New ask Hacker News story: Tell HN: We take LLMs and Diffusion models for granted
Tell HN: We take LLMs and Diffusion models for granted
4 by behnamoh | 0 comments on Hacker News.
Just 3 years ago, the state of the art image generation models were mostly academic papers that would work in specific scenarios, but nowhere near the capabilities of DALL-E 3, Midjourney, SD, etc. Just 3 years ago, the state of the art chatbots were so crappy that no one thought soon they'd easily pass the Turing test. Nowadays we talk to ChatGPT, download a variety of open-source LLMs, talk about some LLMs being better than the others. But just remember: 3 years ago we didn't even have something like Llama 7b!
4 by behnamoh | 0 comments on Hacker News.
Just 3 years ago, the state of the art image generation models were mostly academic papers that would work in specific scenarios, but nowhere near the capabilities of DALL-E 3, Midjourney, SD, etc. Just 3 years ago, the state of the art chatbots were so crappy that no one thought soon they'd easily pass the Turing test. Nowadays we talk to ChatGPT, download a variety of open-source LLMs, talk about some LLMs being better than the others. But just remember: 3 years ago we didn't even have something like Llama 7b!
Comments
Post a Comment