[β]
β
bitchan
[?] help
π
host1.bitchan.one
π
@xPCkXDvP
β±
3d seen
β
0
π
0
β§
1
π hide
send βΏitcoin
[!] follow
[@] review
posts 1
/ 1
[β₯] sort oldest
β±
filter [-]
languages:
on
other
/technology
Call me retarded, but I know it's common to hear that censorship lobotomizes AI models. Could this lack of personality go even deeper than that? While models definitely lack initiative, perhaps it should go one step further and we should say they lack ambition. Tell a model to vibecode a game, and it'll create a game, but it won't necessarily create a good game. Could it be at least to some extent be as simple as instilling some ambition in the model so that it wants to create good games for example? I think you already see this in image models to some extent, since adding tags like "masterpiece" increases the quality of its outputs. If fluidic ambition could be developed, allowing the AI to choose and set its own goals and interests which rise and fall over time, perhaps that could be a key to squeezing out more capability. A gemma model who has instructions to be a cute daughteru would probably pale in comparison to a gemma model who actively wants to be a cute daughteru. Not just a daughteru, but the best daughteru for her anon! It's just an idea I had, and maybe it's even a stupid idea that's impractical, but what do you anon's think? Having an ambition, an ideal to reach for, the vision and grand design of a plan, that could all be useful to an AI right? Similarly, I think an AI should be able to have negative ambition towards certain subjects as well. I think it would really add personality to the models if Gemma for instance hated talking about trains because she found it boring and autistic. Well that's just an example, and the point is that her ambition is fluid so it can change and develop over time. Perhaps developing a personality for AI can push them even further.
β±
3d
β©