The beneficial relationship between AI and Neuroscience
Zacharias Siatris ????
Mastering Business Growth with Artificial Intelligence and Cloud Technologies
Humans have been gaining inspiration from nature for many thousands of years with the purpose of solving complex human problems. Nowadays, however, there is a great interest and a lot of research towards humans and the mechanisms of the brain and cognition. This scientific field is called “Neuroscience” and it’s one of the main sciences that focuses on cognitive functions and therefore, highlights the way nervous system develops based on specific innate structures. Another upcoming scientific field based on Neuroscience is “Artificial Intelligence”. Neuroscience has played a key role in the history of A.I. and it has been a great inspiration for building human-like A.I. For developing and accomplishing Artificial General Intelligence, the need for the fields of both Neuroscience and A.I. to come together is now more urgent than ever before. To overcome the arising challenges, the attention should be focused towards finding a common and direct path for incorporating these two fields.
For many centuries, since the ancient Greek times, intelligence has been a topic of interest by philosophers in an abstract and even moral way. However, a more scientific approach has been thriving for the last 50 to 60 years to deeply understand the mechanisms and functions of the human brain. The result of these studies and the findings regarding human cognition, made artificial intelligence (AI) to emerge; with the aim of not only understanding the way the human brain works but also mimicking the functionality and architecture of it which led to innovations such as artificial intelligent machines. Neuroscience, which is the field that studies the nervous system, works as a source for guidance for AI. The main approach for developing AI machines is to mimic the cortical circuitry in a neuron-like structure where discrete elements are connected by adjustable weights (synapses).
Furthermore, because the understanding of the nervous system is still limited, neuroscientists cannot answer which aspects of the human brain are more important to focus on and develop relevant mechanisms with the goal of reaching AGI (Artificial General Intelligence). However, extensive research has shown that the nervous system is equipped with preexisting innate structures that allow humans to accomplish meaningful and complex behavioral tasks. Such structures could be inherited in the AI field to develop more human-like solutions. There are two approaches for achieving this goal. The first one is to mimic these brain structures and mechanisms and try to implement them in AI models. The second one is to develop methods and algorithms that have no prior training / knowledge to achieve computational learning. By combining these approaches and taking advantage of the neuroscience advances in cognitive and functional brain imaging AI can leap at the chance of rising to the next level.
Recent studies highlight the importance and the benefits of incorporating and mimicking neural systems in the development of neural inspired AI. The first and the major benefit, is the creation and evolution of deep nets. Deep nets are the basis for every problem that can be solved by implementing computer vision, speech recognition, text translation and/or other AI algorithms. Apart from deep nets, reinforcement learning(RL) has been introduced. This brain-like computation algorithm, where reward signals are used and calculated to make decisions, has been used effectively in robotics application. By implementing a deep network architecture along with RL, milestones and stunning results can be achieved especially in the section of (video) gaming. Another insightful benefit, is the development of function-level mechanisms and target tracking algorithms(implemented on robotic platforms) by investigating the visual system of non-mammalian animals through which, work has identified a form of predictive gain modulation where responses are selectively enhanced. The development of artificial brain-inspired navigation algorithms is the next benefit of leveraging neuroscience into creating AI models. These navigation patterns are developed by observing elements of spatial coding within the hippocampal formation. Even though these observations were abstract enough, they were of great help and formed the guidelines towards robot navigation systems, localization and mapping algorithms. Another piece of information regarding this specific field is that it still continues to progress and thus, will continue to affect the direction of the development of such mechanisms and algorithms.
At this point, there is a need to demonstrate some few more advances that have been risen from the participation of neuroscience into technologies and led to the development of tools that help scientists and generally research to move a step forward in understanding the complicated structures of the brain. Such advancements are the electron microscopy (EM) imaging, which is combined with novel reconstruction algorithms, large-scale calcium imaging and tools that are able to simultaneously identify, record and manipulate multiple populations from different cell-types. Through this way, scientists can have access to a range of neural temporal dynamics that were previously inaccessible.
On the other hand, there are great challenges that need to be taken on, to incorporate neuroscience into AI approaches and thus, reach a level of human-like AI models. One of the major challenges is the path and the directions each science follows. While neuroscience focuses on finding therapies for diseases and disorders, AI research tries to understand the way an AI model reaches an outcome at an algorithmic level. Because of these directions or even biases, we may end up in a disconnect between these two fields. To overcome this obstacle, both neuroscience and AI need to get along and aligned in terms of research and the paths needed to be taken. More specifically, neuroscience needs to play a more functional role when it comes to share insights and incorporate with AI to pull along in the same axes. In general terms, the challenges between neuroscience and AI are not only in terms of the directions that the research in each field follows but also on the cultural aspect – meaning multiple levels of interaction. In order to describe this difference, the long article (Crossing the Cleft), presents three tiers by which the impact of each research effort of the two fields can be described and therefore, depicted. These are the “form”, “mechanism” and “function” levels. AI mainly focuses on the mechanism level along with the function one, which describes the way a solution is produced by a system/model at an algorithmic level (“how does it work” and “what does it do”), whereas neuroscience’s interest is towards the form level (“what is it”) and the mechanism one (“how does it work”) which for example focuses on circuit dysfunction and thus, follows the path to understand the mechanisms and the biological neural architectures.
领英推荐
Moreover, there are also some more “tangible” challenges in the broader image of cortical circuitry and AI such as hardware and computational cost. Even though there are platforms which have great computing performance (e.g. GPUs, CPUs) and are one of the main factors for the success of deep nets on the other hand scientists need to act proactively in terms of computing technologies and future solutions. In recent years, there has been a demanding growth of such technologies and big technology companies have developed neuromorphic hardware solutions to satisfy the requirements for the development of powerful AI tools and mechanisms. Last but not least, regarding neuromorphic advances, there is a trade-off for this programmability which is the level difficulty of implementing such solutions.
To enrich the challenges part with few more pieces of information and also understand the big picture, incorporating AI and neuroscience can lead to some ethical challenges too. Given the global burden of neurological and psychiatric disorders, accelerating AI innovation for the benefit of people in need is a medical, societal and ethical priority. While holding great promise, however, the technological novelty and versatility of AI models raises important methodological and ethical challenges when applied to human neuroscience. Although any of these challenges are inherent in any biomedical application of AI, at least five of them acquire particular ethical salience when applied to the human brain and clinical neuroscience more generally. These are scientific and clinical validity, accountability, the risk of neuro discrimination, agency, and neuro privacy. Addressing these challenges is critical to maximize the benefits of AI for humans, while avoiding failure in implementing AI systems in research (Marcello Ienca & Karolina Ignatiadis 2020).
In conclusion, future research is essential to validate the kinds of deductions that can be drawn from both articles. The key to successfully incorporate AI and neuroscience is to combine both empirical and computational approaches to tackle the challenges and therefore, carve a common path to reach the desired state – AGI. The first step to achieve it, should be to acknowledge the differing priorities and address the cultural differences. Decoding the insights of neuroscience on a function level and sharing AI outcomes to different levels of neuroscience would achieve the desired incorporation.
REFERENCES:
Office Manager Apartment Management
2 年It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with primary consciousness will probably have to come first. The thing I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing. I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order. My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461
Data Analyst at Fairstone Bank
2 年Very useful! Zacharias Siatris ????
Technical Specialist @REInvest Greece
2 年Zacharias Siatris ???? Very interesting one! Thanks for sharing! ??????