Last week, Gartner released a unique hype cycle; for emerging technologies. It includes technologies that can completely change the direction of human civilization. “This Hype Cycle highlights technologies that will significantly affect business, society and people over the next five to 10 years,” said Brian Burke, Research VP, Gartner. 

From algorithmic trust to advanced AI, the Gartner’s hype cycle features many new technologies. In this article, we will focus exclusively on the AI segment of this survey. 

1| Embedded AI

Stage: Peak of inflated expectations

Time required to plateau: 2 to 5 years

Thanks to the efforts of top companies to place supercomputers in the pockets, Edge computing systems have garnered significant attention. As we approach an era of IoT, 5G and portable medical devices, it is crucial to facilitate developers to develop and deploy edge applications quickly.

For example, with the NVIDIA Jetson AGX Xavier developer kit, as shown above, one can easily create and deploy end-to-end AI robotics applications for manufacturing, retail, smart cities, and more. Whereas, Google’s Coral toolkit can be leveraged to bring machine learning to edge. The safe, secure and real-time output is the theme of the modern world. And, edge devices offer just that. Check top edge computing products here.

2| Generative AI

Stage: Innovation Trigger

Time required to plateau: 2 to 5 years

AI can now paint auction-worthy art, generate songs and even create faces of people who never existed. All this thanks to Generative Adversarial Networks (GANs). These networks are considered to be one of the critical turning points in the history of ML. In the Gartner Hype Cycle, Generative AI featured in the ‘innovation trigger’ segment of the graph. 

The proliferation of generative networks has had many unwanted results. Malicious online players can now generate disinformation in the form of images and videos that can fool many. Generative networks are all fun and games until they start interfering with matters of national interest. For good or for worse, these networks are here to stay for a while now.

3| Responsible And Explainable AI

Stage: Innovation Trigger & Peak of inflated expectations resp.

Time required to plateau: 5 to 10 years

Machine learning algorithms are infamous for their black-box nature. With increasing business applications such as autonomous driving and medical diagnosis, there is a growing demand from the stakeholders to know what is at stake. Apart from the debugging and auditing of the models, data scientists need to look out for meeting data privacy standards in the context of explainability. Be it medical diagnosis or credit card risk estimation, the amount of personal information that is processed can be very sensitive, and this is being addressed in the organisations which are serious about the implementation of explainability.

From inductive biases to GDPR compliance, AI enterprises have many things to take care. To make the process of inculcating responsible AI systems more viable, tools such as Model cards, AI 360 and many others have been released by companies like Google, IBM etc. Responsible AI too, like Generative AI featured in the first segment bordering on ‘peak of inflated expectations’. Whereas, explainable AI featured on the descending curve of inflated expectations. Though responsible and explainable AI practices overlap in many ways; the end goal is to enable transparency in AI-based decision making. 

4| Self Supervised Learning

Stage: Innovation Trigger

Time required to plateau: 5 to 10 years

Most of the AI R&D endeavours converge at one thing — making AGI possible. Equipping machine learning algorithms with human-like reasoning is tricky. ML models are good as the data it is fed with. In areas such as medical imaging, the availability of pre-trained models are almost negligible. Less data, more accurate results; this is where self-supervised learning comes into the picture. Speaking at (ICLR) 2020, Turing awardee and Facebook’s chief AI scientist, Yann Lecun said that supervised learning systems would play a diminishing role as self-supervised learning algorithms come into wider use. It is no surprise that self-supervised learning is placed in the early stages of ‘innovation trigger’ in the Hype Cycle.

5| AI Augmented Development

Stage: Peak of inflated expectations

Time required to plateau: 5 to 10 years

In a report published last year, referring to AI augmented development, Gartner said that the application leaders should embrace AI-augmented development now, or risk falling further behind digital leaders. AI has now found its way into the software development lifecycle. They are assisting organisations in design, development and deployment of their software products. According to Deloitte, startups offering AI-powered software development tools raised US$704 million over the 12 months ending September 2019.

AI has the potential to take care of the mundane debugging tasks through automation. Today, some tools offer low code and no code services. Some AI-based APIs even search the code as one types it. Going forward, the manual assessment might not match the rate at which the industry is innovating. So, it is safe to say that AI-augmented development will witness a growth in demand.

6| Composite AI

Stage: Innovation Trigger

Time required to plateau: 2 to 5 years

As the name indicates, Composite AI aggregates multiple AI systems trained individually with small data sets instead of pooling data. The training of AI systems typically requires a large amount of aggregated data (‘big data’). As discussed above, it can be challenging to find enough data in niche areas. As shown in this experiment, a composite AI system can be constructed based on multiple neural networks from individually trained AI cores, each using a small data set. These systems outperform individual AI cores in the classification of molecular images. According to the study, this approach can be applied to AI development across multiple institutions without necessitating data sharing or pooling to create an extensive training data set. 

7| Adaptive ML

Stage: Peak of inflated expectations

Time required to plateau: 2 to 5 years

Real world is ever-changing. How can we expect traditionML algorithms to function reliably with inconsistent training data. This is why adaptive ML came into existence.

According to Alan Turing Institute, Adaptive real-time machine learning incorporated efficient reinforcement learning, online learning (dealing with continuous sequences of real-time data), and adaptive learning from a small sample size. The techniques border on real-time meta-learning algorithms that can be utilised to achieve continuous learning, predicting and controlling when functioning in a changing environment. 

Adaptive ML applications include rainfall prediction for urban infrastructure, adaptive yield prediction based on real-time crop vigour analysis in precision farming and more.

8| 2 Way BMI

Stage: Innovation Trigger

Time required to plateau:  5 to 10 years

A brain–machine interface (BMI) is a device that translates neuronal information into commands capable of controlling external software or hardware such as a computer or robotic arm. BMIs are often used as assisted living devices for individuals with sensory impairments. Companies like Neuralink are developing implantable brain-machine interfaces. Building BMI devices requires a combined effort of experts from medical, material, ethical and various other fields.  This in turn will lead to new job opportunities and technologies. The Gartner hype cycle forecasts BMI tech to hit plateau within 10 years. If this is true then we will be closer to deciphering the brain.

9| Digital Twin Of A Person

Stage: Innovation Trigger

Time required to plateau: 5 to 10 years

Digital twin technologies enable authentic digital copy of physical entities that can be represented both in the physical as well as digital space. Digital twin innovation will have great implications in the AR/VR industry and even in brain-machine interface applications.

10| AI Assisted Design

Stage: Innovation Trigger

Time required to plateau: 5 to 10 years

AI is now being extensively used in traditional software like Autodesk and many more. AI high dimensional features can be leveraged to conduct multiple design studies. Algorithms have also been used to generate architectural designs. The designs are not restricted to the physical world. Deep learning can now be used to design web pages, creating brand logos and many more. For example, Volkswagen Microbus uses components such as brackets reshaped in generative design. AI driven generative design has proven to cut time delays by a significant amount.

11| Differential Privacy

Stage: Peak of inflated expectations

Time required to plateau: 2 to 5 years