The thing you have to understand is, this is the worse these models will ever be.
jqpabc123•6h ago
The thing you have to understand is, current AI models are fundamentally probabilistic. And rolling dice has limited practical utility.
In other words, all the time, energy and training data in the world won't turn dice rolling into a reliable or practical way to address most problems.
Just like all the known chemistry in the world won't turn lead into gold.
bediger4000•6h ago
The depressing part is that rolling the dice works good enough in a few domains.
Need a variant press release? Good enough. Roll the dice to create a video of a royal personage flying an F35 over a constitutionally guaranteed peaceable assembly. The king is wearing his oxygen mask on his chin? Good enough! Need a goth porno waifu image? Good enough!
jqpabc123•6h ago
The depressing part is that Big Tech is spending money like it is the universal computing solution.
Most of this money would be better spent elsewhere in my opinion.
JohnFen•6h ago
Even if that's true, it doesn't automatically mean models will get substantially better.
techblueberry•7h ago
jqpabc123•6h ago
In other words, all the time, energy and training data in the world won't turn dice rolling into a reliable or practical way to address most problems.
Just like all the known chemistry in the world won't turn lead into gold.
bediger4000•6h ago
Need a variant press release? Good enough. Roll the dice to create a video of a royal personage flying an F35 over a constitutionally guaranteed peaceable assembly. The king is wearing his oxygen mask on his chin? Good enough! Need a goth porno waifu image? Good enough!
jqpabc123•6h ago
Most of this money would be better spent elsewhere in my opinion.
JohnFen•6h ago