Artificial Intelligence or AI is the latest technology developed to make things easier for humans. The presence of AI technology is able to revolutionize all aspects of life, including work. AI technology provides an opportunity for computers to learn big data so that they can carry out complex tasks. Artificial Intelligence is also effective in optimizing work done independently or in groups.
Artificial Intelligence is not only applied in the business world. This technology is also used in mobile applications, games, and desktops. Operating systems, such as Windows, iOS, and Android also involve AI technology to maximize device performance. In the future, AI technology is expected to be as important as the internet and electricity.
Why AI Needs to be Implemented?
 |
| How is AI being Implemented Today? |
The reason why the implementation of Artificial Intelligence is very important today is because the world is moving very fast, and manual work is considered less effective because it takes a lot of time. With Artificial Intelligence, the work can be completed in a short time. Moreover, AI is also capable of doing repetitive heavy tasks, and even the work can be relied on and has minimal errors.
Moreover, nowadays, the workload continues to increase over time. It would be very easy if repetitive work could be done automatically with the help of machines. Human energy can be diverted to other more complicated and complex tasks such as decision making or business development.
AI which has a wider scope is currently starting to narrow down to Machine Learning because it has advantages over other AI models. This is related to the availability of big data that is easily accessible. Armed with data, the Machine Learning model can be trained to produce a model with better accuracy than the non-Machine Learning AI model.
The following image is a characteristic of the current application of the AI Machine Learning model. The learning model used in the image is Supervised Learning. Other learning models are Unsupervised Learning and Reinforced Learning. Currently, there are many sites that provide models that have been trained, known as pre-trained models. These sites are known as hubs that can be accessed and utilized by the public.
The first type is that users only use pre-trained from the hub. The form is some that use the Application Programming Interface (API) and some whose models can be downloaded for use. One of the advantages of the API is that users can only utilize the intelligence of the model but cannot download the model. While most hubs allow users to download their pre-trained models.
This first type of model, for example, is translating language from Indonesian to English as follows. When a sentence is input, it is translated into English. With API we don't need to go to a translation provider site, for example Google Translate or the currently famous ChatGPT.
The second type of ML model that is also widely used is to retrain the pre-trained model with typical data because it will be used for a particular environment. This mechanism is known as Transfer Learning. When training, the weights that have been trained previously with large data are held so that they do not change, which is known as freeze.
An example is the use of a chat bot with a model from a hub called hugging face. Here the trained model has the ability to answer questions from a series of paragraphs. With AI, different questions about the same thing will produce the same answer based on certain information.
Here specific data such as registration information data such as requirements, when to start lectures, tuition fees, and so on can be given to the system, for example in Comma Separated Value (CSV) format.
If asked a certain thing, the system will answer according to the CSV data. This reduces the burden on public relations in answering questions from consumers or other interested parties.
The third model is a model that is trained with its own data without utilizing a pre-trained model. This usually happens because there is no pre-trained model available. Or the model is not a standard model but a newly created model. Usually intended to improve the performance of the proposed model. Well, here the challenge is that in addition to requiring large data, it also requires hardware resources with strong GPU and RAM.
An example is a model that trains a text category machine. Here a machine, for example SVM is trained with training data that already has labels in the form of categories such as ideology, politics, economy, socio-culture, defense and security. After being trained, an AI is able to categorize a text into one of the five categories.
Maybe our brains are better than the AI machine, but for the millions of scripts that appear on the internet and we have to categorize them manually, of course it is not efficient. This is where one of the roles of AI (think/do rationally) is which is the AI quadrant according to Russell's AI book. Ok, hopefully this post helps.
What is an example of an AI application?
Considering the level of competition in the desktop and mobile application market is quite tight, companies are implementing AI technology to boost customer user experience and gain profits. Interestingly, without realizing it, quite a few frequently used applications have utilized AI technology. Here are some examples:
1. Email
Email providers, such as Google, have gradually introduced AI features. Starting from quick message replies, intelligent email categorization, spam deletion, to eliminating the complexity of unproductive tasks.
2. Face Unlock
Face Unlock technology is a feature that users can use to unlock their smartphones. This success has encouraged Microsoft Windows to follow this trend by introducing Windows Hello to unlock their computers.
3. Google Navigation
With the help of Machine Learning, Google Maps learns your daily work journey. This application also provides information on traffic, weather conditions, and identification of delays in transportation systems, such as trains, flights, and buses.
5. Entertainment and Social
Entertainment and social media applications, such as Netflix, Amazon Prime, Twitter, Instagram, and Facebook have added new features to their platforms to provide a better and smarter experience.
