Our take on the Gartner Hype Cycle: 11 technologies that relate to neural networks
By understanding neural networks, you will understand the core of 1/4 of all emerging technologies!
Predicting the future is extremely valuable. Some researchers even argue that it is the fundamental purpose of the human brain. However, as it turns out, predicting the future in the fast-paced world of technology is difficult.
Every year, the research and advisory firm Gartner release “The Gartner Hype Cycle” a diagram visualizing the development of emerging technologies. The underlying assumption is that all new technologies go through the same pattern of innovation and inflated expectations but eventually reaches the “Plateau of Productivity” at which it finally yields real business value.
We have had a look at the brand new 2018 diagram as well as the one from last year. There are 44 unique technologies on the two diagrams, covering a wide range of technologies from known technologies such as “Virtual Reality” and “Blockchain” to more speculative entries, such as microcomputers creatively named “Smart Dust”.
While the underlying theory might be questionable, the diagram is a highly regarded tool in business advisory, and it can safely be assumed that Gartner spends many resources producing it. The placement of each technology might be up for discussion, but our take on it is not so much the placement of each of the technologies in the diagram but rather, which technologies are considered.
Understanding Neural Networks
The Hype Cycle serves as a good introduction to understanding the landscape
of novel technologies. However, to understand the diagram, you must have a basic understanding of each of the technologies mentioned. This can seem like a big task, but we believe that there is a shortcut.
As it turns out, 11 of the 44 technologies are based on or dependent on Neural Networks!
This means that by understanding neural networks, you will understand the core of 1/4 of all emerging technologies! Trying to understand the technologies without understanding Neural Networks is like trying to learn statistics before you know basic algebra.
We have listed the 11 technologies that relate to neural networks below.
1) Deep Neural Networks / Deep Learning
“Deep Learning” is technically synonymous with “Deep Neural Networks”. “Deep” refers to the fact that Neural Networks stack deep layers of neurons. While single layer neural nets (so called perceptron’s) have their place, the true power of the technology come from stacking multiple layers.
2) Machine Learning
“Machine Learning” is the broad term covering all the algorithms that learn from data. Neural networks/Deep Learning is the most successful kind of machine learning algorithms.
3) Deep Reinforcement Learning
The reward or punishment provided after completing a job is not always immediate. This can make it difficult for the Neural Networks to learn. “Reinforcement Learning” tackles this issue by making the Neural Network remember and learn from past decisions.
4) Neuromorphic Hardware / Deep Neural Network ASICs
Rather than Implementing Neural Networks as software programs on conventional general-purpose computers “Neuromorphic Hardware” or “ASICs” (Application Specific Integrated Circuits) are computers that are built specially for running Neural Networks, thus making them much faster than general purpose computers such as a server or your laptop.
5) Edge AI
Refers to running the AI algorithms directly on the device that captures the data rather than sending the data to a server for processing. That is, rather than having your mobile phone send images to a server for processing, it happens directly on the device.
6) Artificial General Intelligence
“Artificial General Intelligence” is the idea that one day, computers might become as smart humans. This is highly speculative, but our closes attempts at replicating human intelligence are currently based on Neural Networks.
7) Autonomous Vehicles / Autonomous Driving Level 4 & 5 / Autonomous Mobile Robots
Autonomous Vehicles are also known as self-driving cars. The visual systems that allow cars to detect pedestrians, road signs and other cars are all built using Neural Networks The different levels refer to how much human intervention is needed, 5 meaning no intervention.
8) AI PaaS
PaaS means Platform as a Software. This refers to out-of-the-box AI toolboxes that can be used by non-programmers.
9) Virtual Assistants
Siri, Google Assistant and Amazons Alexa are all examples of “Virtual Assistants”. They all rely on Neural Networks to process the utterances from users.
10) Cognitive Expert Advisors
“Cognitive Expert Advisors” are “Virtual Assistants” for specialized business domains.
11) Conversational User Interfaces
“Conversational User Interfaces” is the technical term for chatbots. At BotXO, our chatbots use Neural Networks to understand user utterances.
That is why we believe that if you want to understand what the future holds, it is paramount that you get a good understanding of Neural Networks.
Follow our blog to gain insights into the inner workings of Neural Networks and read our next blog post “What is Machine Learning“.