 |
| Artificial neural network ANN are a grouping of connected software defined nodes, inspired by a data flow optimized simplification analogy of biological neurons in a brain. In the image above each circular node represents an artificial neuron and an arrow represents a connection from the output of one artificial neuron to the input of another, analogues of the axon and dendritic connections in brains and how they cascade data around via activation potential analog biological information processing on slow cycle rate massively parallel neuron brain cell systems that self assemble in nearly all higher lifeforms on Earth from Epigenetic, and Genetic data from chemicals from foods made of atoms shared by all materials on Earth. Actually the human body contains more than 125 chemical elements from the Periodic Table of the Elements. The commercial importance of ANN can be understood through the value of the many applications that can be optimized with machine learning and neural networks. With market capitalization of more than $1 trillion by 2040, we are talking about neural networking technology that will fully develop the fourth Industrial Revolution #4IR #AI #ML #ANN #SNN Trajectory prediction important in many military and defense application during times of war and planning and practice leading up to WWIII. System identification for automated support of many consumer products. Even tracking the epidemiological data of biological weapons used in world war three will be of great strategic and national security importance. Autonomous vehicle control with autopilot modes. I am talking about all weather LEVEL 5 automobile autonomous driving 24/7 in vehicles that are paid for with subscription services, such that one vehicle serves many people at different times of day, in sharp contrast to many privately owned cars & SUV's today that sit idle without moving for 22+ hours per day, rusting, oxidizing, aging and falling apart as one of the most expensive depreciating items in most peoples possession today, well at least in America and other wealthy countries where private automobile ownership common. Meg & I own three cars, a motorcycle, motor scooter, OG Segway MP, two bicycles, 1 e-bike, one E-kick scooter & a few other engine enhanced things like a 4 stoke weed whacker, 4 stoke pressure washer, and 4 stoke backup generator. All of our machines use unleaded regular gasoline except the B5 Audi A4 which requires premium fuel, and I get almost all my fuel from The Grange in Issaquah, Wa, which is distributed by CENEX ( #E0 #PureGas ) because its E0 or ethanol free and most of our engines are older and optimized to run on non-ethanol unleaded gasoline. Our 2022 Corolla Hybrid has many ADAS features known as Toyota Safety Sense 2.0, though in 2023 they already shipping 3.0 in the newer models. I think lane-keep assist and traffic sign camera reading that displays the speed limit persistently on the dash display are novel safety features, along with collision prevention and automatic emergency braking via data the front mounted radar and camera sensor in the top of the front window, that also works for adaptive cruise control and automatic high beam control. Industrial process control mode automation means lights off manufacturing with robots that learn how to work more effectively by communicating with the other robots, IR cameras, computers and equipment to optimized work flow, throughput, uptime, energy efficiency and profitability. This can include tracking the runtime before equipment fails and audio analysis of machines to detect bearing failures, clogs forming, jams forming, equipment drifting out of calibration and when something is getting close to needing a wear part replaced as preventative maintenance. The central intelligence enhanced manufacturing control computer can send service text messages and automated voicemail messages to the engineers or mechanics so as to schedule the needed work or repairs as they come up. To improve natural resource management of forests, fish stock in oceans and rivers, in mines and aquaponics or agriculture. To gain novel insight into how to eco-optimized processed and emissions to reduce pollution emissions at factors by showing cost effective remediation and equipment upgrades that can reduce fines imposed by the EPA or other government law and agency enforcement aimed at improving public health by reducing toxic emissions from common sources of pollution. To increase understanding and generate better models of quantum computing, quantum chemistry, and quantum energy. This will enable novel products with non-linear optics, adaptive wireless with spectral noise filtering and enhanced frequency hopping to get high quality data with good data integrity and excellent signal quality over wireless networks. AI can also be used to optimize fiber optic multi-frequency telecommunications with novel control of optical photonics systems to improve bandwidth more with non obvious control methods the AI generative intelligence can suggest and implement if approved. In security systems for face identification, vehicle tracking, vehicle plate ID capture, animal tracking, thermal imaging mixed with audio and visual imaging and other sensor fusion with radar and occupancy detection. This can be used to automated the switching of LED lights that last much longer if kept cold off when not producing light of use to anyone. This can also reduce light pollution and improve energy efficiency that saves money on electrical utility costs by reducing wasted energy use. For adaptive radar imaging and synthetic aperture radar antenna array control in products like the SpaceX Star Link system and its ground mounted transceivers. Adaptive radar also being implemented in passenger vehicles for adaptive cruise control and autopilot features. Active traffic analysis, vehicle to vehicle communication, vehicle to grid and grid to vehicle communication, many of these systems of data exchange can be enhanced or improved with intelligent learning artificial intelligence. To play GO or Chess with humans or against other machines. To create stories, content, music, videos, movies, images, pictures, even video games with generative AI applied. ANN applied this way can reduce the costs of developing content by orders of magnitude. To recognize objects in airport terminal security scanning. Data sensor analysis to provide useful insight or actionable information to the security personnel at airports, train stations and other public transportation facilities. Improving natural language processing or NLP for voice learning and intrusive human computer interactions with speech, with productions today like Alexa and Siri as common examples of intelligent personal assistant agents that keep improving with software updates continually. To better understand human gestures, posture, body language, slang speech phrases, and to read text messages, emails, blogs, and documents. For financial forecasting, market analysis, future projection, prediction, data mining, machine translation, visualization, social network filtering. Detecting and isolating junk or spam emails. Finding useful signals in a sea of noise, ANN' trained well can do amazing super human intelligence jobs that will seem