Artificial Intelligence - risks and benefits | Sunday Observer

Artificial Intelligence - risks and benefits

Artificial Intelligence (AI) is not a new topic. John McCarthy, a computer scientist at Stanford University, introduced the subject in 1956. The idea behind it was to make a machine operate like a human being, as far as possible. Artificial Intelligence will supposedly make our lives comfortable. Scientists are predicting massive dependency on AI by humanity, in the coming years.

Today, AI provides society with many benefits in the field of economics and law, ( in processes such as verification, validity, security and control). For instance, a minor nuisance such as laptop crashes, or being “hacked” can be rectified. An AI system which controls your automated trading system, bank account, emails, makes it a high priority to correct anomalies as soon as possible.

The danger is In the long term - what will happen if AI succeeds and an AI system becomes better than humans at all cognitive tasks? AI could potentially undergo self-improvement, triggering an intelligence explosion leaving human intellect far behind.

Breakthrough technologies, such a superintelligence might help to eliminate war, civil conflicts, disease, poverty, and interact in many ways in the human life cycle.

Risk drivers in AI

AI, unbeknown to us, is slowly penetrating our homes and lives, in every part of the world. Some forecasts indicate that AI could also take over crucial parts of most businesses such as marketing, customer service and sales in the next five years.

Where AI is concerned, there are five areas that can be identified as risky. They are - data risk, autonomous weapons, social media manipulation, algorithms and human-machine interactions.

Data risk

Big data is available everywhere. We already know that the amount of available data has been growing at an exponential pace. This will connect to several devices to collect the data. The processes such as entering, sorting, linking, and using data have become difficult since a large amount of unstructured data is being used from the web, social media, mobile devices, sensors, and the Internet of Things.

AI is currently good at finding patterns and relationships within big datasets. Crunching the numbers can effectively identify and find subtle patterns in given data, but it is unable to tell us which of those correlations are meaningful. Computing power and statistical algorithms patterns could be analysed with the provided data. However, sometimes, these patterns are not coherent.

Again, spurious patterns and correlation could easily outnumber the meaningful ones; therefore, the final decision made of the issue is in the risk.

Autonomous weapons

AI programs may be dangerous in autonomous weapons since it can be programmed to destroy something or somebody. The main concern is the dangers autonomous weapons might pose in the hands of an individual or government that doesn’t value human life.

Social Media manipulation

Social media, since it has autonomous-powered algorithms, is ultra-effective for target marketing. It discloses who we are, what we like, what we think. By conducting an advertising campaign to individuals identified through algorithms and personal data, AI can target them directly and spread whatever information they want, whether right or wrong, wanted or unwanted.

Interaction conflicts

The interaction between people and machines is identified as another risk area. The possible conflicts are in automated transportation, manufacturing automation etc., for instance. Autonomous vehicles may cause accidents and injuries on the road due to the misbehaviour of other drivers. Even heavy machinery is unable to recognise when systems should be overruled.

Risk Management Framework (RMF)

Managing the risk of AI, requires firms to study business processes and models.

Such processes are not designed to eliminate all AI-related risks, but develop methods and tools to make sure that such risks can be effectively identified and managed within appropriate limits.

Identification

The first step is identifying which risks could have a direct impact on the organisation’s business strategy. This stage includes monitoring the internal/external operating and regulatory environments to identify changes to the natural risk landscape and ensure the organisation’s purpose. It is also important to have periodic reassessments to determine whether the risk profile AI has changed from the date of its introduction.

Access

A risk assessment and risk level identification should be completed at this stage. It is also necessary to agree with the firm’s management before the development of each risk assessment matrix. The process should give careful consideration to the key risk drivers in the industry.

Control

Controlling processes will also need to be more efficient. For the better control of AI, organisations need regular and frequent testing and monitoring of all solutions, even in the development stage of the solutions. (You can use traditional technology solutions for monitoring purposes). A risk-based approach should be used to determine the appropriate level of control for each case and comply with the organisation’s risk assessment framework.

Monitoring includes all legal and regulatory developments that need change in the design of the models, and also, external events that would indirectly feed into the data consumed by the model and influence the outcomes.

The need for white-collar workers in professional, managerial, or administrative work, will be reduced with AI application. Industrial robots predominantly manufacture most physical goods today. The international federation of robotics has said that the worldwide robot density is currently 74 robot units per 10,000 employees.

The automation of tasks from blue-collar workers in manufacturing sites is still similar to the automation for document-centric tasks of white-collar workers in enterprise back offices. For instance, according to the IBM report of 2018, an average Fortune 500 company with around 50,000 employees spends USD 5 million annually on the creation of meeting minutes. However, recording the meeting and document preparation is not a big task with AI, since it has speech recognition and speech to text transcription.

Forecasting the future of AI, everyone will connect to the Internet to obtain various types of information, and AI systems will collect and analyse information on their needs. We will probably be more reliant on these technologies than smartphones.

AI will be widely used for analysing the information collected from wearable devices, which are stored in the cloud. AI will enable wearable devices to be of real assistance in our day-to-day lives, offering numerous benefits to our office and personnel tasks. Wearable devices will have a growing market in the coming years.

AI is already able to defeat human beings. Today, however, machines are able to decide for themselves what it will learn, how it works, etc. thereby making it challenging to divide between the human and machine domains. We must think about what machines can do today, realise that machines surpass human abilities even without the assistance of AI.

For instance, cars are running faster than humans; computers provide speed and accuracy better than a human.

There are an ‘AI booms’, but there is a transformation phenomenon. The development of the Internet and the deep learning incorporates the outcomes of high-volume machine learning and brain science. Most of the technical barriers have been overcome; therefore, we can have positive expectations for future developments. Finally, I can say that AI will surpass human intelligence.

The writer is Head of Consultancy, National Institute of Business Management.

Comments