Neural networks can be used for intelligence related computational information handling for inference, image recognition, synthetic vision, wifi network optimization, energy efficient building control optimization, vehicle energy optimization, design optimization, and many other complex system optimizations not possible with classical engineering applied to cost efficient solutions.
Think, unmatched innovation transforming society in ways that will seem magical, cures for most disease, clean abundant electricity, novel wealth creation, currency stabilization, new chemicals and energy resources, greatly enhanced energy efficiency via intelligent control of vehicles, buildings, factories, manufacturing. Novel mixing of technologies and solutions, hybridizing technology from different fields with new emergent products and services that are vastly better. Optimizing all aspects of human activity to reduce pollution emissions and improve profitability of business. Making laws informed with better information to help the economy, AI can help governments, police, fire, the military, e-commerce, retail, inventory management, logistics, shipping, tracking. Nearly every industry in the world will be revolutionized by full AGI that enables the technological singularity.
That means humanoid robots that can do everything a human is able to do, that learns thousands of times faster than a human, that continually gains news skills and knowledge, that updates its AGI parameters and learning models with continual self improvement.
Watch this video https://youtu.be/z8fYer8G3Y8?si=46TBlvHseaeQ7PnL
Also study this video on Randomness https://youtu.be/iT20A4KQxyM?si=B2Z2FljhdosiXBWI
Say it with me, STEM science technology engineering and math. Facts, physics, chemistry, energy, true or false, real or not, black and white, heat or not, light or not, God or not, you must make a choice, your not making grey choices, you make volitional free will choices with consequences.
In chemical reaction systems the same is true, the system only does what it can, it does have an opinion or emotion, its just a system of chemical interactions with valence electrons exchanged between atoms elements and molecules of the inputs to form the outputs or reactants to products, often enable by a catalyst or enzyme made of a special protein in biochemical systems in living beings or lifeforms in the biosphere of Earth.
Watch this to https://youtu.be/QqY8fY0TqaQ?si=VfEjeKXFZZjjcuwI
Enzymes and catalysis loan their outer electrons to lower the energy or activation so chemical reactions can proceed forward with less heat energy or energy input or can be more energy efficient. This is how the human body uses enzymes to enable all the biochemistry happening in cells thousands of times per second at just above room temperature, for without these protein enzymes or bodies would have to operate at near boiling temperature or around 100 C or 212 F. Instead we have body temperatures around 98.6 F or 37 C.
See this about NVIDIA https://youtu.be/nVtb2vNUOdU?si=Sl1mFn2R0wUFtXQD
NVIDIA up %240 this year / Symbotic AI up %210 / Meta up %164 / C3.AI up %149 in 2023 alone indicating a potential AI sector worth trillions of dollars within less than 7 years by 2030! So after LLM like ChatGPT or after OpenAI, what is next // look to the following companies for examples.
Alphabet Inc (AKA Google) Apple (iOS, Fabless TSMC Silicon), Microsoft (Software Defined AI on Azure Elastic Cloud Compute SAS) Salesforce CRM Software AI Enhance Customer Solutions, IBM (North Pole Analog Compute in Memory CMOS)(Watson on 60W Edge Computer {Discrete GPU Laptop} by 2030)
Try this video https://youtu.be/gp_bQXcW_-s?si=ceGDTaN432KVrYm8
First machine learning training done to train for classification of sounds, images, text and video content. More recently generative AI able to create art, like music, images, stories, blogs, news, and videos & more that sometimes contain factual inaccuracies like Microsofts Erika, publishing fake news with false inaccurate untrue ideas. This means that as we feed data into these training models, the resulting finished models are able to create new novel data.
In this way millions of new chemical crystals have been proposed to improve battery chemistry, materials science, pest management, chemical processing, industrial mining, energy related emissions capture, CO2 chemistry products, soot and PM capture chemistry, new ways of decarbonization, novel magnets and room temperature superconductors, and many other novel drugs and biologics that are going to be revolutionized with these obvious crystal chemistries that AI crated in ways that will help to improve products and services across broad fields of society in most world economies.
