Re: About vibe coding
bixbox <[email protected]>
| Newsgroups | comp.ai |
|---|---|
| Organization | A noiseless patient Spider |
| Message-ID | <[email protected]> |
The True Melissa <[email protected]> writes: > bixbox wrote in article <[email protected]>: >> >> The True Melissa <[email protected]> writes: > It feels so much realer, doesn't it? It is def does. For some reason is the most refreshing part of my day. I don't know how to explain but the feeling is that. > >> The monopolization of the information substrate, and the communities is >> such depressing and limit the ability of innovative idea. > > I have ideas to do something about it, but no one's with me. The StuffAsAService* like "AI as a service" that the current mainstream are using, while I don't question the it make those services a commodities, have a real economic and innovation impact because as of now all those this are build in a way that require a massive amount of memory processing power. Before the AI those *AsAService required economic investment to build from the scratch but is not massive as the current one. Build the current LLM models require several order of magnitude on investment to train, and those company does not start from zero, they already several generation of Model that definitely help the evolution of the next generation. And BTW the cost that we pay at the Inference As A Service is to pay and sustain the cost of the training. That is a monopoly the feed himself, as well as all the input that we feed thought using the inference and sending them prompt. The practically owning the wells. Maybe and maybe, we will see a evolution at some point, like when we went from Main Frame to Personal Computer, we may observe a similar trajectory, from AI Main Frame like Anthropic and OpenAI to Personal AI running on my specialized device. In fact I feel there is a diminished return on larger model, every new generation is not improve linearly anymore and the feeling is that that is flatten out. Many thing have to happen but it can happen. bix