LLAMA2 is out and it’s pretty fancy, and there’s a news report that they’re making a way bigger one. But if it’s just open source stuff, how do they justify the massive costs to make it? It seems like everyone can just use it for free.
In: 156
If you played around with ChatGPT when it first came out, you might’ve noticed some hilariously incorrect answers like it very confidently claiming 3+4=17 or that there are no countries that start with the letter “V”. Since then it has gotten a lot better, and a big part of that is additional data that they can collect from users such as when they click whether the answer is good or not.
Both LLAMA and ChatGPT were incredibly expensive to train (several millions of dollars), but ChatGPT proved the proof-of-concept of the potential of large language models. So Meta is happy to release it for free in order to bring more awareness to their platform like others mentioned, and also to improve the product. Many people in the ML field have become more interested in LLMs, some of the potential issues with GPT or Bard, and the open-source nature would really bring in a lot of contributions.
No doubt they are hoping to incorporate this into their social media platforms and future technologies when it is ready, so it’s a form of investment.
Latest Answers