like magic to most people. ANN can even be used to help cancer doctors diagnose patients with enhanced medical imaging analysis, in order to help the lab pathologists to distinguish sometimes subtle differences between normal cells and invasive cancer cell lines by automated analysis of cell shapes. In order to accelerate reliability analysis of infrastructure before or after natural disasters. This can help local government to maximize the value of roads, bridges, highway, freeway, intersections with smarter traffic handling and novel control of intersection light system to reduce traffic congestion or improve the traffic throughput of existing roadways. ANN's on public vehicles can find potholes, cracks and other road surface abnormalities, even finding tree branch interactions with overhead power wires, in order to give actionable insight to the cities and utility companies to improve the quality of the roads and utilities. To model rainfall to better understand flooding, runoff, soil settling and other hydrology important for hydro power, rivers, land, and in agriculture. Modeling the oceans and coasts or to formally characterize information about the biosphere of Earth in the geosciences. The way that water moves around the Earth and Atmosphere has incredible importance to society as a whole, for where there is water there is life, and most of the water on earth is salty sea water, only a small amount is fresh water, and filtering contaminated water for domestic use with RO expensive energetically and in terms of operating cost, since those pumps for RO are high pressure and costly to buy and operate, and the RO filters are expensive to replace. The same can be said for kidney dialysis machines and the filters they use to filter blood, i.e. expensive. Again ANN's well trained can improve the kidney dialysis machines, and make the operating of RO water filtering facilities more energy conserving to reduce operating costs. Automatic detection of malware, computer viruses, data crimes, fake information, in order to improve online security of sensitive information access. To stop credit card fraud, network intrusions, botnets, threats, bad actors, malicious software or system corruption. Solving partial differential equations in physics and to simulate open quantum systems. To improve compute aided engineering, computation fluid simulations, nuclear simulations and other kinds of data modeling with adaptive component finite element analysis with intelligent solution suggestion that kelp engineers to improve products faster, especially where human improvement hard due to the mature state of the specific technology. Sometimes, like with internal combustion engines, there are novel ways to further improve thermal efficiency, while also improving power output performance, cleaner emissions, more torque easier at lower RPM's, longer oil life, quieter operation, and other benefits. To reduce fuel use in vehicles and energy use in buildings. To enable smart appliances that continually improve with OTA software update over WIFI. |
ANN are not new or novel but have become a mature technology used in smartphones, robotic vacuum cleaners, self driving cars, smart security cameras and many other common widely sold electronics with CMOS chip functions modeled after human brain neurons.
FNN feed-forward neural network of single layer manual data handling used to find linear fit of plotted data points by Legendre 1805 and for the prediction of planetary movement by Gauss in 1875. Lenz & Isling model 1925 a non-learning RNN or recurrent neural network with threshold elements. Using this model Shun'ichi Amari created adaptive architecture RNN in 1972, later made popular by Hopfield in 1982.
A neural network a data framework for information processing that accomplished in digital computers the kind of analog processes that happen when the human brain takes in electrical signals from sensory organs like the ears, eyes and proprioception. Our brains are in a light proof, sound insulted, bone encasement known as the skull, and can only approximate generalized models about what is actually happening with phenomena in the world.
Inner mental projections into brain working memory in order to predict and anticipate based on past learning and experiences (real world training data in life). Consider how small children learn to walk, speak, read, write, and later to do many other things, but how it takes years of brain training with examples and education to give them skills and knowledge to empower then to create positive change and make big positive impacts on the world in the future. This is also why education is so important and ANN can create students custom education programs to optimize the learning rate for each student. By applying neural networks to education we can make more intelligent people in the future.
Today the internet and all computers and compute devices together use about 1% of the worldwide electrical power generation, but with emerging commercial artificial intelligence, the electrical demand for future computing will increase 5 fold because of training AI with big data on large powerful computers. Tesla Motors actually a big data company more than an automaker. Every EV from Tesla has a sensor suit installed collecting real world training data on the worlds roadways where they are operated, for autopilot development, as the data uploaded back into Tesla's super computers and elastic cloud compute services to generate more autopilot development innovation and solution.
Getting to Level 5 autonomy with vehicles not easy or similar, and just like education a baby into a full functioning adult takes 27 years, it will take many more years before cars can drive themselves better than high skill human drivers in all weather conditions. Neural networks are already used by most automakers to implement ADAS & other early forms of autopilot, like Nissan ProPilot.
Analog just means how things relate to one another. Digital means we represent things on computers with bit streams of 1 and 0 to turn the transistors on and off in the IC logic chips or integrated circuits made of systems of billions of transistors etched as complimetary mixed oxide systems or CMOS into silicon wafers at high volume chip fabs that operate 24/7 making billions of computer chips. CPU, GPU, RAM, SOC and ASIC are the big products today. Mixed signals, digital to analog and analog to digital chips and application specific processors are lower volume products. CMOS imaging sensors are another big category in the chip fab industry.
Chip fabs are very busy to keep up with the growing demand for these important valuable chips that the Chips and Science Act of the USA seeks to make in the USA or America with help from Intel, AMD, NVIDIA, Apple, IBM, HP, TI and other companies like TSMC that is also busy building a chip fab in Arizona. Lovely little ASML the most important company in the world since it's the only group making leading edge chip fab equipment.
Wikipedia entry on ANN says:
"Artificial neural networks are a branch of machine learning models that are built using principles of neuronal organization discovered by connectionism in the biological neural networks constituting animal brains"
No comments:
Post a Comment