Peter van der Made is the founder and CTO of BrainChip Ltd. BrainChip produces advanced AI processors in digital neuromorphic technologies.
getty
The artificial intelligence (AI) revolution is upon us, and companies must prepare to adapt to this change. It is important to make an inventory of the current skills within the company to identify which additional skills the employees need to learn. The company does well in developing an AI strategy to outline the areas where AI is most effective, whether in a product or a service. Failing to act inevitably means falling behind. The training should include an introduction to AI, its capabilities, and its shortcomings (AI is only as good as its training data). This article gives a view of the current state of AI and what lies ahead.
Artificial intelligence found its purpose in 2012 when AlexNet won the ImageNet challenge with a 16.4% overall error rate, compared to over 26%. The ImageNet challenge is a collection of 1.4 million images in 1000 categories, such as dogs, cars, plants, etc. A neural network is the internal engine of all artificial intelligence technologies. The neural network is said to be based on the way the human brain functions; however, this is far from the truth. Brains are way more complicated and efficient than neural networks. Brains have awareness, imagination, inventiveness and creativity, all missing in neural networks. Brains are also dynamic, consisting of specialized cells called neurons.
Neural networks have grown from a few million to nearly 200 billion parameters. Each parameter must be computed, causing an increasingly great demand for high-performance computing resources and energy. Artificial intelligence programs have beaten humans at chess and in the more complicated game of Go. Programs such as ChatGPT can weave exciting stories and answer complex questions. Training a large network can take months on powerful servers with hundreds of thousands of processors.
The increase in computational resources has made new AI tools and neural networks possible. However, the neural networks responsible for all these impressive results are unaware of what they are doing. There is no awareness, just computation.
Machine learning is a subset of artificial intelligence. Training algorithms need vast data sets to create good-quality parameters that define the function and accuracy of a neural network. Machine learning continues to advance as more data becomes available and algorithms become more sophisticated. AI is used in many fields, including healthcare, finance, manufacturing and transportation.
The future of artificial intelligence appears bright with continued advancements in technology. Investment in artificial intelligence reached $93.5 billion in 2021, according to Statista. The current trend for neural networks to grow larger will likely continue into the near future as more functionality is required.
One of the most promising new technologies is neuromorphic processing. Neuromorphic means "like the brain." Dedicated circuits are used to mimic the way dynamic cells in the brain operate. They do not run any programs but are capable of learning, and just like actual brain cells, they all work simultaneously rather than sequentially. Neuromorphic cortical models of artificial intelligence are based on the structure and function of the neocortex, the brain's outer region responsible for complex cognitive processes and are small, faster and less power-hungry than computers.
Research into these and other brain structures is expected to lead to greater levels of intelligence and better cognitive performance than in previous types of artificial intelligence. With several million nodes, these artificial cortical networks are still far from simulating human intelligence.
Like the composition of the brain, which contains many different structures, it may be necessary to use different types of neural networks to perform specific functions. The neocortex is just one part of the brain responsible for cognition and intelligence. It has massive connections to the thalamus, the hippocampus and the cerebellum, all examples of brain regions important for different cognitive aspects.
Modeling these regions and the neocortex could lead to more advanced AI systems. The thalamus is the central hub that receives sensory information in the brain. Modeling the thalamus could improve the ability of AI to process sensory information, such as auditory, tactile and visual data.
The hippocampus is involved in spatial navigation and the creation of long-term memories. Modeling the hippocampus's functions could improve AI systems' ability to learn and selectively form long-term memory.
The cerebellum has massive connections to all regions of the neocortex. Modeling the cerebellum could lead to the simultaneous processing of incoming data while the AI is learning something new, like driving a car.
While the neuroscience understanding of these brain regions is still incomplete, sufficient information is available to build models that may answer open questions and fill in some of the blanks through experimentation. One day, cortical neuromorphic neural networks may displace the neural networks driving artificial intelligence today and have been responsible for its many successes.
One key difference is their training method. Current neural networks require millions of examples and an error feedback algorithm to adjust the parameters. These training sessions can take weeks using expensive, powerful computers costing millions.
Cortical neuromorphic neural networks learn from few examples and are thus cheaper to deploy. Neuromorphic processing eliminates the need for colossal computer resources. Continuous learning adds experience, creating more accurate outcomes.
Cortical neural networks are expected to be used in products that range from speech recognition to image processing, space exploration, healthcare, and robotics within the next five years. The development of cortical neural networks may lead to the emergence of artificial general intelligence (AGI), the holy grail of artificial intelligence. Humanity will benefit from the emergence of AGI in turbocharging the global economy in providing a multiplier for human ingenuity and safety. AGI is likely to benefit humankind much like previous economic revolutions. The internet and computers changed the way we do business. There will be a similar shift in how humans occupy their time in the future.
AI is a fast-moving field, and companies cannot afford to stand still. To enable personnel, it is imperative to take action now to enable employees to upgrade their skills to meet the challenges of the future. Seek out opportunities to train your employees on this new technology in order to fully embrace it across your organization.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
