Simulating the human mind
Brand new technologies are leading to results of emulating the human brain, with all the positive spin-offs.
New hardware chips and software concepts enable machines to more closely mimic the mystery behind human reasoning.
The present and future of AI
Laboratory brains
Intel Labs and other development centers are successfully applying Neuromorphic Computing to develop robots capable of learning increasingly complex tasks, in increasingly shorter time frames.
At the moment, neuromorphic AI hardware and software are in an experimental stage, but state-of-the-art neuromorphic chips promise rapid developments for next-generation simulated neural networks (SNNs), allowing them to move out of binary logic and approach the extraordinary flexibility of our brains. The goal is to make AI that can think creatively and recognize patterns, objects, and contexts they have never seen before, connecting past and present experiences, just like a human being.
Can an Artificial Intelligence generate ideas for Bags inspired by Fast Food Chains?
APPLE VISION PRO and Tourism – everything will change if everything changes
Memcomputer, the computer that mimics, really, the Mind
Tourism and Artificial Intelligence. Fact Checker for False Hotel Reviews
The various generations of AI
The first generation of AI, which developed until the mid-1990s of the last century, was designed to classify data and make decisions, solving problems within a very specific domain. The second generation of AI attempted to generalize results as far as possible, using deep neural networks and statistics. The third generation of AI, which is currently experiencing strong application growth, involves generalist, explainable models that are able to improve their capabilities, thanks to competing for generative systems.
The next key challenge for future AIs is to more accurately emulate the functioning of the human brain and its ability to learn patterns and patterns from unstructured stimuli, retaining only the relevant part of the information and creating associative networks based on experience. Neuromorphic AI, Neuromorphic Computing, attempts to match the flexibility of human neural networks with the computational power of modern computers, basing its decisions on Spiking Neural Networks (SNNs), which are networks that involve independent neurons that fire depending on stimuli and are able to send signals to other elements in the network.
How Neuromorphic Computing Works
The concept of Neuromorphing Computing (or neuromorphic computing) is by no means new: as early as the late 1980s, researchers have been exploring the idea of making computer systems inspired by the human brain. By definition, a neuromorphic computer is a system that does not follow the Von Neumann machine standard, i.e., the concept on which all the computers we are used to using every day are based. In neuromorphic computers, there is no conceptual separation between memory and processing units: these machines are based entirely on the concept of a neural network, i.e., a composite system that is not programmed with a sequence of instructions to be executed, but with a set of parameters defined for each neuron/synapse pair. At present, inputs to the network are still handled with binary values, but new systems are being sought to make these chips even more flexible.
Currently, AI neural networks are operated by traditional computers that have a clear hardware separation between memory and processing. This division represents a bottleneck during the training phase and results in considerable energy consumption for the continuous movement of information from one component to another. The management of the processing (neuron) and memory (synapse) units of neuromorphic chips is much smoother since there is no hardware separation. This has a major advantage for machine learning systems based on deep networks: it dramatically reduces processing time and training costs, which are currently critical issues for using AIs in real-world environments. Compared to traditional AIs, neuromorphic AIs are more sustainable and much more efficient, features that are even more relevant in the midst of an energy and environmental crisis such as the one we are experiencing today.
Another key peculiarity of neuromorphic systems is the parallelism of processing: by their very nature, neural networks are inherently parallel; therefore, information will not follow a static flow as in traditional computers, but will flow through the entire processing model in parallel. This detail makes neuromorphic systems easily scalable, because adding additional processing chips only increases the number of available neurons/synapses, and thus the system as a whole can be treated as one large network.
The Technological Leap
Currently, neuromorphic computers are made based on silicon, the same semiconductor technology used in CPUs and components of the electronic devices we use every day. But research is moving rapidly to identify materials more suitable for neuromorphic technology such as, for example, resistive memory systems, optoelectronic devices, and innovative materials such as biomembranes, which can be processed to react like semiconductors to the passage of electrical currents. Research in this area is in an exciting experimental phase: hardware implementations of the very different neuromorphic systems are being investigated, using digital, analog, or even hybrid systems to efficiently implement SNNs.
The new NeuRRAM chip, presented in a Nature publication, is one of the most promising fruits of this research. The chip is capable of running different AI models, consuming only a fraction of the energy required by traditional computers to achieve the same level of accuracy in the result. The prototype is able to break down the barrier between memory and processor, performing all the elementary operations needed by each simulated neuron, directly in memory. NeuRRAM’s performance will enable bringing AI to all those devices that require compactness and low power consumption, such as edge computing devices, IoT systems, wearable technologies and smart sensors.
The features of neuromorphic computers attempt to mimic the functional configuration of the human brain not only at the software level, but also at the hardware level. However, we do not know in detail all the biological mechanisms that make the brain such a perfect machine, and it is also on this aspect of the problem that research will focus in the near future. A proper abstraction of the biological workings of our “processing unit” will advance the development of machines that aspire to mimic it.
Neuromorphic Computers and Tourism
Neuromorphic computing: thinking of AI as neuromorphic systems is a substantial and drastic paradigm shift: not only the basic algorithms need to be rethought, but also the very architectures that are the foundation of artificial intelligence as we know it today.
But the opportunities in performance and energy efficiency are unprecedented. A new vision of parallel computing that could change the face of the technology we use every day, improving its usability, making it closer to our expectations and needs.
For example, in Tourism one could calculate with unprecedented estimation the profit margins per proposed package; think of perfectly honed hotel cost control techniques; eliminate with appropriate territorial policies Overtourism; calibrate marketing techniques and product sales through Big Data derived from social and everything else that a brain that lives infinite generations can create.
Address: via Ammiraglio Millo 9 (Italy )
Alberobello, Bari. (Italy)
📞 +39 339 5856822
