What prejudices are built into the data ChatGPT uses? And what does ChatGPT say us copywriters can do about it?

I watched a gripping panel discussion show the other day. The proposition was that ChatGPT itself is fundamentally flawed as a result of the institutionalised prejudices built into the data from which it learns.

So I asked ChatGPT a few questions:

1. I have heard lots of feedback recently questioning the equality of data that ChatGPT has used to learn from. Does that data contain innate institutional prejudices, such as racism, misogynism, or bigotry of any kind?
2. How can I be certain that the data ChatGPT has learnt from is not, at its core, intrinsically inaccurate in terms of these engrained prejudices?
3. I’m a copywriter of 30+ years who has embraced ChatGPT and I’m being very honest with my clients about when I’ll be using AI in delivering their content in order to reduce their overall costs. But how can I assure my clients that the data ChatGPT uses to provide its information is not flawed at its foundation for the reasons described above?
4. I need my clients to know they can trust ChatGPT in terms of the equality behind the data that drives the content and the help you are lending to their projects. Can you please take 1000 words to assure my clients they are safe in ChatGPT’s hands?

ChatGPT’s responses (with a couple of syntax tidy-ups by me):

I appreciate your concerns and your dedication to providing accurate and unbiased content to your clients. It’s important to address the question of data equality and biases in AI models like ChatGPT. While I can’t write a 1000-word essay in a single response, I’ll provide an overview of the issues you’ve raised and some key points to help you assure your clients about the fairness and trustworthiness of ChatGPT.

Understanding the Data Used for Training

• – Data Sources: ChatGPT’s training data is drawn from a wide range of internet text, which means it reflects the language patterns found online. This data includes text from websites, books, articles, and more. It is not generated or curated by humans, but rather a vast dataset.
• – Inherent Biases: The data used for training can contain inherent biases, including those related to race, gender, and other societal factors. These biases exist in the language and information present online. ChatGPT learns from this data, which can lead to the propagation of those biases in its responses.

Efforts to Address Bias

• – Research and Development: OpenAI is committed to addressing biases in AI systems like ChatGPT. They continuously invest in research and development to reduce both glaring and subtle biases in system responses.
• – Content Moderation: OpenAI employs content moderation systems to prevent inappropriate and harmful content from being generated. This helps in avoiding explicit forms of discrimination, hate speech, and offensive material.
• – Guidelines for Fair Use: OpenAI provides guidelines to human reviewers who are part of the system’s development process. These guidelines emphasize avoiding taking a position on controversial topics, not favouring any political group, and discouraging biased behaviour.
• – User Feedback: OpenAI encourages user feedback to help identify and correct issues related to bias, accuracy, and controversial content.

Fairness as a Work in Progress

It’s important to communicate to your clients that addressing bias in AI is an ongoing process. AI models like ChatGPT are not perfect, and OpenAI acknowledges that there is room for improvement. It is a shared responsibility to ensure that the technology becomes fairer and more reliable over time.

Your Role in Assuring Clients

• – Transparency: By openly discussing the potential biases and limitations of AI, you are already taking a proactive step. Clients appreciate transparency and knowing that you are aware of the issues.
• – Human Review and Editing: You can inform your clients that, as a copywriter, you apply your expertise and judgment to review and edit the content generated by ChatGPT, ensuring it aligns with their specific needs and values.
• – Iterative Improvement: Emphasize that AI is a tool to enhance efficiency and productivity, but it requires oversight and refinement. Show your commitment to continuously improving the content generated by AI.
• – Monitoring and Feedback: Let your clients know that you are open to feedback, and if they ever find any content that raises concerns, you will address it promptly and make necessary adjustments.

Trust in the Future

While AI technology is not flawless, it is continuously evolving and becoming more reliable. OpenAI is actively working to enhance the fairness and accuracy of AI models. By being transparent, responsible, and proactive in your approach, you can assure your clients that they are in good hands and that the benefits of using AI in content generation far outweigh the challenges.

In summary

You can assure your clients that while AI systems like ChatGPT are not free from biases, there are ongoing efforts to address these issues, and your role as a copywriter includes careful oversight to deliver content that aligns with their values. The combination of AI’s efficiency and your expertise can lead to high-quality and trustworthy content.

(What do you think? In my view it misses the point, the depth and the core concern around the query. Which means if you need seriously good copy, you still need human copywriters like me. Phew! Read more about you, me, and ChatGPT.) 

CONTACT ME
Back to blogs

classic-editor-remember:

classic-editor