How can artificial intelligence change cloud computing?

I haven't conducted any literature review on above yet & haven't seen any empirical evidence yet in which I hypothesize AI (including its subsets of Machine Learning, Deep Learning etc.) can impact cloud computing in the following manners:

There are many optimization problems associated with cloud computing. Resource allocation, Load balancing, Minimization of SLA penalty cost, Workload in multimedia cloud are some examples of optimization problems. All these are NP-hard problems. These problems can not be solved by direct methods or formulas. Many authors have been trying to get optimum solution using metaheuristic algorithms.

Regarding Han Ping Fung's answer and questions: For (1), there seems to be products in the market that claim to do this - Densify (https://www.densify.com/) claims to use machine learning to optimize cost and other parameters in buying and deploying cloud based applications. Densify's Cloe product (https://www.densify.com/service/cloud-optimization-engine) appears to do (2) and (3). Using AI for Cloud operations has been called AIOps by some - and this web page lists a number of products that are said to use machine learning to optimize cloud operations (https://wikibon.com/building-ai-optimization-cloud-computing-infrastructure/).

There is no doubt that AI can change cloud computing by enabling Cloud machines/applications learn, think, act, and react like human beings. This facilitates the learning and analysis of data-sets from the historical data while identifying patterns and make real-time decisions. This leads to process automation which will eradicate the possibility of human errors.

Again, AI via machine and deep learning drive predictive analytics in cloud domain. This leads to an endless possibility with respect to Cyber-physical systems, HPC, Osmotic computing, Fog, Edge, and other forms of Cloud based computing.