Will AI Cement the Oligopoly of Big Tech Or Become "Democratized"?

In the short run, unfortunately, it is likely that Microsoft, Google, and perhaps Meta and Amazon will become even more dominant. Not only do they have the resources to develop Large Language Models (LLMs) themselves, they can even use their control of these resources to acquire large stakes in start-ups (Microsoft OpenAI, Google Anthropic) that could potentially rival them. The deal is simple: we let you train and run your models in our cloud, and you give us a big share in your company. In the end, should you become dangerous for us, we will just swallow you.

In the medium term, there is reason to be more optimistic, however. Meta is breaking ranks and has open-sourced its Llama-Model, at least part of it. And some clever people have already hacked and compressed it and made it possible to run it on a MacBook (and some other machines as well). The local version is by far not as good as ChatGPT, but it's a beginning. Install it https://github.com/nomic-ai/gpt4all , and you might feel a little bit empowered.

In the end, it will however not be important to have local models. What will be important is that there are a lot of them, easily available at low cost. That will make regulation of AI no easier, but it will reduce the need for regulation of Big Tech.