The AI landscape is about to get a whole lot more interesting. OpenAI is hinting at making their GPT-3.5 model open source. Yes, you heard it right. The same model that’s been the talk of the town, the belle of the AI ball, might soon be available for everyone to tinker with.
Now, this isn’t official yet. The news comes from a Twitter thread where Andrej Karpathy, a bigwig in the deep learning world and a part of the OpenAI team, was asked why he’s been playing with Llama 2 instead of building Jarvis for OpenAI. His response? A cryptic hint that OpenAI might release its models as weights.
The Llama in the Room
This whole conversation comes in the wake of the recent release of Baby Llama aka llama.c. Karpathy has been exploring the concept of running large language models (LLMs) on a single computer as part of his recent experiments. This was inspired by the release of Meta’s Llama 2.
Karpathy’s approach has been successful in achieving highly interactive rates, even with reasonably sized models containing a few million parameters. He’s trained a 15 million parameter model of the TinyStories dataset.
The Return of the OpenAI Jedi?
If this pans out, it could signal a return to the original ethos of OpenAI, which started as an open-source non-profit company. Karpathy was one of the initial founding members who played an active role in contributing to the open-source community.
So, what does this mean for the AI world? It means more access, more innovation, and potentially more llama-themed AI models. But remember, this is all speculation at this point. We’ll have to wait and see what OpenAI officially announces.
In the meantime, keep your GPUs warm and your data sets ready.
