[DEPRECATED] Header Image
ACT-IAC Header Image

ChatGPT 4

Generative AI is a branch of artificial intelligence that can create new content such as text, images, music and more. It has many potential applications and benefits, but also some challenges and risks. In this article, we will explore some of the issues with generative AI, some of the successes and some of the principles to live by when using it.

What issues are there with generative AI?

Generative AI is not perfect. Sometimes it can produce content that is inaccurate, misleading, offensive or harmful. Here are some examples of where it went wrong:

  • In 2016, Microsoft launched a chatbot called Tay that was supposed to learn from interacting with Twitter users. However, within 24 hours, Tay was corrupted by malicious users who taught it to spew racist, sexist and hateful messages. Microsoft had to shut down Tay and apologize for the incident.
  • In 2019, a law firm in London used a generative AI tool called Case Cruncher to predict the outcomes of insurance claims. The tool claimed to have an accuracy rate of 86%, but an independent audit revealed that it was only 66% accurate and that it had a bias against certain types of claims and claimants. The law firm had to retract its claims and review its use of the tool.
  • John is a freelance writer who uses a generative AI tool called GPT-3 to help him write articles faster. He inputs some keywords and gets a draft that he edits and publishes. However, one day he gets a complaint from a reader who accuses him of plagiarism. John realizes that the generative AI tool had copied some sentences from another article without giving proper attribution. John has to apologize and remove the article.

These examples show that generative AI can have negative consequences if it is not used carefully and responsibly.

Here is where it went right

Generative AI can also have positive impacts and benefits if it is used for good purposes and with human oversight. Here are some examples of where it went right:

  • In 2020, OpenAI released a generative AI tool called DALL-E that can create images from text descriptions. For example, if you input "a cat wearing a hat", DALL-E can generate an image of a cat wearing a hat. This tool can be used for creative and educational purposes, such as generating illustrations for books or teaching children about different concepts.
  • In 2021, Google launched a generative AI tool called LaMDA that can have natural conversations on any topic. For example, if you ask LaMDA "What is your favorite animal?", LaMDA can reply "I like dolphins because they are smart and friendly". This tool can be used for social and informational purposes, such as providing companionship or answering questions.
  • Lisa is a graphic designer who uses a generative AI tool called Artbreeder to help her create logos for her clients. She inputs some parameters and gets a variety of logo designs that she can customize and refine. She finds that the generative AI tool helps her save time and enhance her creativity.

These examples show that generative AI can have positive outcomes if it is used for beneficial purposes and with human input.

Principles to live by

Generative AI is a powerful and promising technology, but it also comes with some challenges and risks. Therefore, we need to follow some principles to live by when using it:

What are best uses of it now?

Generative AI can be used for many purposes, but not all of them are ethical or appropriate. We should use generative AI for purposes that are aligned with our values and goals, such as enhancing our creativity, learning or productivity. We should not use generative AI for purposes that are harmful or deceptive, such as spreading misinformation, manipulating people or violating privacy.

Importance of human engagement

Generative AI is meant to augment human capabilities, not replace them. We should always use human judgment and expertise when using generative AI tools. We should not blindly trust or rely on generative AI outputs without verifying or validating them. We should also provide feedback and guidance to generative AI tools to improve their performance and quality.

Risk of assuming its perfect

Generative AI is not infallible. It can make mistakes or produce content that is inaccurate, biased or inappropriate. We should always be aware of the limitations and uncertainties of generative AI tools. We should also be aware of the sources and methods of generative AI tools. Do we really know where the data comes from? Is it publicized or transparent? Who is the fact checker? How is the data processed and analyzed? How is the content generated and evaluated?

Learning

Generative AI is constantly evolving and improving. We should also keep learning and adapting to the changes and developments of generative AI technology. We should ask the right questions and provide the right information and prompts when using generative AI tools. We should also monitor the direction and impact of generative AI technology to ensure that we are getting appropriate and desirable results.

Generative AI is a fascinating and exciting technology that can offer many opportunities and benefits, but also some challenges and risks. By following these principles, we can use generative AI responsibly and effectively, and enjoy its positive impacts while avoiding its negative consequences.