Using AI to make content more expensive and time-consuming
Does generative AI make Satya Nadella an unnecessary expense?

Using AI to make content more expensive and time-consuming

OpenAI AI recently released a guide on how to prompt generative AI to write your content. Essentially, using this guide and about 15-20 hours of spare time you can produce a piece of content that is grammatically correct and possibly factual if not readable.

Here are the five steps:

  • Write clear instructions
  • Provide references
  • Split complex tasks into simpler subtasks
  • Give the AI some time to think

Test changes systematically.

Without an AI you have to make a general outline (instructions), do research (references), expand the general outline with subpoints (subtasks), start writing (thinking), and edit (testing).

In other words, the AI user has to do everything an experienced and competent writer would do if hired to create a 1000-word piece of content, which would cost between $500 and $1000.

Now that might sound like a bargain, but if you are not a professional writer, you still have to schedule two or three days of your time to complete the article, which is two or three days not doing your job. If you make less than $50 an hour, it might make sense. However, Satya Nadella, CEO of 微软 , recently admitted he uses ChatGPT to write emails. Considering Nadella makes the equivalent of $20,000 an hour (assuming a 50-hour work week), using generative AI in that manner is an expensive use of technology … if Sadella actually went to the trouble to effectively prompt the content as OpenAI suggests.

Let's assume he doesn’t but has someone else to do all that because, after all, he is a very important man. That begs the question, however, if an AI can be trained to do Sadella's work for him, why have a CEO at all, much less pay him $49 million a year?

Perhaps you think I've oversimplified the steps. Let’s take a look at the full list:

1. Write clear instructions:

  • Include details in your prompt to get more relevant answers
  • Specify the steps required to complete a task
  • Provide examples

As a journalist, this all describes what I do to research a story. I do some research to ask relevant questions, interview experts, and provide examples to the interviewees to move the conversation along. I pull all that together in the story I am writing, which for a 1000-word article will take me about 5 hours to complete. So putting that into an AI might save me 4 hours and 45 minutes, assuming the AI can produce it within 15 minutes.

2. Provide reference text:

  • Instruct ChatGPT to answer using a reference text, such as a link to a PDF or a website
  • Instruct ChatGPT to answer with citations from a reference text

OK, that’s what I would do with the final online story: link to relevant sources. I probably already did that under the “provide examples” bullet point, so no time was saved there. This step probably wipes out some of the time I've saved under the instructions.

3. Split complex tasks into simpler subtasks:

  • Since there is a limit on how much text you can insert into ChatGPT, summarize long documents piece by piece to stay within the limit
  • For prompts that involve multiple instructions, try breaking the prompts into smaller chunks

That pretty much describes how I write now, but having to do it for the AI duplicates the effort, so another 2 hours of saved time is eliminated.

4. Give ChatGPT some time to think:

  • Instruct ChatGPT to work out its own solution before rushing to a conclusion
  • Ask ChatGPT if it missed anything on previous passes

That’s my internal discussion during the process. Now I have to wait for the AI to duplicate my process. There goes another hour. We are at one hour of saved time now.

5. Test changes systematically:

  • Evaluate ChatGPT’s outputs with reference to gold-standard answers.

Um, what constitutes the “gold standard”? Are they saying compare the generated content with duplicative content that already exists? Does that mean the best AI content isn’t much better than plagiarism? Just figuring this out is probably going to add another couple of hours of research. So now using an AI to write my content makes me less efficient. Add to that the reality that AI hallucinations could insert false data that I might miss and now I have an inferior product.

So, in the end, using generative AI neither saves time nor money in the process of developing content. Instead, it adds layers of cost and time and produces mediocre material, at best.

On the other hand, it does make an argument for eliminating the overhead of C-level management, so there is that.

Meghan Lacienski

Digital Marketing Manager

10 个月

I couldn't have said it better myself.

回复
Ken Durazzo

Building the foundation of the AI-driven business, unlocking the value of AI to increase both insights and productivity.

11 个月

Hi Lou! Very interesting perspective, coming at it from someone that builds content, it provides an insight to a creative angle for the use of GenAI. While I am not a professional ’content builder’, I follow a similar approach to the one you outline for developing content. It seems that where one might find time savings is in the editing process (at least if you’re not a professional writer!), where you can leverage GenAI to perform copyediting tasks like grammar, spelling and clarity, instead of using it to build the content from scratch. We are all heading down the path of AI-augmented work, the ‘gold’ will be in finding the best ways to leverage augmentation to save time / cost instead of defaulting to use it for every task. Looking forward to following your journey here, you’re likely to be at the forefront of this ‘AI-augmentation’ transformation.

要查看或添加评论,请登录

Lou Covey的更多文章

  • The future of AI is not the Terminator or the Matrix

    The future of AI is not the Terminator or the Matrix

    At Cyber Protection Magazine we are trying to figure out how to incorporate #AI into our process. It isn't easy.

    11 条评论
  • How 'listicles' are created

    How 'listicles' are created

    You see them all the time in news feeds and social media, but do you ever wonder where those articles based on lists of…

  • Kamala Harris wants to save corporations from themselves

    Kamala Harris wants to save corporations from themselves

    A lot of economics pontificaters are, like Josh Barro in The Atlantic are calling Vice President Kamala Harris'…

    31 条评论
  • A note on real "Hillbillies"

    A note on real "Hillbillies"

    J.D.

    5 条评论
  • Let's call the whole thing off

    Let's call the whole thing off

    I was listening to one of my favorite podcasts, The AI Fix with Graham Cluey and Mark Stockley, and they were arguing…

    3 条评论
  • A meditation on "woke"

    A meditation on "woke"

    Every year at this time I stop and have a long think about my experience with Dr. Martin Luther King Jr.

    5 条评论
  • Would AI have stopped the invasion of Iraq? Nope.

    Would AI have stopped the invasion of Iraq? Nope.

    Some articles about generative #AI are about what it could potentially be. Some articles try to explain what it is.

    7 条评论
  • We're coming to the end of the year and I'm exhausted

    We're coming to the end of the year and I'm exhausted

    2023 has been a helluva year. I've written more than 200 articles, produced about 50 podcasts, reviewed hundreds of…

    4 条评论
  • How to fix MSM in 5 minutes

    How to fix MSM in 5 minutes

    You can restore trustworthiness and objectivity to mainstream media Yes, I’m absolutely serious. You can do it all by…

  • Social media destroys credibility. So why are you still spending money on it?

    Social media destroys credibility. So why are you still spending money on it?

    The primary benefit and goal of advertising is not, and never has been, what most companies use it for. Instead, the…

    2 条评论

社区洞察

其他会员也浏览了