Future Development Trends of Artificial Intelligence (Part 1)

With the advent of artificial intelligence (AI) It is becoming increasingly common and embedded in life-changing decisions, driving a growing demand for transparency. In recent years, there have been numerous cases of AI exacerbating bias and discrimination—using facial recognition for law enforcement is just one example. Over the next few years, it is highly likely that we will shift from loose self-regulation to government involvement in AI. Conversely, major tech companies are increasingly leveraging AI to address the very privacy and bias issues that AI itself has created.

 

01Technology Trends

 

  Below are the key technology trends identified by global data companiesAIKey technological trends in the theme:

 

Explainable Artificial Intelligence(XAI)

 

  Artificial intelligence is increasingly involved in life-altering decisions, such as welfare payments, mortgage approvals, and medical diagnoses. Consequently, transparency and interpretability have become paramount. Around AIAIAIdevelopment guidelines to build trust. Simply expanding application recognition (XAI) is insufficient to mitigate bias in data, as it merely enables companies to identify potential discrimination. Businesses will refine their models as needed, and expanding the application of AI-powered recognition has become a top priority for most AI-related companies, such as Google and defense giant Lockheed Martin.

 

Federated Learning

 

  It may seem counterintuitive that AI could serve as a solution to the privacy issues it itself creates; this approach was pioneered by Google in 2017Launched this year, it aims to leverage the rich datasets available on local devices—such as smartphones—while safeguarding sensitive data. In short, a device downloads the current model from the cloud, refines it by learning from the data stored in its memory, and then updates the model based on localized data. Google is the most well-known application of federated learning; for instance, keyboard apps use it to suggest words or phrases based on users’ past input. In the healthcare sector, federated learning enables different organizations to collaborate on a single model that enhances patient care without sharing sensitive data.

 

Facial recognition

 

        2020 年 6 In the month, Amazon and International Business Machines Corporation (IBM) and Microsoft have both announced that they will no longer sell facial recognition technology to law enforcement agencies or the general public.(FR)Tools. Although Amazon has postponed the use of facial recognition by law enforcement for one year, Microsoft stated that it will not sell facial recognition technology to police departments until national laws are enacted to regulate its use. In turn, IBM announced that it will no longer offer general-purpose facial recognition or analytics software. George Floyd(George Floyd)Following the killing and the subsequent anti-racism protests, scrutiny has intensified, prompting major tech companies to distance themselves from facial recognition. However, instances of ethical concerns and bias have been widespread for years.

 

Machine autonomous learning (AutoML)

 

  Machine autonomous learning aims to reduce the need for autonomous learning (ML) The human effort and energy required to apply it to real-world problems. It achieves this by automating the time-consuming, iterative aspects of building autonomous learning models, enabling developers and data scientists with limited ML expertise to work with autonomous learning models and technologies, while also reducing human errors and biases in machine learning models. Fraud detection, pricing, and sales management are common application areas for autonomous machine learning. Google Cloud, DataRobot (DataRobot) and H20.ai is an example of a leading machine autonomous learning software solution.

 

Efficiency, achieving twice the results with half the effort

 

In 2019 , Google’s head of AI, Jeff Dean, said, “We hope to build systems that can generalize to new tasks; it will be both interesting and important to accomplish things with less data and fewer computations.” Most people agree that the regulatory and commercial advantages of learning from fewer data points are significant. The emotion-recognition startup Affecta , Kaljubi, stated that data synthesis methods will enable companies to take already collected data and synthesize it to create new data. For example, a video of someone driving a car could be used to generate new scenarios, such as a driver turning their head or texting—edge intelligence for the Internet of Things (AIoT）

 

One of the key areas where disruptive themes converge is artificial intelligence and the Internet of Things. This combination has given rise to AI for the IoT. By integrating data collected from connected sensors and actuators with AI, it’s possible to reduce latency, enhance privacy, and enable real-time intelligence at the edge. Apple Inc. (Apple) launched 2020 年 1 acquired Xnor.ai, the company offers image recognition tools that can run on low-power devices to strengthen its edge‑computing portfolio. AI, to bolster its edge‑computing offerings.

 

quantum computing

 

Quantum computing is beginning to mature, with Google, Microsoft, and IBM taking the lead. With the support of quantum computers, AI applications can operate much faster. According to Dr. David ·Oshlom, a professor of quantum information at the University of Chicago, the use of quantum computers can accelerate many computational tasks that rely heavily on machine learning. They will enable faster computations and allow for the development of more resource-efficient algorithms, while also enabling AI to tackle increasingly complex tasks. IBM has been providing quantum cloud computing services to its customers since 2016 , and in 2019 , Microsoft (Azure Quantum) and Amazon’s Quantum Solutions Lab have all announced that customers can now test their experimental algorithms on multiple quantum computers.

 

Next-generation chip

 

The focus of chip design has shifted from in 1The race to pack more transistors onto a square millimeter of silicon has shifted to building microprocessors as systems, each composed of multiple components, with each component performing a specialized task. In the years ahead, computers will increasingly emulate the way the human brain stores and processes information. This is where neuromorphic computing and neuromorphic chips come into play. Neuromorphic chips are ideally suited for handling massive amounts of data and powering deep-learning applications that rely on neural networks, all while consuming far less energy than traditional processors. IBM, in collaboration with Truenos (TrueNorth) collaboration, Intel and Rishi (Loihi) collaboration, currently leading in R&D but facing competition from well-known companies such as Qualcomm, HP Enterprise (HPE) and Samsung, among others, as well as GraphcoreCompeting with well-funded startups like Cambrian.