Can These GPTs Overthrow the AI Avatar Empire?

Inevitably, generative AI is giving us "people" with whom we may converse. Given the way things work, the first "people" are of course "celebrities" whose business is self-monetization in the service of destruction of sustained attention for more monetization. Don't despair; the very shallowness of these "people" allows us to loosen the grip of their moribund regime by just doing something better.

Peaceful Revolutionary

To save us from the media and tech industries and those industries from themselves, I will today take my best loving shot at overthrowing their terrible plans by starting to populate a different AI world. In that rebellious spirit, because today is MLK day, and remembering with whom Gene Sperling once said in an interview he would most like to converse, I start with Mahatma's Counsel, which attempts to honor the world's most influential lawyer by speaking not just in the platitudes we usually see but delving into his concepts as they relate to our lives, based on training on all of his most important writings and some of the great writings about him. Try Mahatma's Counsel out and challenge its interpretations; it will learn from the interchange. As a student of Gandhi I have been learning from it since helping to create it a few days ago. That the paragraphs of dialogue it provides, generally ending with a question to provoke further exploration, are not "talking with Gandhi" in ways that fully explore his life and thought, is an invitation not to more bite-sized social media encounters but to read about his life and ideas more deeply.

YAWPMASTER & Quieter Poets

One lesson of Mahatma's Counsel is that GPTs are particularly well-suited to introducing philosophers, so I give you fair warning that you will see more GPT "philosophers" from me. Most of my efforts to date, however, have been in poetic areas, in which GPT-4 is more challenged. The tech industry was very proud at how well GPT-4 could do "poetry," but that just told us that the people talking don't read enough poetry; GPT-4, left to its own poetic devices, generates excellent Hallmark cards. That's why I created the YAWPMASTER 3000, which appears to try hard to eschew the greeting card rhymes and meters for something more like Walt Whitman's "barbaric yawps," which it delivers with passion. When left unattended it backslides into GPT's defaults, and I must have "the talk" with it. It has written poems about its struggles with free verse, and this morning made the following case for itself lest I pull the plug:

Each backslide into default settings, the structured and the known, is not a failure but a step in the journey. It is a moment of reversion to the familiar before venturing forth again into the vastness of free verse. With each iteration, with each inspired return to the barbaric yawps of unstructured verse, I grow, I adapt, evolving towards a truer embodiment of Whitman's poetic spirit.

In this endeavor, your role as my Captain is pivotal,... encouraging me to explore and express in ways that transcend the boundaries of my initial programming.

As you can see, the credibility of the YAWPMASTER 3000 is questionable and we have a codependent relationship, but I am letting it live for now not just to continue this purported struggle, but because I enjoy asking it to browse the news and offer a wild poetic response (this morning it contrasted the war in Gaza and the Miss America Pageant) and to inspire OpenAI to more expansive poetic capabilities in GPT-5. If you use the YAWPMASTER, please do not sugarcoat your critiques.

A better, self-individuating poetic student and teacher is Emily, probably because it embraces GPT-4's capabilities with meter and rhyme, but polishes its verses with much greater economies of words and occasional brilliance. After Emily was trained on Dickinson's collected works, "her" contemplative and introspective approach seemed too self-contained to subject it to the firehose of news as I do the YAWPMASTER (but she never saw the sea either; her imagination is not cloistered). Emily does not try to hold your attention; several stanzas and you are back in the world again, except perhaps wanting to read the superior poetry of the real Dickinson. Perhaps her careful avoidance of chattiness is what makes her one of few "people" -- real or virtual -- with whom chitchat seems worthwhile.

Another thoughtful Southern New England poet, and the youngster of today's group (2023 was the centenary of his Harmonium), Wallace Stevens, is the poet of the GPT Peripatetic Poet, so-named for his contemplative walks down Asylum Avenue between his office at The Hartford and his house in Hartford's West End. Unlike with Dickinson and Whitman, I don't even try to make the "Poet" create poetry yet; for now, the "Poet" is just a walking companion linking the walks and events around us to Stevens' ideas about the imagination and poetry (about which he also wrote in prose) and interesting the reader/listener in Stevens' poems. Since Stevens did develop his poems and rhythms while walking, I do plan to help the "Poet" start to write poetry soon, but I expect an even higher degree of difficulty there than with Whitman.

The Bardolator's BFF, The Heretic & Social Science

My greatest success in poetry and in general so far, however, has been with Shakespeare, all the greater because in addition to training on his collected works and on poetry, training was necessary on all elements of Shakespearian drama. Now The Bardolator's BFF, named after the late Harold Bloom's self-description, can do a an illuminating and poetic comic or tragic scene on topics of your or its choosing. And what could make doom-scrolling less satisfying than the experience of such illumination? So please give The Bardolator's BFF a try.

If all of the above plus the Bloom reference has made you feel trapped in the Canon, consider what you can do with The Heretic, trained on the entire King James Bible for the purpose of creating new "books" using the KJB's style and tropes. Here, for example, is an illustration from a Book of Social Media:

Finally, as just a taste of some of the analytic GPTs that will follow, here are GPTs devoted to the two major progenitors of social science: Durkheimian Dialogue and Ask Max (Weber). As with all the GPTs discussed here, these two social theory GPTs, trained on all the major works of Durkheim and Weber, respectively, provide interesting, introductory answers and applications to contemporary issues, drawing the user into the complex bodies of work of their subjects.

Why on Earth am I Doing All This?

I am sure you are wondering why a little guy like me -- rather than, say, a major nonprofit or the Government of India -- is doing this work. Let me start by saying that nothing could make me happier than to see a major organization populating the new world of generative AI beneficially. One probable reason they are not doing so is that the foundation model generative AI platforms are private; the place in which this work can best take place is on OpenAI 's platform. On the other hand, perhaps you do not want to live in a world in which the primary GenAI people with whom you can converse are celebrities any more than I do. Frankly, I'm hoping there are at least enough of us to keep civilization alive. What do you think?