I’ll say that on youtube, the implementation has been problematic.
There’s a channel where the creator has used a speech synthesis for long before the modern crop of AI text to speech. He still uses the same old speech synthesis, not even one of the newer models, but flagged as ‘AI’.
Conversely, blatant slop appears with zero indication. With no option to report as AI slop.
But the point is that if someone used an LLM to do some code completion, technically AI but I don’t think anyone would consider something like ‘multiply all members of this array by this other array’ to be ‘slop’ when turned into what an LLM would turn it into. There’s some point where it’s lost the plot entirely, but it is going to be relatively unlikely that folks ever give up the ‘still micromanaging, but letting an LLM bake in the very specific syntax needed’ level of AI.
On the other hand, we have people turning a total lack of idea into big, time-consuming content that requires a minute to figure out how hollow it is. It’s a continuation of what I’ll retroactively call ‘asset-slop’ where people just tossed in a bunch of existing assets, but harder to tell at a glance. Saw an ad recently for Perplexity, saying you can just ask Perplexity to make video slop and post to youtube with zero effort and make $10k/month, brazen in the worst of the worst of the AI users.
All that said, Sweeney is pretty firmly in the ‘let the slop flow’ end of the AI spectrum, so screw him.
I’m not quite sure I understand what your point is.
Almost all modern game engines will potentially have AI code in. Most of them are either be managed by a company like blizzard or unity. The rest are open source projects like Godot.
If that is so then disclosing the fact will carry little if any stigma.
However I don’t see why you would suppose that any amount of slop code being published might make (for example) the Quake engine become slop.
I’ll say that on youtube, the implementation has been problematic.
There’s a channel where the creator has used a speech synthesis for long before the modern crop of AI text to speech. He still uses the same old speech synthesis, not even one of the newer models, but flagged as ‘AI’.
Conversely, blatant slop appears with zero indication. With no option to report as AI slop.
But the point is that if someone used an LLM to do some code completion, technically AI but I don’t think anyone would consider something like ‘multiply all members of this array by this other array’ to be ‘slop’ when turned into what an LLM would turn it into. There’s some point where it’s lost the plot entirely, but it is going to be relatively unlikely that folks ever give up the ‘still micromanaging, but letting an LLM bake in the very specific syntax needed’ level of AI.
On the other hand, we have people turning a total lack of idea into big, time-consuming content that requires a minute to figure out how hollow it is. It’s a continuation of what I’ll retroactively call ‘asset-slop’ where people just tossed in a bunch of existing assets, but harder to tell at a glance. Saw an ad recently for Perplexity, saying you can just ask Perplexity to make video slop and post to youtube with zero effort and make $10k/month, brazen in the worst of the worst of the AI users.
All that said, Sweeney is pretty firmly in the ‘let the slop flow’ end of the AI spectrum, so screw him.
I’m not quite sure I understand what your point is.
Almost all modern game engines will potentially have AI code in. Most of them are either be managed by a company like blizzard or unity. The rest are open source projects like Godot.