Fantastic read! I have one question regarding the OpenAI/Google moat. If "details of the architectures" don't matter much, and setting aside their current advantages in user data, distribution and pricing, what should we expect the technical moat be based on? Superior access to hardware and compute? If the best future models are driven by new ideas, you essentially need to hoard most of the top AI researchers to sustain an advantage and it isn't clear any company is able to do that.
Great read, thank you for sharing! OpenAI is definitely in the lead, and we're already seeing an open-source explosion with agents and llamas, but I agree that it'll take time for anyone, esp. open-source, to earn B2B trust, and that's what makes money. 90% of consumers won't pay a penny for advanced LLMs given the good-enough quality of open-source products. Bing and Bard will remain free anyway for customer retention, I predict. Signed, a Microsoft employee 🤓
Fantastic read! I have one question regarding the OpenAI/Google moat. If "details of the architectures" don't matter much, and setting aside their current advantages in user data, distribution and pricing, what should we expect the technical moat be based on? Superior access to hardware and compute? If the best future models are driven by new ideas, you essentially need to hoard most of the top AI researchers to sustain an advantage and it isn't clear any company is able to do that.
Great read, thank you for sharing! OpenAI is definitely in the lead, and we're already seeing an open-source explosion with agents and llamas, but I agree that it'll take time for anyone, esp. open-source, to earn B2B trust, and that's what makes money. 90% of consumers won't pay a penny for advanced LLMs given the good-enough quality of open-source products. Bing and Bard will remain free anyway for customer retention, I predict. Signed, a Microsoft employee 🤓
But MosaicML didn't publish GPT-3, OpenAI did, using Microsoft tech, didn't they?