Using ChatGPT and Generative AI Tools in Journalism - with Paul Bradshaw!

Summary

In this insightful webinar, Paul Bradshaw, a data journalist and course director, delves into the transformative potential of generative AI tools like ChatGPT for journalism. He explores various applications, from text summarization and ideation to coding assistance and bias identification. Bradshaw emphasizes the importance of effective prompt design, maintaining journalistic skepticism, and addressing inherent biases. The presentation offers practical guidance on integrating AI into journalistic workflows while underscoring the need for fact-checking and editorial oversight. As AI reshapes journalism education, Bradshaw envisions a shift towards news gathering, editing skills, and an increased focus on primary reporting.

Takeaways

• 😃 ChatGPT and similar AI tools can assist journalists in various tasks such as summarizing long documents, explaining systems, improving writing style, generating story ideas, and coding/programming.
• 😐 However, these tools deal in plausibility rather than facts, and can generate biased, racist, or false information if not used carefully.
• 🤔 Effective prompt design is crucial to getting useful outputs from AI tools. Provide clear context, specify audience/style, and iteratively refine prompts.
• 🔎 AI can help identify potential bias or differing perspectives in news stories, acting as a mirror to check one's own biases.
• 💻 Coding and troubleshooting are areas where AI excels, as it can generate functional code and offer fresh perspectives on coding issues.
• 📝 While AI can assist with text generation, it should not be relied upon for writing entire news stories, as it cannot reliably determine newsworthiness or factual accuracy.
• 🚫 Transparency is important when using AI in journalism. News organizations should have policies on disclosure and attribution of AI-assisted work.
• 🏫 Journalism education will need to adapt, shifting focus from storytelling techniques to news gathering, fact-checking, and prompt design skills.
• ⚠️ AI-generated search results are increasingly presenting misinformation as fact, highlighting the need for journalistic skepticism and verification.
• 🤖 While AI tools offer many potential benefits, human journalists remain essential for ensuring accuracy, ethics, and high-quality reporting.

Q & A

What are some key advantages of using ChatGPT and generative AI tools for journalism according to the speaker?

The speaker highlights several advantages: 1) Summaries - AI can effectively summarize long documents and reports. 2) Understanding systems - It can explain complex systems and processes, aiding story ideation. 3) Text correction and subbing - It can help improve writing style and catch errors. 4) Source ideation - AI can suggest diverse sources for stories. 5) Advanced searches - It can generate complex search queries. 6) Coding and programming - AI excels at generating, explaining, and troubleshooting code.

How does the speaker recommend addressing the issue of bias in generative AI tools?

The speaker emphasizes the need to be aware of the built-in biases in AI language models and suggests several strategies: 1) Being transparent about AI use where appropriate. 2) Building diversity into prompts to counteract biases. 3) Asking AI to identify potential biases or perspectives in stories. 4) Reporting mistakes and biases back to the system. 5) Maintaining journalistic skepticism and fact-checking AI outputs.

What are some potential downsides or limitations of using generative AI in journalism mentioned in the script?

The speaker highlights several limitations: 1) AI cannot distinguish between factual and persuasive information, so fact-checking is essential. 2) It has a tendency to make up sources or information, requiring verification. 3) It can't always identify truly newsworthy events or stories. 4) There is a time lag between real-world events and AI's knowledge base. 5) It may reinforce existing biases and lack diversity in its outputs.

How does the speaker recommend using generative AI for idea generation and story development?

The speaker suggests several techniques: 1) Asking AI to generate ideas for features or potential story developments rather than breaking news. 2) Prompting AI with specific contexts or roles to generate relevant ideas. 3) Using AI to identify weaknesses, angles, or perspectives in existing story drafts. 4) Asking AI to generate plans or approaches for reporting on specific ideas or sources.

What are some examples of how generative AI can assist with coding and programming in journalism?

The speaker provides several examples: 1) Generating initial code that can be customized and adapted. 2) Helping identify and explain problems in existing code. 3) Generating code for data visualization or system diagrams based on input data or prompts. 4) Accelerating the learning process for journalists interested in coding.

How does the speaker suggest attributing or citing the use of generative AI in journalistic work?

The speaker suggests several approaches: 1) Following organizational policies or guidelines on attributing AI use. 2) Adding a footnote or separate methodology section explaining how AI was used in the story or reporting process. 3) Linking to a separate page with a detailed methodology. The key is transparency about AI's role where appropriate.

What impact does the speaker foresee generative AI having on the teaching and learning of journalism?

The speaker anticipates a significant shift in how journalism is taught: 1) Moving away from a heavy focus on technical writing skills towards news gathering and editing skills. 2) Incorporating prompt design and effective prompting as a key skill. 3) Changing assessments to be 'AI-proof' or leverage AI as a tool. 4) Increased emphasis on fact-checking, editing, and primary news gathering to differentiate human journalists.

How does the speaker suggest addressing the issue of generative AI potentially spreading misinformation or inaccurate information through search engines?

The speaker emphasizes the need for journalistic skepticism and fact-checking, even when using AI tools or search engines. The speaker warns that AI-driven search results may present misleading or mythical information as facts, underscoring the importance of verifying information and not blindly trusting AI outputs.

What are some of the key considerations or guidelines the speaker suggests for effectively prompting generative AI tools?

The speaker offers several guidelines for effective prompt design: 1) Specifying the context, audience, and role for the AI. 2) Clarifying the type of output desired (e.g., summary, story ideas, code). 3) Setting style rules or guidelines. 4) Asking AI to explain its output and changes. 5) Providing context and making iterative improvements through multiple prompts.

How does the speaker compare the use of generative AI to tools like Google or Wikipedia in journalism?

The speaker suggests treating generative AI similarly to how journalists have traditionally used tools like Google or Wikipedia: as a starting point or complement to their own research and fact-checking. Just as Wikipedia can provide summaries but requires verifying references, AI outputs should be treated as potentially useful but requiring verification and fact-checking by journalists.