Is OpenAI dead?

Day 221 / 366

When this year began, OpenAI was the most exciting company in the world, miles ahead of its competition. Now, it’s August and we still have not heard anything of significance from them yet. On the other hand, Facebook and Anthropic have made huge improvements.

Facebook came out with llama3, and now we have an open-source model that performs just as well as GPT-4. Anthropic’s latest sonnet model is also out-performing GPT-4. It seems that OpenAI has lost its moat, everyone else has caught up.

To make things worse, there have been so many people from the core team leaving OpenAI.

It is not as if it's making tonnes of money either. It’s losing billions this year and needs to raise some cash soon in order to survive.

Their whole bet is that transformers and LLMs would alone be sufficient to reach AGI. This is looking highly unlikely now. LLMs have not gotten that much better recently. We got GPT-4 in March last year, and there are still no signs of GPT-5.

I don’t think the situation is that hopeless. Their biggest advantage is their higher reach. And they have the backing of Microsoft as well, who can fund them for a long long time.