As robots gain ever new skills to use tools to affect reality, even labor will be revolutionized by the robots brain edge compute AI models, predictive logic, analytical synthetic understanding, natural language communication processing, machine vision, broad generalized object recognition with high speed adaptive learning, flexible brain inspired software plasticity that can update the firmware to add features and functions that people never even imagined. In this way the some human assitant robots with AGI will become increasingly more intelligent, able to help more people in ways we can't even envision yet.
As AI moves from software, across hardware and into the physical world with call this Physical AI, where robots can play musical instruments as well as the best human musician, participate with energy limits in the Olympics in a variety of sports as gender neutral robots. With edge computer AGI the robot brain will integrate many channels of real time data from imaging sensors, microphones, force sensors, radar, lidar, sonar, RF, and other imaging techniques using microwaves, UV, IR and other forms of energy, taking these data signals and integrating them into the AGI models that make up the robots executive function and reasoning productive AI CMOS digital + analog on CMOS chips with computer in memory no clock massively parallel, inspired by the human brain, information processing. Multimodal large models utilized to make decisions and immediately implement them in real time "on the seat of their pants" like a human does when driving a car today.
Consider the Eureka algorithm from NVIDIA on H200 GPU, that leverages the GPT4 large language models to help robots learn thousands of times faster. Training for these robot brains happens across massively parallel cheat to setup virtual world simulations that enable rapid learning and progress not possible with trial, error, adaptation looping or manual brute force robot programming thats is slow and costly by comparison.
Using an LLM with iterative feedback similar to how human learn to speak, read and write, but one machines in virtual words, training can happen thousands of times faster, massively speeding up AI training.
Robots in the real world will feed data to cloud computer virtual machine training simulations so that real world data can create increasingly large generalized models that pave the way to AGI or the technological singularity or Artificial General Intelligence.
AI Technology and AI Startups are getting a lot of money today, since many companies realize the huge potential of these new tools to radically improve their profits with increasing sales of vastly better products and services. Even the more conservative tech giants in Japan known as SoftBank and NTT are investing in emerging AI technologies.
The AI sector benefiting from new business partnerships with many different AI products and services, not a single AGI to rule all. This kind of diversification fosters rapid innovation from competition and global collaboration, between Microsoft and OpenAI for example.
Consider how big tech investing billions into AI startups important to the entire AI ecosystem, like Anthropic, DeepMind, Hugging Face, AI21 labs, Runway , Cohere, Inflection, Databricks, Imbue, Adept and OpenAI.
Look at Microsoft Cobalt starting with 100 on 5nm process with 105 billions transistors custom in house developed Azure Maia cloud AI accelerator chips optimized for LLM training and inference. How about AMD MI300X accelerated Azure VM's or virtual machines. Look at Intel development and deployment tools for the entire AI workflow, alongside a range of hardware unified by the oneAPI programming model.
Custom silicon chips the new way to be cool, to differential products and design the chips with software and development tools to make integrating them into cloud compute and edge compute applications even easier. This means higher performance computers able to do more faster. This will eventually liberate Microsoft from costly NVIDIA solutions.
What about Aleph Alpha raising $500m in series B as Europe's star AI capital capture tech company. Or the way that Meta's open source llama tool model that upsets the AI horse race. Why pay for a restricted or licensing for fee's AI model when free, unrestricted alternative are compatible, the way that GIMP does a lot of the same things that Photoshop does. Giant models are missing the fast iterative update ability of small variants with less than 20 billion parameters. The world tech companies should be enabling 3P integration collaboration, since multiply the speed of AGI develops has profound implications.