6. Banking and Finance
AI and Machine Learning technology play an important role in the banking industry. The presence of this technology provides a sense of security and comfort for customers. Machine Learning effectively prevents fraud by monitoring spending habits regularly and considering factors such as the interval between transactions, location, value, and others. AI technology helps determine whether transactions are legitimate or fraudulent.
Equipped with facial and voice recognition features, banks can protect payments made by customers. AI technology also optimizes the use of digital banking applications. This allows you to make the entire banking process fast, safe, and paperless.
What are some opportunities of AI?
While the implementation of AI in Indonesia has not been fully realized, its potential is quite promising, especially in addressing health care inequality and revolutionizing various industries. Health care inequality is a significant challenge in Indonesia, but AI can be a game changer by expanding its implementation to remote areas. By studying large amounts of historical data on doctor care and patient data, AI can replicate the advice of many doctors at a lower cost, enabling accurate diagnoses and improving standards of care.
In addition, AI can enable scientists and engineers to make breakthroughs in various fields. Industries such as agriculture, manufacturing, and mining can benefit from AI capabilities, revolutionizing practices and optimizing operations to increase efficiency and productivity.
In agriculture, AI applications in crop monitoring, disease detection, and crop yield optimization can revolutionize agricultural practices. By analyzing satellite imagery, weather data, and soil conditions, AI can help farmers optimize their approaches, especially in the face of climate change. In addition, the financial services sector also has the opportunity to use AI to address low financial literacy in Indonesia. By analyzing diverse financial needs and behaviors, AI-powered solutions can customize financial products and services, expand access to financial resources, and enable more accurate credit assessments for underserved communities.
What are the challenges of implementing AI?
There are many obstacles or challenges in the development and implementation of AI in business. In the AI implementation process, this needs to be considered in detail. These challenges include the following;
1. Lack of support from stakeholders
There are many possible causes of lack of stakeholder support. For example, the team does not want to take risks, whether the technology adopted is capable and can be applied. Doubts arise from the team who do not understand the benefits of AI. They doubt why AI is important in helping business processes. In addition, the lack of knowledge of AI including in terms of implementation and unclear expectations in its implementation can cause doubts from stakeholders in implementing AI in the future.
Not only that, after getting support, the next problem is the lack of clarity in its execution. This AI execution includes; where and the starting point of execution, what use cases are used, and what AI strategies will be implemented, the use of in-house solutions or vendors, the formation of data teams and knowledge of the company's current state. Should the company invest first? Should they invest in dataflow, and how will the infrastructure be run? These problems need to be reviewed in the implementation process developed by the next system.
b. Not having organized, consistent, and appropriate data
This problem includes various components, including the following: important data may not be available, the company may not store the data needed and must be available for AI development, the data owned is still in paper form and has not been digitized, and the format system is unstructured; even the main problem may be data that has not been collected at all. The existing data may not be clean data, or it may still be noisy data. Noisy data can cause instrumentation errors, resource errors, and errors that cause machine learning to be incompatible.
When viewed from the quality of the data, the data flow in noisy data is still scattered everywhere so that it is not centralized. Bureaucracy in data documentation is still poor and unstructured, so ETL data is unclear. If you want to build an AI architecture, this problem is very complex and needs to be fixed from the start of execution.
c. Data Privacy and Security
The use of data is currently highly regulated by regulations, including parts of privacy and security. Do not let the data used by the AI Application to be studied violate the ITE Law, including privacy laws. Pay attention to the legality and copyright of the data so that it does not violate the rules and policies related to user privacy.
Then, the period of time for which the data will be stored, including the storage of data from the company must be considered. This regulation must be clear and well-structured so that there are no wrong policies in the future that violate data misuse.
Conclusion
Artificial Intelligence itself is created by humans through complex programming algorithms, and there are two categories of AI that need to be known, namely Weak AI and Strong AI.
- Weak AI is specifically designed to do a specific job, where AI systems in this category include video games to virtual assistants like Alexa on Amazon and Siri on Apple.
- While Strong AI is more complex, where Strong AI runs commands or tasks as similar as possible to humans. This system is more complicated and is specifically programmed to handle situations where they may be asked to solve problems. Strong AI can also be found in types of applications such as self-driving cars or hospital operating rooms.
The application of Artificial Intelligence is indeed very broad now. Various sectors of life have begun to adopt AI technology to streamline work processes. Are you also interested in starting to implement AI in your company or business?
