AIaaS: A new trend for AI? By Aditya Abeysinghe
Today, the use of cloud services has changed the way people build, process and store applications. Many businesses have deployed applications to the cloud and use different models in the cloud to host and process applications. Deploying of these apps to the cloud is possible with cloud service providers providing services required. Cloud-hosted Artificial Intelligence applications are a new trend in the cloud to build models and obtain services.
What is IaaS?
IaaS, Infrastructure as a Service, is a service for managing, monitoring, and accessing resources in the cloud. Resources such as servers, storage, etc. are purchased on-demand using IaaS. IaaS vendors provide a graphical user interface from which users can manage all resources and monitor their usage. The benefit of IaaS over on-premise is that the user does not need additional hardware and network to manage resources. Instead, only costs for all resources purchased need to be paid by the user. Also, scalability, disaster recovery and hardware provisioning are managed by the cloud vendor.
What is AIaaS?
AIaaS, Artificial Intelligence as a Service, is a service of Infrastructure as a Service which provides AI, Artificial Intelligence, services in the cloud. Like with other IaaS services, AIaaS tools are provided by the vendor and the user pays for the usage of tools and hardware. Most AIaaS tools can be integrated with other services in the cloud. Therefore, the user can easily use other services to host datasets, to secure their AI systems etc.
Benefits
Less costs in investing for software and hardware to process AI models is one of the advantages of AIaaS. Most AI models need large computation power and storage to process datasets during building and deployment. Most small-scale businesses, users who train models for non-business uses etc. often do not have the ability to buy hardware that could process datasets with the required power. For these users using AIaaS in the cloud is a low cost and reliable option to train these models.
Another advantage of using AIaaS is the ability to easily add or remove resources when required for computing and storage. With the cloud, users do not have to pay for resources unnecessarily and they can easily scale the resources. Since most services in the cloud are managed by the cloud service provider, most of the necessary services can easily be integrated with the scaled services. Also, by scaling into multiple clouds, disaster recovery and availability of the AIaaS can be ensured.
Challenges
The security of applications in the cloud is a challenge when compared to security of applications when they are hosted on-premise. Most AI models require datasets which need to be downloaded, then hosted in the cloud to train models and then store data used by models during processing. Security of data and applications must be secured since data is stored in remote locations where users do not know who has access. Also, when data is stored in the cloud data can be vulnerable to network attacks.
With the cloud, there can be hidden costs when services are being used for a long run. Businesses need to often train their users to use AI services in the cloud or use people who are specialized in cloud services. There could be also costs when applications used currently need to be adjusted to be used in the cloud. Different cloud vendors provide APIs, Application Program Interfaces, to integrate services of customers with the cloud. Therefore, costs for managing services to be usable in the cloud need also be added.
Image Courtesy: https://bernardmarr.com/
