci., 29 March 2019 |, downloading information could be like in "the matrix" or help us connect to a "global brain" - large AI based systems or superbrains that can help make sense of massive data. Neural nanobots may handle 6x10^16 (ten to the power of 16) bits per second wireless and 10^18 (ten to the power 18) through nano optic cables, changing the way we look into brain monitoring and data extraction.

With the rapid focus on AI, and within the next five years, there will be a convergence of how we build neuron like models, circuit like structures, information processing algorithms from labs across the globe and India should stay in lead. Urge in exploring computational principles from neuroscience and models from data will restructure a better future for several thousands in our IT job industry as well.

Funding agencies could start massive data and algorithm challenges and that could keep us in the top. Professors and industry experts should retrospect not just in publications or case-based products but on the necessity of their interventions. As we see with the recent photography of a black hole, students will dare to go beyond and media could highlight more such causes connecting our people, scientific traditions and institutions. This may probably redefine our country's roles within and beyond the scope of AI and neuroscience.

*Dr Shyam Diwakar is a senior faculty member of Amrita School of Biotechnology, Kerala and an international expert on Computational Neuroscience

Share it
To Top