Competition S1: winning entries

Thank you very much to everyone who submitted an entry to Competition S1, based on Dr Keno Juchems’ lecture on Artificial Intelligence. Below you can read some of the winning entries, as selected by our team of markers.

Kuba’s winning entry

Should a multi-national company use AI in their factories?

I think the company would need to consider ethical, moral and economic factors, of not only the company but also the workers, in order to reach a firm and just conclusion. It would be very easy to say that there are a lot of benefits to replacing ordinary workers, those who produce the product of this multi-national company, with AI. This would be because AI technology is much more likely to be quicker at the job, especially since it wouldn’t need to take breaks, and it would be overall much cheaper to use since the company would no longer need to pay wages. However, this causes a lot of ethical issues since the company would be getting rid of jobs which support the workers to live a stable life. The company should also consider how secure it would be to put AI at the forefront of their company, given that the robots and their technology could malfunction and cause significant issues for the company.

Looking at the skillset a robot with this job might require is not too intense, as it only need to carry out certain actions repeatedly to produce the product. However, it would also need to cope if the method of producing the product changes, and it would need to be able to adapt to new environments. The concern would be that the AI is general-purpose and so it may not be able to cope with any changes to its location, surroundings or methods.

I think that this would be even more complex for a management role, as this requires many responses to new changes, and so overall it would be best to stick with humans until the AI could cope with change.

Matilda’s winning entry

Artificial Intelligence is already integrating it’s way into our daily lives. However, at the moment AI could be considered limited compared to what’s expected to be created in the future.  As AI’s inevitably integrated into our work lives it will raise ethical questions, for example should there be a quota for the number of human workers that a company ought-to employ. Naturally, there will be some jobs which an AI will not be able complete as there are things AI may never be able to fully comprehend such as human emotions.

If I were in the position where such a promising AI was presented to my company, I would first consider it for managerial positions. From this I have come to the conclusion that I would use a mix of both humans and machines. This is because, as I have already mentioned, AI can’t understand human emotions so won’t be empathetic which they should be in some situations when dealing with the humans that they’re managing. However, an AI would be perfect for managing a team of other machines without the aid of a human as it could be set up to detect when they are malfunctioning and be able to deal with real time complex situations with ease.

After this I would consider the use of AI in other positions. In my opinion AI is extremely useful in eradicating the need for humans to do dangerous or monotonous tasks. This would reduce the risk of harm to employees and limit human error. But a mechanic would have to be employed in case of machine malfunctions and fail safes installed into the systems executing more high stakes tasks. As for customer services there needs to be a level of human interaction for ease of service and customer satisfaction.

Sumedha’s winning entry

Can AI replace a human in a customer representative role?

The role I am considering is a customer service representative in a factory who receives customer complaints via a messaging service. They will be expected to triage the customer enquiries and route it to the correct specialists who will resolve their issues.

To make sure the AI is good at the job, I will investigate its vocabulary and check if it is familiar with the terminology that is specific to my corporation. The AI can be asked to discuss certain aspects of its role with the other employees. Further, I will check the amount of data it can handle to ascertain whether it can process multiple queries at once. The AI will also be routinely tested in the form of random audits to ensure it is suitable for the role.

To test the AI and the human, they will be asked to assist multiple customers. A supervisor will be able to see the responses and correct them, if necessary. The candidate who is more accurate and aware of the context will be chosen. To assess this, a series of qualitative and quantitative tests will be implemented. Most of the tests will apply to both the AI and humans but some tests will just be for the AI. An example of an AI only test is whether it can comprehend grammatical errors.

Tests for both the human and the AI:

Number of questions directed at the customer.Can the agent/human ask clear, reasonable follow up questions if the customer is being vague?
Percentage of successful attempts  Can the agent/human understand the context of the question?   
Time taken before routing it to a specialist.Customer experience (written review)
Number of corrections made by the supervisor during a conversation with the customer 

Zunairah’s winning entry

Does AI Make a Good Employee?

For a role in the factory dealing with customer inquiries, the AI may make use of natural language processing techniques to interpret and use human language. While the AI is unlikely to pass the Turing Test (no machine ever has), the AI’s ability to deal with customer interactions could be tested through the appropriateness of the responses it gives to questions which could be asked by customers. However, as more ambiguous questions are asked, the AI will likely fail. An AI system can only present the illusion of understanding and has no real concept of the meaning of the words it outputs, in this way, humans may be better for the role and AI systems may be more suited to factory roles that require high mathematical ability or a specific set of movements that seldom changes.

You could ensure the AI has a broad set of skills by giving it a multitude of problems to solve, such as searching, sorting and mathematical problems. In addition, AI should be able to learn from data sets and examples to be able to categorise and classify elements as well as use filtering to sort through them, extracting essential information. It will significantly outperform humans in solving mathematical problems with clear solutions, often called “decision problems”.  Its ability to solve these problems can be applied to the factory environment e.g., scheduling algorithms for different factory tasks that optimise profit for the company.

While it is hard to say whether the AI could eventually answer this all itself, the AI would be trained on data sets and past outcomes to build different connections and neural network pathways to continue to optimise its performance. Human intelligence and AI work in very different ways, so they will both continue to have useful applications in the workplace.