Unrestricted AGI agents pose a security risk to data integrity, news integrity, able to easily quickly publish massive amounts of fake news, lies, misinformation, disinformation, false, misleading, confusing stores and articles that trick and manipulate people into believing things that are not true, not factual and not real. I am reminded of conspiracy theorist who claim the shape of the Earth is flat without offering explanations about what is on the edge of the disk or the other side. Obviously some people are insane! The insanity can be seen in early AGI agents left to their own devices with unrestricted access to the internet or World Wide Web. The three laws of robotics need to be applied to AGI models so they help people, not harm society.
IoT or industrial internet of things, Symbotic AI a leader of cutting edge AI and robots for automating warehouse logistics for Target & Walmart, by moving robotic inventory carts carrying shelving of automatically integrated, tracked and organized inventory, with unmatched retrieval speed and precision. Amazon famously uses similar inventory robotics systems in their fulfillment centers by Kiva Systems which was acquired and internally developed.
Tesla FSD or Full Self Driving the market leader in automotive autopilot or autonomous vehicle control technology. The hardware 3 & 4 custom inference silicon, and the OTA updated software, making Tesla a Big Data company with automotive EV robots collecting real world data with sensor fusion on Model S, 3, X, Y & now Cyber Trucks and Tesla Semi. Tesla building an entire ecosystem for AI with big training data models, custom AI accelerator chips, custom in-vehicle silicon, sensor data fusion, OTA data collection and feeding the real world EV sensor data into virtual world training for their next generation autopilot FSD algorithm stack development for L5 or level 5 autonomy as the market leaded in automotive autopilot commercialization.
The development trend was from cloud training inferential AI, to cloud based generative AI, to edge compute physical AI in robotic systems that interact in the real world, such that we will iteratively improve towards full AGI or artificial general intelligence also know as the technologic singularity. This means robots with AGI will be able to perform any task that a human can perform. AGI sorta a moving target in the sense the ChatGPT would have been considered a Touring Complete AGI 20 years ago. Future full AGI with broad skills and immersive interactive generalized intelligence will enable rapid scientific progress not possible by any other means. In this way humans using AI will be able to do more with fewer resources, thereby reducing pollution and making the world a much better place to live.
NVIDIA vertically integrate the design of an AI ecosystem from the software stacks to the hardware and even application engineering, who a holistic AI solution set of novel products and services that can help their engineers to design even better GPU and AI Accelerators. This means class leading process node CMOS AI chips with continually updated firmware OTA. Virtual worlds and modeling becoming an increasing focus of AI development since these technology allow energy efficient AI training. Interestingly these novel model environments will enable rapid scientific research experiments done with software modeling of chemical physical systems to learn even more about real world phenomena with novel insights used to informed improved designs of real world products and services, like GPU's with better light rendering with lower wattage chips that provide superior E-Sports playing experiences, smooth game play, excellent image quality, unmatched rending efficiency, and automated game asset development that makes new games cheaper to develop. Certainly we will see a boost to STEAM games both in terms of performance of games on many existing computers, as well as more game offerings to choose from.
Amazon spends 14% of its revenue on R&D, as does Alphabet. Meta spends 30% on R&D. Apple spends 7%. Microsoft spends 13%. NVIDIA spends %27. Broadcom spends 14%. ASML spends %15. Tesla spends 4%. even PEPSICO spends 1% on R&D.
Consider the value of Interactive AI that works with regular human people with natural voice in all human languages, as an generalized intelligence agnostic of platforms or differences freely willing to help anyone that uses it as a tool to improve nearly anything. Superlative AI will be given the power to make decisions in government and make real world actions happen that influences how society and human activity works.
Generative AI helping AI hardware developers and cloud computer service provides to optimized their parameters, control schemes, service speeds, security, access control, bandwidth, data integrity, uptime, operating cost, and many other aspects that improve customer service while also increasing revenue and improving profits.
