Thoughts and Best Practices for Generative AI
My thoughts on all this AI-hype--this blog post was written with the assistance of slightly-tuned GPT-3:
Generative AI is the newest gold rush on the web. Companies like Jasper.ai have scaled to $100 million in yearly revenue in less than 2 years, while at the same time the price of software has never been lower (~70% deflation as a category since 2000!). So there's clearly something special going on in the generative AI space right now! Unfortunately, differentiation between companies using generative AI (as in, making API calls to companies like OpenAI/HuggingFace) is incredibly low.
Most of the SaaS companies built off OpenAI's models are simply marking up API calls 10x with a basic user interface (UI). The release of free access to chatGPT created even more competition in this space, and we can expect more free models this year that will further reduce prices (as of this writing, Bing just announced an updated version with GPT integration, and Google released Bard).
So what does this mean for those who want to capitalize on generative AI? How do you prevent your generative AI SaaS from going to 0? Here are three things to realize:
1. The cost of creating content has never been lower - Every business should have a social media presence across every platform and a blog if they want to succeed online; it’s never been cheaper thanks to generative AI.
2. Marking up an API call 10x won't cut it anymore - To stay competitive, businesses will need build their own unique AI model (like Midjourney fine tuning Stable Diffusion), or do more with GPT technology than they’re doing now.
3. The web is about to get much larger - With reduced costs for content production and improved models available for use, we can expect an influx of new content flooding the internet from authors hoping to seize their share of the market’s attention - while at the same time the overall pie (in terms of the web's TAM) won’t be growing nearly as fast.
As a result I predict four things will happen:
1. Companies that build off LLM's (Long Language Models) will be doing much more work per user prompt - each prompt could result in hundreds of model calls which could then be put together into impressive outputs over time or periodically (almost like Bot as a Service).
2. We’ll also see acceleration in the creator economy as niche human communities thrive due to increased noise from automated content generation tools; finally allowing us to witness first-hand those “1 000 fan” theories come true!
3. Unfortunately however it may become increasingly difficult for ecommerce companies or anyone selling products/services online as SEO may look drastically different than what it does today due to increased competition from automated content generators and lowered barriers-to-entry caused by generative AI technologies!
4. We’ll see generative AI-based social media companies that allow users to be the star of their own show.