Enterprise AI Ethics – What Makes the Problem Complicated?
Artificial Intelligence had always been in the boundaries of research and academic communities until recent times. Now with advancements in technology today, AI plays an important role in all enterprise across various verticals such as Healthcare, Insurance, Retail, Manufacturing, and Consumer Applications.
This ability to embed Artificial Intelligence methodologies into the core and data governance strategies of the organization is called Enterprise AI. Enterprise AI stimulates the cognitive functions of the human mind like learning, reasoning, perception, planning, problem-solving, and self-correction with computer systems. It comprises of business applications such as expert systems, speech, and image recognition, etc.
Enterprise AI has taken up a varied range of responsibilities in various industries across all domains.
- Rolls Royce analyzes data from internet-connected sensors to spot any wear and tear in its plane engines to carry out predictive maintenance.
- Amazon uses facial recognition to recognize what shoppers buy in the Amazon cashless store.
- Walmart uses facial expression recognition and sentiment analysis to find out the satisfaction level of customers at Walmart stores.
- Citigroup uses AI to spot fraudulent transactions and errors in payments.
- Microsoft PowerPoint uses Image processing AI to identify, tag, and add a description to the images used in the presentation to ensure that PowerPoint is accessible by visually impaired audiences.
- Amazon and Netflix use the power of AI to show future recommendations based on the analysis of customer activities.
Such Enterprise Applications of AI has helped these companies grow their reach and enhanced customer opportunity conversion.
Companies are increasingly using AI and ML make decisions, such as the university admissions, allocation of jobs or loans, that directly or indirectly affect people’s lives. Algorithms are also used to recommend a movie to watch, a person to date, or an apartment to rent.
The AI Black Box
When talking to business customers, these decision-making algorithms and models are a complete black box. For example, let us imagine an ML model for hiring is used for recommending the top 20 candidates from 150 applicants for a job post. Before trusting the model’s recommendation, the recruiter wants to check the results. If the recruiter were checking human work, they would look for summary or clues like underlines, circles, pluses, or minuses around salient elements. This is different for ML models. Depending on the depth of the algorithm’s neural network, there would be limited explanation and transparency.
The black box problem of artificial intelligence has grown with modern, more powerful machine learning solutions and complex models. AI models can outperform humans in complex tasks such as classification of images, transcription of speech, or translations from one language to another. These models can also perform a lot of complex tasks in less time without human intervention – leading to automation of several industrial processes. But, the more sophisticated the model, the lower its explainability level.
In some cases, the black box issue does not matter because the users have no choice other than leveraging the machine’s intelligence blindly to save time, effort, and human intervention. For example, manually translating a lot of text from Chinese to English cannot be done by an explainable and straightforward model.
The Coming Age of Augmented Intelligence
The practical adoption of AI systems in enterprises that are making a move to Augmented Intelligence includes empowering not just AI engineers but also the System Integration (SI) engineers and business stakeholders as well.
The Present AI systems, which primarily involve an AI engineer as the only human being in the loop, leave out these important constituencies. Enterprise AI should include the following main aspects of transparency.
- Accessible AI – Business stakeholders should be able to understand and ask questions without going via AI engineers.
- Explainable AI – AI models should be simpler and be accompanied with some explanations which can be understood by business stakeholders and users.
- Interactive AI – The business stakeholders should be able to use the AI by making it interactive and get the output without the help of AI engineer.
- Tunable AI – The AI system should provide easy interaction with users.
What about AI Ethics?
An AI system in the enterprise grows, so does the question about AI ethics.
The black box problem of AI forces us to think, how can we trust the decisions it makes? The more an AI system leverages machine learning and neural networks to crunch immense amounts of data, the less likely it is for a human to understand how the AI has arrived at its conclusion.
A lot of major tech companies, aware of these growing concerns, have already started taking steps to mitigate the AI ethics issue. For example, Microsoft has had FATE (Fairness, Accountability, Transparency, and Ethics in AI) since 2017, a committee that aims to remove discrimination in algorithms, enhance fairness in outcomes.
Ultimately, human decision-making can only be explained to some degree. It is the same for complicated algorithms. However, it is the responsibility of the companies and software developers to improve on AI model transparency to build further trust in intelligent software.
The onus is on the major tech companies embedding Enterprise AI into their core systems to build transparent and secure AI systems and grow towards a trusted AI world.