The Covid-19 pandemic has altered the traditional work models for both employers and employees resulting in a hybrid work environment. This has been achieved by more and better automation and AI-based initiatives. Employers have recognized the opportunity to automate many jobs to a large degree to allowing them to work with a part-time or hybrid workforce. Many of jobs are at risk from AI. There are various factors that will help employers understand how it will impact their employees and business some of which are discussed below:
Singapore has formulated an AI governance framework. It lays a great foundation for understanding the issues related to AI’s impact on jobs. Following are the two main factors in this framework:
- The probability of harm caused by AI
- The severity of harm caused by AI
Since we have considered the potential harm of AI on a job, we will be able to better comprehend the implications of how it will impact roles in various areas. For a better understanding, let’s consider three situations where humans are working with AI:
- Human-over-the-loop: Humans may intervene in AI-based decisions, as and when required.
- Human-out-of-the-loop: AI makes decisions without human interaction.
- Human-in-the-loop: The decisions require human interaction.
We will now discuss how each of these situations will affect jobs.
Human-over-the-loop (Quadrants 1 and 4)
This AI model accounts for human interference when they don’t agree with the decision or AI has failed. Human interaction is usually required when the probability or severity of harm is high. Nevertheless, if humans are busy elsewhere, the AI model will continue without human interaction. In such situations, AI may replace humans or supplement their jobs.
The first example of this situation are traffic prediction systems. AI will recommend the shortest route to the destination. However, humans can disregard that decision when they want to intervene.
Most security systems protected confidential data by various security protocols such as a firewall and encryption. Hackers are unlikely to breach both your firewall and decrypt your data successfully. Nevertheless, if they manage to do that, the severity of the attack will be high. In such an instance, humans will be expected to concentrate at the task at hand. (See If Microsoft Can Be Hacked, What About Your Company? How AI Is Transforming Cybersecurity) Most companies need to define a framework to handle AI-based cybersecurity solutions for such attacks.
In compliance audits, the probability of harm is high since the system will most likely not be flawless. Having said that, the severity of harm is lower since the business may be given an option to correct the issue or simply fined for it. It is important to be cognizant of the fact that some compliance audits are more important than others. Based on the importance, it is best for humans to ascertain whether (any) intervention is required or not.
Human-out-of-the-loop (Quadrant 3)
Human intervention is usually not required when both the probability and severity of harm is low. In such instances, the AI runs independently without human supervision. These are the areas where there is risk of humans losing their jobs is the highest.
Recommendation engines are a great example of such systems that require little or no human interaction. Many e-commerce sites recommend products to consumers are most likely to buy. Furthermore, companies such as Spotify and YouTube will recommend you songs or podcasts that you would like to listen to next. For such recommendation engines, both the probability and severity of harm is low eliminating the need of human intervention. The worse possible outcome is being recommended something you won’t like.
Another example of such systems is translators. Except for situations with legal implications, both the probability and severity of damage is low. AI-based translation systems are improving at a rapid rate where they will soon replace humans. AI is not only learning how to translate basic language but also the meaning of slang and localized words. Perhaps the most fascinating is that it is also learning how to read body language, facial expressions, and tone of voice.
Human-in-the-loop (Quadrant 2)
The main function of this type of AI systems is to assist humans in making decisions. Since the probability and severity of harm are high in such instances, a human is required to be a part of the decision. The AI-system cannot run independently and makes recommendations to humans. If the human intervention isn’t available, no decisions will be made.
When AI systems are used for medical diagnosis and treatment of severe issues, they usually fall in quadrant 2. The AI system will most likely assist the doctor in determining the patient’s issue. Moreover, it may also recommend a course of action by reviewing the results of patients with similar diagnosis. The AI-system will have the results of more patients than the doctor and will help them make better and more informed decisions.
AI-systems can help a company looking to expand identify potential targets for M&A and to undertake their due diligence. However, since the probability and severity of damage are high, the AI-system will help identify potential opportunities that may be hard for humans to see. Lastly, AI offers exceptional predictive models for potential acquisitions. You will, however, need a human to participate in this decision.
Last but not the least, AI-systems are helpful in designing products since they can help foresee and design options not envisioned by the humans. In case you want to design a drone, you will provide the AI-system specific parameters such as the cost, maximum altitude, the amount of weight it can carry, and its range after which the system will provide you design ideas. In conclusion, AI is better at humans at certain tasks. If you perform a job that can be completely automated, it is at risk. Still, for most jobs AI will perform the tasks it is best suited for an allow human intervention to make optimal decisions. Both the employees and employers garner benefits in case where AI is augmenting humans’ decisions.
This is a summary of the original Forbes article. To read it, click here.