I don't think AI should have non-thinking modes. They are so much worse, not just in conclusions, but in tone as well. We should focus on making reasoning faster and replacing non thinking models entirely, as they are only useful for the most basic stuff.
>>107921185They should for some things. If I want to know the cooking time and temp of a casserole I just want the answer, I don't want it to spend 30 seconds deep reasoning about the melting points of every kind of cheese.
>thinking model No such thing
>>107921221>can't readsad to see maybe you should reason about it a bit
>>107921214That's why I said they are useful for the most basic queries.
>immediately look at the rubix cube>missing a corner, center of one side is multicolor, seems to have 7 colors total, has a small chunk of one next to itlol
>>107921185thinking mode is a marketing gimmick
>>107921185I don't think Indians like you should be on people's internet, you should be on IoT.
>>107921258Also look at the desk, lmao.
>>107921185Thar generatice ai that you/many use for images music videos (...) usually has no special thinking mode.
>>107921637The point is about text LLMs, image models are wack
>>107921185“Thinking” mostly makes AI look smart, but not reallyThey really do not think, so it just merely skews token generation towards a certain direction, sometimes even the wrong one, but at least it’s good for “debugging” the little retards
