Rush or Reflect? What Deadline Reporters and Investigative Journalists teach us about AI

Rush or Reflect? What Deadline Reporters and Investigative Journalists teach us about AI

Deadline reporters and OpenAI's ChatGPT 4o both deliver fast, excellent results. Same day for a deadline reporter, milliseconds for the AI.

Investigative journalists and OpenAIs ChatGPT o1 take longer and dig deeper. Sometimes weeks for an investigative journalist. Often seconds for the AI.

After hours and hours of banging on both models, as well as decades leading deadline reporters and investigative journalists, my experience so far leads me to an immutable conclusion: one is just as valuable as the other.

And, counting Emmys is no way to measure their utility.

  • GPT 4o is a so-called “System 1 AI”— with honed, intuitive and near-instant predictive outputs.?
  • GPT o1 is a newer “System 2 AI“ and, as with an investigative journalist, we afford it more time to be deliberate and analytical.?

By design, System 2 AI’s are proudly slower, patiently thoughtful and harder to manage (in terms of compute resources and expense).?

I used the exact same “zero-shot” prompt in both ChatGPT 4o (System 1) and ChatGPT o1 (System 2):?

So-called "zero-shot prompt" as the input includes no examples

  • ChatGPT 4o delivered an excellent output in about two (2) seconds and 462 words (approx 616 tokens). ?
  • ChatGPT o1 delivered a more detailed, thoughtful, output in about six (6) seconds and 684 words (approx 912 tokens). ?
  • Twice as long to process the same prompt. Along the way, indicating its step-by-step thinking.

Expressing its own "Chain of Thought"

System 1 Deadline Reporters generally work with more straightforward data and cached insights. "I can see it. It's a fire. They are getting a handle on it now. I know because my experience tells me the color of the smoke changes from black to grayish white with water."

System 2 Investigative Reporters, however, sift through vast amounts of data, double-check against training instructions and deliver more nuanced outputs—even if it takes longer. "This is the 6th such fire in the same block in the last 415 days. Three firefighters have been injured here in that same timeline. Although arson is blamed for the previous fires, no arrests have been made. Damages to properties here already exceed 14 million dollars."

Were we talking about AI Systems or newsrooms? Both. The point is to conflate the two, making the analogy memorable and clear.

System 1 AIs keep interactions snappy, accurate and fast enough to have a real-time dialogue with it. It's not better. Just different. More instant utility. Fewer costs.

System 2 elevates the quality of results. Again, not better. Just different. And, more costly with added compute.

As we integrate AI into business strategies—TV journalism, marketing, HR, Finance and analytics—understanding when to optimize for speed vs. depth can make or break the experience or the budget or both.

Days, milliseconds, weeks or seconds. It's about time.

*I'm not a 10-thousand hour AI subject matter expert, but I am a super-user with about 2,000 hours of "fingers on keyboards" using AI tools and many more hours of online courses and certifications.


?

要查看或添加评论,请登录

Kerry Oslund的更多文章

  • I Love AI—Let Me Go Last

    I Love AI—Let Me Go Last

    AI preachers want to move faster. But, the entire choir isn’t quite ready.

    9 条评论
  • The Making of an AI Video

    The Making of an AI Video

    If Porsche ever wondered about its impact on ordinary people, particularly the Pisano family and their impressively…

    9 条评论
  • The Patagonia Principles

    The Patagonia Principles

    And with that single word, Patagonia, I got every CEO's attention. I did.

    9 条评论
  • Thinker-Doer Revisited (Post Steve Jobs)

    Thinker-Doer Revisited (Post Steve Jobs)

    The days of separating thinkers and doers are ending. There's just no reason for it anymore.

    4 条评论
  • "Baja, Baja" The Unsung Half of Strategy

    "Baja, Baja" The Unsung Half of Strategy

    "Baja, baja" (come down, come down), the statue cries. Laddering-up has become go-to corporate phraseology, signaling…

    7 条评论
  • A.I Bias is Good (said the Brazilian Hot Dog)

    A.I Bias is Good (said the Brazilian Hot Dog)

    *Written by a Human (fact checked by GPT) When we talk about AI for broadcasters (large language and natural language…

    14 条评论
  • Carr's Incentive

    Carr's Incentive

    Spectrum is the broadcast equivalent of gold and FCC Commissioner Carr (R) did a little test panning last week at the…

    16 条评论
  • Translation, Voice Cloning and Lip Synchronization

    Translation, Voice Cloning and Lip Synchronization

    Dear ATSC Members-- Triple Play Translation is the confluence of three IP technologies, representing a possible "killer…

    1 条评论
  • Core Curriculum: Power of Core Networks in ATSC 3.0

    Core Curriculum: Power of Core Networks in ATSC 3.0

    101: A core network, in ATSC 3.0, is the pivotal orchestration layer that provisions spectrum & bandwidth across…

    1 条评论
  • FCC Task Force: Full IP transition to ATSC 3.0, including affordable devices and use of ancillary fees as a funding mechanism

    FCC Task Force: Full IP transition to ATSC 3.0, including affordable devices and use of ancillary fees as a funding mechanism

    CHARTER (as generated by Chat GPT) 1. The official designation of the federal advisory committee is the “Task Force for…

社区洞察

其他会员也浏览了