Commercial applications for ChatGPT

Source: Marginal Revolution.

I agree with the observations in Tyler Cowen’s note.

For more than 20 years as a proposal writer I contended with dreamers who claimed that proposal writing soon would become automated. They had a point: Most winning proposals contain chunks of re-usable content. If only those chunks could be stored in a database and made easy to retrieve and re-assemble, the dream of push-button proposal writing could be achieved. Simple.

And yet every project I ever saw that attempted to create the dream application failed.

I eventually figured out why. The most important part of every proposal development project is figuring out how to answer the question, What does the customer want?

That, however, is the one question that cannot be answered in a purely mechanical way. Crucially, it is the one question for which no database of canned answers can be created, especially in an industry where customers request proposals for products and services that have never existed before.

ChatGPT or something like it can solve this puzzle. I can imagine a customer using a vendor’s ChatGPT interface to write the RFP that kicks off the proposal development process. This controlled interaction would produce the all-important detailed answers to the question, What does the customer want?

The remaining questions proposal writers deal with are:

  • What do we give the customer?
  • How do we do it?
  • What is the benefit?
  • What proves the benefit is real?

All these are easy to answer with re-usable content that can be stored in a database which ChatGPT can retrieve, assemble, grammar-check and print.

Those who are astonished by the potential of AI to replace knowledge workers are right to be. The industry I retired from is a prime candidate for such a conceivable transformation. The key for starting the process is finding the specific context or approach in which AI technology can be helpful.

16 thoughts on “Commercial applications for ChatGPT

  1. To paraphrase, wisdom is the result of making bad choices.

    It seems to me that humans make decisions based on memory, experience and situational awareness. What has been tried before, what worked and what didn’t, and why, and the applying the decision in the realm of the current problem. The last being what I would define as ethics influencing desired outcomes.

    The common denominator is data, past, present and predicted. Essentially, 1’s and 0’s.

    AI can work in seconds what few people are able to solve after years of learning, experience, a bit of luck (that the bad choices were not terminal ones) and then studying the problem and all it’s relevance to past actions.

    You were a proposal writer as I understand it. Coming from a business that required marketing in a very competitive field, advertising photography, I have some experience similar to that.
    Since no product or service is perfect, the pitch is to emphasize the benefits and downplay the drawbacks. Massaging the data, as it were. The client would do the same. Then it is a question of who knows what and how it applies. The more creative the field, the more subjective the data…maybe. AI could evaluate creativity, however, by scanning data of what images influence todays buyers then predict what might work best and then which photographer can produce it.

    With the advent of photo realistic rendering, the product photographers are becoming an endangered species. Engineering drawings on a CAD can be turned into a photo with all the lighting, reflections, shadows, colors and endless variations. My Stihl account went that direction because new products that had not been prototyped could still be marketed via rendering. No studio, assistants, lights, cameras, props or any overhead other than a Mac and software. AI struck again.

    This is a complex issue if we consider historical, cultural, ethical and legal data affecting our decisions. But all that information is still data. And data is quantifiable no matter how huge the amount and that is what AI can do.

    Hawking’s cautionary approach to AI should be a guard rail as we progress. We shouldn’t fear AI but make sure we understand it.


    Liked by 2 people

    1. RE: “My Stihl account went that direction because new products that had not been prototyped could still be marketed via rendering.”

      I saw the same thing in my world, but applied to ships, ship’s spaces and equipment suites. Also to aircraft flight simulators and other types of mechanical trainers.

      ChatGPT is a little different from other AIs in that its domain model is linguistic rather than spacial/visual. Since most people (including writers!) have difficulty expressing themselves accurately, this may be where ChatGPT can prove most useful.


      1. I understand ChatGPT is linguistic but I think the same principle applies to visual or audibles like music.

        There are few new things, but an infinite variety of arranging them. Even brilliant “out of the box” thinking requires know what’s in the box and what is not.

        Liked by 2 people

      2. RE: “I understand ChatGPT is linguistic but I think the same principle applies to visual or audibles like music.”

        No doubt. I’d put it this way: A visual AI that can produce a work of art might be useful to the human being who wants a specific painting (that he can imagine) but can’t produce himself.


        1. That’s the objection by some. But as a photographer, I realize that before it was considered its own art form, some denigrated as a crutch for those who can’t paint realistically.

          It worked for me, however.😎


  2. I think that modern iterations of AI are going far beyond cut and paste. What makes it revolutionary is that they are not programmed per se, they are trained on data and create their own algorithms for doing tasks.

    Liked by 2 people

    1. “…they are trained on data and create their own algorithms for doing tasks.”

      Isn’t that what humans do daily?

      The problem is whether the data is good or not. Will AI discern, ignore or create conspiracies if the data is questionable or just flat wrong?

      Liked by 2 people

      1. You can test drive ChatGPT for yourself, if you’d like.

        1. Go here:

        1. Click the Sign Up button to create a free account.
        2. Navigate to the ChatGPT user page, or visit the Playground. (Along the way you’ll need to enter your name in a form.)

        3. You use the app by asking a question or stating a request or command. The app displays its response after a short delay. You can then tell it to modify its response in some way, or add to it, or you can change the subject, etc.

        I think you’ll find that ChatGPT is very good at understanding your words and talking back to you responsively, but that it clearly doesn’t “think” creatively.


        1. Thanks for the link…

          Here is a prompt I tried.
          “Write a defense of American support of Ukraine in its war with Russia.”

          Here is the text it generated in about ten seconds . . .

          “American support of Ukraine in its war with Russia is essential in preventing further aggression from Russia against the Ukrainian people. Russia has been the aggressor in the conflict, illegally annexing Crimea and supporting separatists in eastern Ukraine. The United States and its allies must stand with Ukraine in its fight for freedom and democracy.

          American support of Ukraine serves as a deterrent for Russia and sends a strong message to the entire international community that aggression and human rights abuses will not be tolerated. By providing military aid and diplomatic support to Ukraine, the United States is demonstrating its commitment to the protection of international law and the defense of democratic values.

          The United States is also providing economic aid to Ukraine, which is necessary for Ukraine to rebuild its economy and strengthen its defense capabilities. This will help Ukraine become a more self-sufficient nation and better able to stand up to Russia.

          Finally, American support of Ukraine is a symbol of American solidarity with the Ukrainian people and their efforts to build a better future for their nation. By providing aid and support, the United States is standing with the Ukrainian people in their fight for a free and independent Ukraine.”

          Not bad!

          Liked by 1 person

        2. Thanks I saw the tutorial the other day. I plan to play with it at some point. But creativity can be quantified and fed back by AI as far as I can tell. Let’s just take painting. Realists will interpret the data they see. Abstract art will do the same based on the memories and experiences of the artist. Data.

          We could tell a computer to paint an image in the style of Dali and if enough data were offered for AI to quantify the style there is no reason for it to not create a fresh image.

          Then we expose the AI to a thousand artists and it could create something fresh by either blending styles or avoiding all 1000 and offering a new view.

          Which is what in effect happens now. Art is not created in a vacuum. Visual, audio and sensory clues abound in the brain all based on experiences. Experiences tempered by culture, environment and DNA.


          Liked by 2 people

  3. I have also written ship repair specifications. There usually is a lot of cut-and-paste involved because most specs for similar classes of ships are interchangeable. HOWEVER, there were many times when the writers, including myself, had to address certain nuances to a particular repair. Aluminum failure in CG-47 Class Cruisers come to mind.

    Not a yea or nay concerning the app. Just a comment based on experience


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s