The Future of AI Is Small, Specialized Models — Not Giant Chatbots
Smaller AI models could soon run on your laptop, saving money and energy.
A Reddit user argues the future of AI is smaller, specialized models that work better when they’re smaller and are easier to train, so datacenter-level power may not be needed. They note tasks like OCR and segmentation are already being solved by single-digit-B-parameter models, and say a 9B model trained by a lab tops the benchmarks.
They don’t call LLMs a gimmick, but predict that ten years from now we’ll have specialized models for the tasks we care about, while LLMs will fail to do everything all at once unless it’s one model with a hundred under the hood. They also say current LLMs like Opus 5.5 are in a league of their own but are losing a lot of money on subscriptions.
- Small AI models (like 9 billion parameters) can outperform giant models on specific tasks like reading text from images.
- These small models are cheaper to train and run, and can work on your own device without sending data to the cloud.
- Big AI companies may struggle to make money from their huge general-purpose models as specialized models become more common.
Why It Matters
Smaller AI means cheaper, faster, and more private tools that can run on your phone or laptop.