Main menu


Top 10 most popular AI trends of the 2022 year

Top 10 most popular AI trends of the 2022 year

Top 10 most popular AI trends of the 2022 year

The tech media outlet Toolbox featured the views of 10 experts on “How will AI evolve in the next year?” Edge technology that experts should pay attention to next year was also intensively discussed.

Top 10 most popular AI articles of the year

1. Deep Learning Revenue Decline

The first place was occupied by MIT's Neil Thompson research team featuring an article on the cost of energy to train deep learning systems.

As a result of analyzing the improvements of the image classifier, the research team found that "to cut the error rate in half, it can be expected that 500 times more computational resources are required."

"The rising cost requires researchers to devise more efficient ways to solve these problems, otherwise we will give up research on these problems, and progress will be difficult," he said.

2. 15 graphs to look at to understand AI in 2021

The article was reported with an attached 222-page report organized into 15 graphs covering jobs and investments.

Emphasizing the diversity of the AI ​​workforce and the ethical issues of AI applications, it offers a variety of perspectives from academia and industry.

3. How DeepMind Innovates Robots

This is an article about 'Alpha Fold', a protein folding AI model developed by DeepMind.

DeepMind, a subsidiary of Google's artificial intelligence (AI) development, has developed an AI model 'Alpha Fold' that can solve the 'protein folding' problem that scientists have not solved for 50 years.

According to scientists, each protein, a component of the body, has a unique folding structure according to the base sequence of the linear amino acid (amino-acid) complex.

Scientists have tried to elucidate the three-dimensional structure of the protein, but the number of cases of this protein folding is astronomical, making it a labyrinth.

Alphafold developed by DeepMind used neural network (DNN) technology to learn vast amounts of genome data. Through this, the three-dimensional structure of the protein was predicted based on the amino acid sequence.

4. AI's Eventful Past and Uncertain Future

From 1956 to the present, it was introduced in a way that tells the story of this field. In this article, we look into the potential of hybrid neuro-symbolic systems, paying special attention to the past feud between the symbolists who focused on expert systems and the connectionists who invented neural networks.

5. Andrew Ng X-Ray AI Highp

This is an interview article with AI pioneer Andrew Ng. Andrew Ng, a Chinese-American, majored in computer science, statistics, and economics at Carnegie Mellon University in the United States in 1997 and completed his master's degree at MIT. He currently heads a company he calls Landing AI.

He mentioned an AI system developed at Stanford University, which "can detect pneumonia on a chest X-ray," and explained that "the AI ​​machine outperforms radiologists."

6. GPT-3 in Open AI

Last year, when the AI ​​Lab located in San Francisco opened, GPT-3, an AI language generation system, was unveiled. GPT-3 was able to generate fluid and consistent text in any subject and style, given the smallest prompts.

However, as they were trained with texts from the Internet, they became biased towards certain parts of the Internet.

Professionals use GPT-3 for applications such as customer support, online tutoring, and mental health counseling, and companies that are integrating it into their products have come to the fore.

7. Fast and Efficient Neural Networks Copying the Sleeping Brain

Francis Chance, a researcher at Sandia National Labs, presents research on developing algorithms using the dragonfly's unique brain to build neural networks of ever-increasing size and complexity.

"By leveraging the speed, simplicity and efficiency of the dragonfly nervous system, we aim to design computers that perform these functions faster and with less power than conventional systems consume," Chance said.

8. Deep Learning, Difficult to Advance Unless Replicated in the Brain

Jeff Hawkins interview article. The question-and-answer session with Hawkins included the assertion that consciousness and conviction that super-intelligent AI does not pose an existential threat to mankind is not so difficult. These are some of the most controversial ideas he has published.

9. Algorithm to Create Instacart Rolls

This article is about the 'Instacart' role algorithm. Instacart is an online-based agricultural delivery service company. "The company's AI infrastructure has to predict product availability from thousands of different data points," explained Sharath Rao and Lily Zhang, the company's engineers. "How many shoppers can work We predict what will happen,” he said. 

10. 7 AI Failures

Contributor Charles Choi is an article explaining the weaknesses of today's AI by collecting examples of AI technology failures.

2022’s AI forecast from 10 experts

1. AI will transform citizen development with low-code technology

Jean-Franchois Gagn?, Head of AI Product Management and Strategy at ServiceNow, said, “As enterprises apply AI to their industries, they begin to leverage ready-to-use underlying models to create value for language and vision AI solutions. It has grown significantly,” he said. 

“If AI advances with low-code technology, anyone can become a developer,” he said. “Citizen developers will be able to type in the problem they want to solve, in plain English, and interactive AI will generate the code. It can be done,” he predicted.

2. New regulations may limit the application of AI in some areas

Omkar Kharkar, manager of data science and analytics at NTT Data Services, predicts that “in the coming year, the regulatory aspect of AI that entails trust and ethics will be a bigger factor.”

“The EU continues to pour in regulations, and we have recently published more draft AI regulations,” he said. “The EU has even banned certain types and uses of AI.”

In the case of the United States, Hakar said, "It remains to be seen how the regulatory environment continues to grow and how it can be used as a regulatory framework for AI and in the non-military sector." More recently, he argued, "we are also focusing on the types of 'trustworthy AI practices'."

3. AI Helps Businesses Cut Costs and Overcome Skill Shortage

Tom Shea, CEO and Founder of OneStream Software, predicted that “to get more value from AI, institutions will prioritize democratizing the technology.

“Data scientists are expensive and difficult to find, so organizations are looking for alternatives to create value from data with limited resources,” he said.

