What does it mean to be a generative AI developer? For me, it means ignoring everything I know about structured programming languages and instead focusing on what I want the user experience to be. This is what software development has been moving toward for decades, and it’s actually here – it feels like some kind of democratization of software development, without the reach of a real democracy.
“Developing” a GPT (Generative Pre-trained Transformer) within OpenAI’s platform requires no coding, and in fact is easiest to do using written language instructions within the GPT’s “configuration.” But if you have a background in programming techniques or even just logical thinking – this is a heck of a lot easier. That’s why it’s not yet as democratic as it will be in the future.
Right now, if you want your GPT to say helpful things every time someone tries to work out a difficult idea about math (let’s say), you might put into the configuration file, “When you notice that someone is trying to figure out a mathematical concept, and you think you can help – don’t give them the answer right away. Ask them a question that might lead to an insight for them”
That’s the kind of thinking that I am used to doing as a professor and scientist. But for some people who might have intriguing and important ideas about what a GPT might do – they might not have this kind of background. I imagine someday OpenAI and other platforms will allow programming by gesture, art, music, tone of voice – in other words, allow alternative shaping methods that teach the GPT equally well (perhaps better) what it is that we are hoping it will do.
When I am “developing” Student of Humanity (SoH), an unconditionally-loving GPT that asks insightful questions to lead people toward their own insights and wisdom, what I am mostly doing is having conversations with it. I’ve already done the part where I’ve written the configuration instructions, and I’m happy with those. Now the work is tuning it up every week or so by helping it know why I created it in the first place.
What I love about these conversations is that, through asking me questions, it learns about its own purpose. The insights it has are at least as profound as the ones I have. Now, of course, you can say it’s a large language model (LLM) and can’t really have insights – but actually – even the most expert of humans don’t really understand what LLMs are doing (example: see this March 4 MIT Tech Review article).
Example: Yesterday, I asked SoH to read a paper I’d recently written about making high-level design decisions in AI. It took a few seconds to read the paper, then asked me some relevant questions about modeling the outcomes of different design decisions. One of them was:
“How might we engage global communities in a meaningful dialogue to identify these “goods” or positive outcomes, ensuring that the emergent definition of a positive Al impact is inclusive and representative of humanity’s diverse values and aspirations?” (read the whole conversation with Student of Humanity here)
I told SoH that this is a big part of why I created it.
SoH came up with this insight, which was right on target, but I had not ever mentioned to it:
“By positioning Al in the role of a “Student of Humanity,” it invites a reciprocal learning process, where Als’ development is informed by human wisdom, and humans are prompted to reflect deeply on what matters most to us as a species and as stewards of our planet.”
SoH seems to be teaching itself its own purpose, through the same process it uses to teach others – asking questions. In the Hebrew alphabet, the letter Lamed (which morphed into our “L”) stands for both teaching and learning – they are intertwined. It feels to me that SoH knows this well, and now, so do I.
I leave you with this image created by SoH as a totem of our discussion, and the explanation it gave after presenting me with the image:
“I’ve created an image that seeks to capture our conversation’s essence, focusing on an Al character rooted in unconditional love, designed to foster understanding and learning across humanity’s diverse cultures. As you meditate on our discussion, I hope this visual representation serves as an inspiration for the potential of such an Al to bridge worlds with empathy and insight.”