Artificial Intelligence is taking the world by storm: from defeating the world’s Go Champion to enabling driverless cars, discovering new drugs, predicting crop yields, flying drones to detecting fraud, curating news feeds on Facebook to suggesting music on Spotify – the possibilities are endless. IDC predicts that global spending on AI will reach $57.6 billion by 2021. However, on the flip side, discussions around AI ethics are becoming more and more widespread amongst AI researchers.
The Discussion on Ethics
Although AI looks good on the surface, it gets increasingly rigid as it expands across larger systems. The possibility of creating thinking machines raises a host of ethical issues: issues to ensure these machines do not harm humans (or take away their jobs) and issues around the moral status of the machines themselves. Regardless of its applications, there is an impending need to create regulatory frameworks to ensure AI complements humans and does not control them.
- One of the biggest fears of AI is that it is going to make humans obsolete. Isn’t it scary to imagine AI making complex decisions about how we live our lives? How do we stay in control of such intelligent systems? Pulling the plug will not be an option because these systems would have learned to anticipate and defend.
- Another fear is about how AI can be acutely biased, as it makes decisions based on only a chunk of data, and not an entire mass. IBM’s Watson for Oncology for instance provided cancer diagnosis treatment only for Americans and could not be applied to other races of the world.
- With voice-recognition systems and robots taking over our lives, there is a likelihood that AI can be used to manipulate people. How the next generation uses or interacts with AI systems could affect their development and the future of humanity.
- Since AI machines perform depending on the information that is fed to them, it is easy to manipulate them in order to benefit just a section of the society. According to an MIT study, three of the latest gender-recognition AIs could recognize a person’s gender from a photograph 99% of the time, but only for whites. For non-white people, the accuracy dropped to 35%.
- Granting AI machines themselves are not biased, because they are created by humans who can be biased and judgmental, the output can be prejudiced. For instance, while predicting future criminals, the COMPAS AI system tends to racially profile people based on the data it is fed.
- The capabilities of AI can also be used to misused by people with malicious intent. If no security measures are taken to keep AI safe from adversaries, cybersecurity wars can become highly prevalent – after all, dealing with a system that is faster and more capable than us isn’t straightforward.
What you can do
For all the things AI can do, it is imperative that we only use the advanced intelligence of AI for the good, and not for the bad. Humans need to guide AI on an ethical basis throughout its lifecycle. While AI can provide us with the information, we need for better ethical decision making – the onus is on us to make those decisions. In order to ensure ethical AI, here’s what you can do:
- Develop AI across different user groups: Since humans possess inherent social, economic and cultural biases, it is important to develop AI across diverse user groups. This can help eliminate biases and build capabilities from relevant and trusted data sources that embrace a diversity of culture, knowledge, opinions, skills and interactions.
- Ensure a wide data set: It is advisable to be aware of the biases from the beginning — or, at the very least, address them as soon as they begin to influence data. By taking in data from diverse systems, you can ensure that AI is being trained on the right information while simultaneously making it aware of the limitations of a given data set.
- Incorporate responsible practices: Since AI performs tasks based on the information fed to it and the things it learns on the job, it incorporates human social norms into its work. Therefore, when AI is developed with responsible business and social practices in mind, it can consistently and ethically deliver outcomes.
- Recruit people who understand AI: Employees join and quit. Teams change. Technology advances. Change is the only constant. Hence, focusing on efforts to develop and grow a diverse talent pool who can apply strict guidelines can ensure AI performs tasks alongside human employees across industries and geographies.
- Filter out human judgement:Since AI is only as good as the data it is given, it is important to not feed in human judgment but only into the results it returns. Developing effective mechanisms to filter out biases, as well as any negative sentiment in the data that AI learns from can ensure the technology does not provide biased outcomes.
Leveraging AI for the Better
In spite of the myriad benefits that AI offers, the fears around it are also widespread – from unemployment to bias, security concerns to malicious intent. While AI has the potential to improve the way we live our lives, considering the associated risks is vital. Developing strategies around how the benefits of AI can be spread to all of humanity is critical to ensure AI ethics. Artificial intelligence has vast potential; its ethical implementation is entirely our responsibility.