AI on AI — A new sales model
AI is the pitch, UI is the product
We all know by now that AI is the hottest buzzword this year. If you don’t know what ChatGPT is, then you haven’t been on the internet. In an attempt to hop on the bandwagon, some companies are popping up at record speed claiming to be the next ChatGPT or the best AI for your company, etc… What are they claiming? Did they source trillions of data points and build a complex LLM that was fine-tuned to produce better results over billions of tests? Did they source multiple data sources to lessen the natural bias of the original dataset? I could go on and on about what they are leading you to believe.
NO. They didn’t do any of that. They built a great UI and signed up for ChatGPT 4 and piped it through the API. Yes, that’s it. They charge a little more than their cost of using it and TADA profit.
It is similar to white-labeling any product and reselling it at a higher (higher misspelled intentionally for human content proof) price point. It really is absurd, or is it brilliant?
We have been white-labeling products for years. Most of our food comes from a small number of corporate farms and is branded a little differently. With the internet, affiliate marketing is the way people make money with links and then kickbacks after the sale. Are they really selling the product themselves? No, they are linking out to the place that is with a tracker so they get paid.
But with AI being so “buzzy” right now, people are building UI on top of them to make it a product, get seed money, and hopefully figure it out by the time they have real clients.
What’s my point in all of this? Simple — UI matters now more than ever. Most startups can’t afford to create LLMs and need to leverage existing models. ChatGPT lost a half billion dollars generating their LLM: https://www.thewrap.com/open-ai-lost-half-billion-making-chatgpt-cost/
What startup that is getting started can afford to tell a seed investor, “We need 500 million to start and then we can start building a product on top of our learnings.” Actually, in today’s market that may actually fly. But if they come in with a sexy UI that returns something real-time (regardless of how they got the data, they could outsource 1000 people to respond real-time 24/7 and claim it is a genius model, we call that customer support — 😄)
But it is the UI that is what people are selling to the masses, but they are selling an underlying AI to the investors. That underlying AI may or may not be original, but as long as it is accurate and fast, do the investors actually care?
There are more articles about these thoughts and here are a few that I encountered:
https://www.linkedin.com/pulse/chatgpt-vs-gpt-wrapper-tools-which-one-right-you-pravash-chandra/
We all want to believe in magic, but when you realize it is just a wrapper, is that enough? In some cases, that may be ok. I love that UI is finally getting recognition of its importance in software development. UI was so undervalued that they had a term called “pixel pushers” for front-end engineers. But, if that “pixel pushing” is what the client sees and ultimately buys, how valuable is it now?
My point is this: Be aware of what these AI companies are actually doing. Are they building a new innovative model that no one has ever seen with datasets that are unattainable by anyone else or are they connecting to a couple of APIs making a UI on top of it and calling it innovative?
