10 Comments
Apr 9Liked by Devansh

This was interesting to read now, a few months after it was written. We seem to be going through the first period of disillusionment. I predict that the curve is actually non-convex, so there will be different periods of disillusionment (as well as other peaks of inflated expectations and so on) as things are scaled up over the next few years.

Expand full comment
author

That's a good prediction. With powerful technologies, we always find new possible uses, leading to multiple hype cycles before we eventually settle.

Since you're reading this a few months later: what do you think of the diagnosis. Do you think it matched up well with reality?

Expand full comment

I think you did well. I do not think we have reached the bottom of the first period of disillusionment (as you elegantly called), and let's hope investors and pushers do not get too impatient before things start picking up to increased efficiency (in the sense of model utilization by users) too. They will likely pick up again by July, when state-of-the-art models are released again. Then after that, if your prediction of multiple cycles applies, we would expect a sharp increase in expectations, probably corresponding to the time when many people start using the new models (given that the current set of users is large, they will quickly find the increased performance with it, and thus they would not be considered the ones hyping the model expectations in July). What do you think of that analysis?

Expand full comment

Appreciate this fresh angle on Prompt Engineering and what people misunderstand about it.

It's very helpful for creating a safer and fairer AI.

Expand full comment
author

Glad you found this valuable

Expand full comment

Prompt engineering course-sellers are like the pick-and-shovel sellers during the California Gold Rush of the 19th century: they're the first ones to get rich, and everyone else who comes later is going to have a LOT more competition.

Expand full comment
author

Great analogy

Expand full comment

First of all, full disclosure, I’m an artist and poet, a designer for over 20 years. I have completely embraced ai imaging in the past several months. It has been so exciting! I work with ai imaging every day and prompt language is serious business for imaging. I get asked a lot on Discord what prompts I use to create. I have to say it’s really about the basics that I learned in art classes in high school and college plus my own knowledge of poetry and language that has helped me. I love creating with ai, whether it is with bing dall-e copilot, adobe, Fotor, google, etc... I do find barriers though.

Expand full comment
author

" I get asked a lot on Discord what prompts I use to create. I have to say it’s really about the basics that I learned in art classes in high school and college plus my own knowledge of poetry and language that has helped me."

Exactly

Expand full comment

Barriers seem to be subjective and intentional, however. The other day I asked the ai to create an image “feature political parties and us elections” Well, the ai yelled at me saying that I didn’t have permission. Really? Ultimately, who’s pulling the strings here?

Expand full comment