3 ms·
It’s not that it’s worse, it’s just refusing to do coding without persistent prodding and the right prompts. Some think they are trying to do something with ali
by twobitshifter 3y ago
It’s not that it’s worse, it’s just refusing to do coding without persistent prodding and the right prompts. Some think they are trying to do something with alignment, and maybe prevent it from giving code away so that they can upsell.
- MattDaEskimo 3y agoThe new GPT-4 model has a context length of 120k. For consumers this equates to slightly more than $1/message input-only. If ChatGPT is using this model then it's more reasonable to assume that they are bleeding money and need to cut costs. People really need to stop asking ChatGPT to write out complete programs in a single prompt.
- twobitshifter 3y agoInteresting, how is writing less code cutting costs for them? Does this get back to the rumor that the board was mad at Altman for prioritizing chatgpt over money going into research/model training?
- afpx 3y agoCode is very token dense, from what I understand.