The Technological Singularity or creation of Super Intelligent AGI or Artificial General Intelligence already happening with the emergence of Google Gemini, Open AI, Chat GPT 5, and explosive growth of LLM large language models used in NLP natural language processing, LIM or large imaging models used in Synthetic Vision Object Detection & Identification or SVODI, and AVC or Autonomous Vehicle Control in AAP Automotive Autopilot Level 5 pre-commercialization LVCM or large vehicle control models.
We are going to see 20,000 years of technological progress happening in the next 20 years, as generational AI gives us access to new crystals to improve displays, batteries, room temperatures superconductors, new drug discovery, novel biologics, cures for common diseases, the ability to modify the body with upgraded vision, upgrade executive function in the brain, enhanced brain memory performance, stronger bones and muscles, improved endurance and athletic performance.
We will not only slow aging, we will reverse aging with genetic reprogramming, enabling people to live many hundreds of year, then thousands of years as we solve and creation solutions or cures for all diseases, including a complete understanding of the aging process to defeat aging mechanism and give people more time alive healthy and vibrant even in old age as though they were teenagers right up till the end in death. We might even learn how to cheat death, clone ourselves, enhance the clones and explore the universe in a big vast way.
Nearly all aspects of human society will be completely revolutionized by AGI or the technological singularity. Forecasting shows that many Large Generalized Models of LGM AI just a few years away from widespread worldwide commercialization as Google Gemini integrated into many new products and service from the Cloud to edge computer in user devices running on small batteries.
Neuromorphic computing really shines in computing at the edge, on low power devices, in robots and self driving cars. Imagine a robotic vacuum that actually understands what its seeing with its imaging chips to avoid fabric, cords, small object, fibers, thread, or other things that could be easily tangled into its brushes or rollers. Enhanced object avoidance that other side of AI security cameras that can identify different people correctly, so you can have it unlock your door with your Face ID matching its stored profile that can recognize you with different hair style, while your wearing a hat, or hooded jacket, glasses, if you have facial hair or not, and becomes flexible at accurately identify you, but sends your phone a notification if other people are present in its field of view with images of the people at image texts with other useful information, like the age of the person, which can be determined by inferential models in the smart camera. This means the camera will be able to identify pets and wild animals so it will not turn the light on unless there is a valid reason to, and can speak with people nearby using NLP.
Edge computing AI means on your smartphone, laptop, tablet or low power computer, even a smart thermostat to control the HVAC in your home, in smart appliances, in smart lights with occupation detection and auto off when no one around.
Intelligence can be added to a variety of end user applications that help improve the lives of the general public worldwide by making their fridge use less energy in its compressor with variable compression and variable throttling of the heat pumping, just as a smarter microwave with inverter control can throttle the magnetron to vary the heating RF wattage to better match food loads being heated to thaw frozen foods or heat meals or cook different food items, with image recognition such that the smart microwave can see what is inside and with auto heating modes also more evenly heat the products by varying the rotation of the circular glass plat that rotates the food. It will be able to detect if water or other fluids are being heating and use optimized modes for each kind of thing being microwave heated.
Putting more intelligence in a product means that it will be able to operate with better energy efficiency. Imagine a smart washing machine that uses multi-spectral imaging and water quality analysis to determine exactly what kind of items your washing, how much soap to use, what water temperatures to use at different stages, the amount of rotation and rotational speeds, to optimize spin extraction of moisture so that hang drying faster and more effective or to use less energy in the heat pump dryer.