“Businesses implementing built-in automated artificial intelligence (AutoAI) tools can help analysts with varying levels of experience analyze data streams, rapidly create machine learning models, and find actionable insights like data scientists. do,” he said.

This will allow businesses to increase revenue, optimize costs, and strengthen operations, and moving forward, others will begin to jump on the curve.”

4. IT Talent Needs

Bernd Rueker, co-founder and chief technologist at OneStream Software, advocated for the need for IT talent. He pointed out, “There are currently no projects using AI for many companies due to a shortage of talent.” 

5. Human augmentation to play an important role

Viralspace Vice President Hiro Tien argued that while AI has definitely advanced in terms of technological innovation and effectiveness, it still requires a human touch.

"Sometimes AI can be harmful or perpetuate bias," he argued, so "human augmentation must play an important role in the effective and responsible use of AI."

6. The healthcare industry needs to leverage AI-driven technology

Sanjeev Agrawal, president and chief operating officer of LeanTa, said, “Health care will also continue to find ways to deliver innovative outcomes by investing in AI. “We need to recognize that adopting new technologies can be truly worth the investment.”

7. AI automation effect, sound data and trust play an important role

According to Xenia Falke, Head of Airspace AI, the impact of AI in 2022 will depend on building consumer trust, adequate adoption and ample historical data. That said, trust will play an important role in leveraging AI to automate operations. “AI automation will depend on accessing large amounts of relevant, clean data because data has always been at the heart of AI,” he stressed.

8. Cybersecurity field where AI detects anomalies

Experts argued that AI provides important functions such as anomaly detection and pattern recognition. He also said that natural language processing allows instant information on the latest threats.

9. Ethics becomes more important for AI and big data

Experts predicted that “more than 90% of AI/ML projects next year will not achieve their business goals due to incorrect knowledge of AI and ML, insufficient planning, and low execution power, which will create negative results.”

10. AI Rooted in Road Traffic Safety Business

“By 2022, AI-enabled cameras and sensors will be embedded in both commercial and personal vehicles to prevent distraction, drowsiness and impaired driving,” said Ryan Wilkinson, Chief Technology Officer, IntelliShift. will take root even more.”

He continued, “In the industry, the technology used in trucks will continue to increase and become safer,” he said. Analytics Insights has published an article on the Top 10 Edge AI Trends and Predictions for the New Year in 2022.

Edge AI outlook for next year

Edge computing is a distributed computing paradigm that introduces computation and data storage where needed to improve response times and conserve bandwidth. Edge AI solutions and applications range from smartwatches to production lines, logistics, smart buildings, and cities. Easy and versatile, most organizations are leaning towards edge AI. Here are the top 10 edge AI trends for next year worth paying attention to.

1. IT-focused edge management

Edge computing is becoming a necessity for many enterprises, but deployment is still in its infancy. Moving on to production, edge AI management will be handled by the IT department, which will transition to cloud-native technologies to address edge computing challenges related to manageability, security and scale.

2. Expanding AI Use Cases at the Edge

Computer vision is active in the field of edge AI. For example, NVIDIA Metropolis, an application framework and developer toolset that enables the creation of computer vision AI applications, has grown its partner network 100x since 2017, and now has more than 1,000 members.

3. Convergence of AI and Industrial IoT Solutions

The intelligent factory is another space where new edge AI applications will run. Factories can add AI applications to cameras and other sensors for inspection and predictive maintenance. After detecting anomalies or defects, AI applications can alert humans to intervene.

For safety applications and other applications that require immediate action, real-time response is possible by connecting an IoT platform that manages assembly lines, robotic arms, and picks and AI inference applications. Integration between these applications relies on custom development work. This is expected to further strengthen the partnership between AI and existing IoT management platforms that simplify the adoption of edge AI in industrial environments.

4. Increased adoption of AI-on-5G

The AI-on-5G Unified Computing Infrastructure provides a high-performance, secure connectivity fabric that can integrate sensors, computing platforms and AI applications on-site, on-premises and in the cloud. Key benefits include very low latency, guaranteed quality of service and improved security in non-wired environments.

5. AI lifecycle management from cloud to edge

For organizations that have deployed edge AI, machine learning operations (MLOps) will be key to helping drive the flow of data from edge to edge. Gathering new and interesting data or insights at the edge, retraining the model, testing the application, and then deploying it back to the edge improves model accuracy and results.

While existing software can be updated on a quarterly or annual basis, AI does it on a continuous update cycle.

6. Edge Data Center Growth

One of the top edge AI trends for next year is the rise of edge data centers around the world. Edge AI predicts that more than 5 million servers will be deployed by 2024.

These edge data centers will grow further with 5G networks, IoT proliferation, data gaps, SDN, NFV technologies, and video streaming with augmented and virtual reality. Higher demand due to lower latency by overcoming intermittent connections and data storage near end users.

7. IMT2000 3GPP - High Presence VR Service Technology Using Edge Computing Resources

The process sequence from user terminal services to 3D video display must be completed in milliseconds (ms). VR motion sickness should be avoided in VR applications that produce 3D images in real time. Edge computing can remotely deliver even high-quality VR with low latency.

8. Changing from Advertising to Digital Advertising

Edge computing based on real-time targeting data will enable a programmatic digital out of home (DOOH) marketplace.

9. Edge-ops

The move to edge computing entails a streamlined 'end-to-end client process from development to implementation. This is called EdgeOps. This will allow all Stripe developers to improve their current applications and use the power of the edge more efficiently. 

10. Digital twin

Updating the digital twin requires a wide bandwidth to relay data. Digital twins can transmit smaller subsets of data, and edge computing does some of the necessary on-site processing.