Sr SDE @ByteDance Seattle, Ex-Amazon | YouTuber | Mentor | Teacher System Design & Programming | Full Stack Java SpringBoot React AWS Developer . Email : [email protected]
It's fine to create your startup or project fully on top of ChatGPT or any other GenAI models. It doesn't matter whether people says your startup is just a wrapper over LLM. All you need is to solve a "problem". Millions of software developer just write wrappers on top of CRUD API's, using frameworks in some programming language and have created multi-trillion dollars industry. Hardly any software developer created a new programming language. So, you also don't need to create a new GenAI model. Just use any model, solve a problem and generate millions of $$. 🤝 Share/Repost if you agree.
I disagree, like many others. If your wrapper product isn't good enough, it won't generate revenue. If it is good, LLM giants will integrate it as a feature. Unless you're aiming for short-term profits or have substantial funding like the likes of Perplexity, it's hard to sustain a long-term business.
Bingo, there's no need to invent the wheel again :)
whole company will be dusted the day OpenAI provides the same feature
Creating LLM isnt creating a class.. it has to go through whole life cycle: training,, tuning, accurancy, security etc etc. so your advice is okay to dev but not to corps..
Only when your solution will not be added in the next iteration of that GenAI model.
The ultimate motto should be to solve people's problems.
Well said Pushkar
That's definitely right 👍🏻 great pov
Computer Vision Research @ Akridata
1yHard disagree. If your entire product is based on wrapping a model's API in a nice, presentable UI, then you have no moat, regardless of how much finetuning or prompt engineering you may have done. And if you have no moat, eventually someone else will do the exact same thing that you are doing but slightly better. Or even worse, there will end up being open source alternatives. Oh and let's not forget about ChatGPTs GPT store. Unless you have some crazy VC backing and extremely good marketing team, your product won't survive a day, especially in the world of AI which is changing every single day. LLMs should be a feature/addition to your main product, not the product itself. Using LLMs as a product is equivalent of using SQL as a product. Except, SQL is usually a requirement in the product, LLMs on the other hand, are nice-to-haves.