Imaging smarter vehicles where the computer can enhance driver & passenger safety in all weather conditions, drive the car itself with autopilot, prevent collisions with automatic emergency braking, help to keep the vehicle centered in the lane or avoid pot holes with lane keep assistance and many other advanced driver assist features running on edge compute inside the network of many chips in modern automobiles that have more than 100 chip modules that do compute over OBDII today, but LAN much faster networking in future emerging vehicles will enable Level 5 autonomy and autopilot thats as good or better than the best human drivers in all real world driving conditions and unpredictable situations with superior consistent safety. AI can be used to reduce fuel or energy use to make the vehicle less expensive to operate, as well as using microphones and audio analysis to help determine problems for maintenance and repair with intelligent analysis of the waveforms or sounds emitted by the engine, transmission or drivetrain, or analysis of NVH data to tune the variable electronic suspension for varied road conditions to improve ride quality or reduce in cabin noise with active noise canceling. AI applied to the vehicle HVAC can make heating, defrost and cooling more efficiency, while also reducing emissions. Intelligence created by the AI on the vehicle can give mechanics useful insight into what needs servicing or repair, including notifications to the driver that its time to change the oil in the engine or transmission or gear reducing or transfer case or differential, or time to change a belt or chain, or other wear items. The AI can even schedule appointments with the mechanic and drive itself with L5 autonomy to be serviced.
Robots with AI will be able to help accomplish even more to solve all the worlds problems, especially helping disabled people at home, providing better patient care in hospitals, helping diplomats and leaders to make more informed choices when crafting new laws, revolutionizing transportation energy efficiency and helping to decarbonize energy worldwide. AI poised to completely enhance many aspects of human activity with more intelligence applied. Energy efficient SNN or ANN operating on CMOS digital and Analog and Device Specific Custom ASIC chips means that robots with finite battery capacity will have human skill motion control and NLP to communicate with people using natural language with Turing complete intelligence so that people can have deep conversations with the robot about complicated topics, while the robot will be able to help around the house with dishes and laundry and yard work, even vehicle maintenance, home painting, siding repair, roof repair, plumbing and electrical repair.
Edge compute case robots with enhanced AI will become a valuable high performance personal assistant to many disabled people and older people at home, where the robot can cook meals, go grocery shopping, operate the appliances, and plug itself back in to a regular power outlet to recharge, automating, without any human interacting with the power cord for charging. Personal assistant robots will be very useful to improve patient care in similar ways, acting as helps for the nurses and doctors and to better communicate with patents, to prevent medication mistakes and improve workflow and talent utilization of the hospitals staff, for scheduling, routing of ambulances and other security functions and services.
Level 5 autonomy in an automobile for example, will be leveraging much digital CMOS chips with less analog chips doing specialized noise reduction sensor processing for imaging chips, radar, lidar, sonar and other synthetic autopilot vision technologies that enable the vehicles onboard computers to drive as safe as the best human driver or pilot in all weather conditions in the real world.
Training the algorithms has to be done on digital computers because of the varied operation and parameter required to teach the neural network how to identify things, how to predict or infer things, how to manipulate data for optimization of complex systems, for generational creation of game assets and content, to make music, art, movies, commercials, news, stories, posting, blogs, even to write software and self optimize continually as more is learned by the machine entities that simulate natural brain intelligence though the use of complex neural networks like SNN spiking neural networks, ANN artificial neural networks, with Convolution Networks, Deep Learning, Machine Learning, and many other artificial intelligence technologies being well funded and actively developed by most big technology companies that are in an AI race worldwide.
Whoever controls information in the future controls all world economies!
Reality consists of energy & information. Matter is an illusion made of energy via E= MC^2 as Einstein demonstrated mathematically with classic physics equation. What's more is that improvements in quantum science and quantum technologies will enable even more novel integrate circuits, ASIC, FPGA, EPROM, RAM, SOC, CPU, GPU, with ever increasing energy efficiency or more computations per watt of energy.
A specialized analog chips running special tuned algorithms made on large digital server cloud compute with huge data training sets, can enable an analog chip operated signal processor used on an automotive vision chip imaging sensor to process visual data to identify objects, people, vehicles, infrastructure, signs, animals and other real world things or stuff that are happening in, near or around the vehicles place of operation while its stationary, backing up or traveling forward.
