Article by engineer Sara A. Al-Emadi, Research Associate at Hamad Bin Khalifa University’s Qatar Computing Research Institute (part of Qatar Foundation), an expert in AI, deep learning, and cutting-edge technology.
Thanks to the availability of large amounts of data, the advancement of computer storage systems, sensors and networking technologies, Artificial Intelligence (AI) solutions, based on techniques such as Machine Learning (ML) and more specifically Deep Learning (DL), have been growing rapidly. Recently, ChatGPT has disrupted the natural language processing field, offering businesses new, efficient and cheap ways of integrating such a system into their on-ground products, assessing workflow management, designing smart dashboards and forecasting future events.
Examples of AI technologies that have reshaped traditional business in terms of day-to-day applications varies, below are few examples:
- Recommendation systems: Learning patterns from data gathered from users have been the essence of the success of modern-day AI systems. An example of this case is streaming videos and content or browsing products on online platforms such as YouTube, Spotify, Netflix or Amazon. In general, a recommendation system works as follows: the user will search for items/movies/songs or click on an item in the webpage and their search/selection data will be fed into a DL model where it will learn the user’s preferences. Next, it will predict and recommend items with a higher chance of the user liking them or tailoring towards the preferability of the user. Such systems have improved the probability of commercial and social media platforms significantly.
-
Mitigating security risks through monitoring and anomaly detection: Due to the advancement of technology, cyber-security attacks are becoming very complex such that the current detection systems are not sufficient enough to mitigate or address them. However, thanks to AI, these anomaly attacks can be detected before they even take place. Therefore, in one of our studies, we analysed the use of AI, more specifically the DL model, to detect network intrusion detection attacks [1]. We observed that such a system was able to detect attacks with a very high accuracy. Recently, AI systems have shown their effectiveness in protecting physical premises against drone attacks by detecting and identifying malicious drones using the noise generated by their propellers [2]. This illustrates the promising usage of AI systems to enhance cyber and physical security.
- Financial analysis: Historical financial data, customer behaviours, market trends, and economic indicators are fed into AI systems to assist in investment decisions in the financial and property markets.
- Marketing and advertisements: AI has enabled businesses to tailor their offerings to individual customers’ preferences by analysing historical data. Therefore, enhancing the customer’s experience through increasing their engagement in the content provided which increases the profits generated by these companies through exposure and the creation of effective targeted marketing campaigns.
Despite the outstanding performance of modern AI models in various applications, these models are designed based on the a training samples collected from the same distribution, consequently, experience a significant performance degradation when deployed in real environments. Scaling AI systems across businesses is a challenge and their ease of deployment is a challenge. A key contributor to this phenomenon is the gap between the distributions of the samples which it was exposed to during the design process and the samples observed in the deployed environment. For example, a self-driving car can be designed based on samples gathered from Europe. However, when deployed in Asia, the AI model will fail. To address this issue and while being limited to small data samples during the design process, my current research at QCRI focuses on proposing new techniques to address this issue in order to bridge the gap by improving the generalisation ability of modern Al to new and unseen data.
Current trends in AI, such as the boom in the integration of ChatGPT in many businesses across different sectors such as chatbots which can be found in healthcare systems, restaurants and hotel reservation systems started to have a visible impact on businesses in terms of how they are operating, the expertise required to attract and hire and the costs associated with such an integration. Additionally, transparency on how these models are learning along with the lack of explainability provided for how the decisions are been made by these models remain as open research questions. Without a clear understanding of how these decisions are being made, deploying these models could lead to many ethical and in many cases dangerous decisions that can harm the consumers, clients, patients and communities. Therefore, in my opin ion, we will witness more and more integrations of these systems in many businesses. However, these systems are unlikely to diminish job opportunities for individuals, in fact, they will enhance how people do their jobs, the same way computers have enhanced our business and led to great breakthroughs in the industry.
References:
[1] S. Al-Emadi, A. Al-Mohannadi and F. Al-Senaid, “Using Deep Learning Techniques for Network Intrusion Detection,” 2020 IEEE International Conference on Informatics, IoT, and Enabling Technologies (ICIoT), Doha, Qatar, 2020, pp. 171-176, doi: 10.1109/ICIoT48696.2020.9089524.
[2] S. Al-Emadi, A. Al-Ali, and A. Al-Ali, “Audio-Based Drone Detection and Identification Using Deep Learning Techniques with Dataset Enhancement through Generative Adversarial Networks,” Sensors, vol. 21, no. 15, p. 4953, Jul. 2021, doi: 10.3390/s21154953.