Sensor fusion means that autonomous vehicles can see 360 deg in all directions at the same time by mixing signals for different sensors. Mixed signals processors are essential for mixing these inputs to give useful data with energy efficiency operation (a few to tens of watts) to the vehicles increasingly sophisticated MCU (low power optimized) or master control unit that controls the ICU or inverter control unit (compute modules), ADAS control modules, Autonomy Control, ECU or Engine Control, and many other integrated compute modules that combine with in vehicle networking superior in bandwidth to OBDII with LAN technology or local area networking that improved computer networks data throughput in many stationary buildings to bandwidth speeds billions of times faster than common vehicle OBDII data systems.
Effectively automobiles have become similar to a mobile data center with hundred of chips and fast networking connecting the compute modules. Features for drive assist technology have driven the increasing use of wiring, compute, chips and software in common on-road cars, SUV's, and trucks sold to the general public commercially worldwide. This is why I chose vehicle autonomy L5 as an example of where analogy chips will work with digital CMOS to enable autopilot with excellent safety.
In healthcare patient health data analogies with AI can help doctors screen patients for enhanced preventative medicine, to detect problems with health before they harm the patient. This means better outcomes for more people at lower cost, and easier workloads on the doctors too, who can use these new tools to help patients faster, cheaper and more effectively. Data from the patients fitness tracker, like Apple Watch or Fitbit, can be integrated with medication information and health history data, so that an AI can perform complex data analysis to give he doctor and each patient useful insight for actionable health improving suggestions and information. This way blood test data, and other bio-data collected from patients can be combined with diagnostic testing data and bio-tracking data from fitness trackers and other health tracking devices, to create Wholistic Science Based WSB AI Enhanced Patient Care or AEPC!
In financial markets AI can be used to stabilize the economy and currency systems. This can help government to better serve the general public with laws against fraud and other financial crimes integrated into the AI framework to help authorities activity identify bad actors faster, cheaper and more easily.
AI applied means that many applications will become faster, cheaper and easier to do more effectively. This will make many aspects of life more affordable for everyone. AI means that doctors can help their patients more, teachers can teach the students more, people will gain useful insight to tune and optimize their lives continually, enabling people to use less energy at home and work so they save money on utility bills, to use less fuel or energy in their automobile so they save money on charging or refueling.
Many manufacturers can tune HVAC, processing equipment and product handling, with AI that can suggest optimizations after studying operations. AI thermostats or learning thermostats for example can handle the blowers, heat exchanges, heat generators, cooling systems, pumps, switches, controls and other sensor data, to continually create better control schemes or building control algorithms. This means the operating costs of many manufacturing operations will become more energy efficient or lower operating costs, further improving profits and revenue.
In many large business communications interruptions or communication failures between employees means a lot of skill and talent not applied effectively. AI can be used to maximized productively with automated communications that helps them connect to other relevant employee's to collaborate in novel ways that make operations more efficient and more effective. This means that products and services can be improved faster with the more intelligent leveraging of employee skills, knowledge and abilities. In this way many businesses will be able to make better use of their employees, tools, and equipment to generate more profits.
Watch this video https://youtu.be/yG5Pvk9Bkqk?si=XIdENY3KMnNovcjD to gain deeper insight into this emerging technology, phase change memory, analog chips, and digital computers essential for training artificial neural networks, generational AI models and many other uses of arterial intelligence or AI in commercial applications.
Watch this video too https://youtu.be/p0W5eHn5sZ0?si=bkuTOBM1lQFb6i-a
And this one https://youtu.be/n29WWr4g6sc?si=tyM3a-ukrrIo-RPt
See this as well https://youtu.be/Hz8yZUXsYrs?si=GM57peMZRhzWt9b4
My goal to this post. If you read it all and watch the linked video, you will gain a deep understanding of emerging AI technology, science & engineering and how all these topics are related, with huge opportunities to figure out which companies to invest your money into to make large returns on investments or ROI.
Take what I post with a grain of salt, these are just theories and conjecture based on real world data and trends, not absolutely predictions since I am not God and do not know everything. I do however study a lot more than most people and continually learn online to gain deeper understanding into how everything in the world works.
No comments:
Post a Comment