Are we allowed to like AI art?

[Ciderpress]

Haha yeah but synthesizers also do things that a human can't possibly do, and have been since the 60s. So I would have to assume that, if that study is correct, humans are also more aroused by music with lots of synths in it than say, bach.

It's just a wildly dramatic conclusion without really much evidence, and I don't even have a problem with AI music in principle. Good shit is good shit I don't care if a human made it or an AI or like dolphin noises can be pretty too, who cares.

[BradZax]

Originally Posted by Ekco [You must be logged in to view images. Log in or Register.]

>because the one i use is retarddded.

write better prompts, or write your question/task into one LLM like Deepseek and ask it to turn it into a better prompt then copy paste that into the other LLM

like for images i tell deepseek what i want, go back in forth with it then it spits out a literal page and a half of instructions that i paste into chatgpt

I thought bigger prompts just made it more confused?

[BradZax]

Originally Posted by Ciderpress [You must be logged in to view images. Log in or Register.]

Haha yeah but synthesizers also do things that a human can't possibly do, and have been since the 60s. So I would have to assume that, if that study is correct, humans are also more aroused by music with lots of synths in it than say, bach.

It's just a wildly dramatic conclusion without really much evidence, and I don't even have a problem with AI music in principle. Good shit is good shit I don't care if a human made it or an AI or like dolphin noises can be pretty too, who cares.

I bet AI djs could be super good

[Ekco]

Originally Posted by BradZax [You must be logged in to view images. Log in or Register.]

I thought bigger prompts just made it more confused?

It's a middle ground, depends what your doing and with which model

short Google search like prompts of a couple keywords and just letting it wing it will give mixed results

Google overview/ai mode if you prompt it like a Google search or ask a simple question won't even respond other than just providing 10 blue links half the time

Gemini/chatgpt/deepseek seem to work better the more instructions I pack into then prompt

"Act as a [PERSONA/ROLE]. Your task is to [MAIN GOAL]. The context is [BACKGROUND INFO]. The audience is [TARGET AUDIENCE]. You must adhere to these constraints: [FORMAT, TONE, LENGTH]. Here is an example of the style I want: [EXAMPLE]."

my prompt for Claude to check Gemini's coding work on Kaia is 1,395 words lol
https://pastebin.com/hNp39UWS

[Ekco]

[You must be logged in to view images. Log in or Register.]

Gemini 3.1 Pro is an
exceptionally capable AI model, offering a massive leap in reasoning—more than doubling the performance of Gemini 3 Pro on complex, novel logic tasks (77.1% on ARC-AGI-2). It excels in advanced reasoning, coding, and creative tasks, providing high-tier, fast, and, to some, verbose performance for complex, agentic workflows.
Key Performance and Capabilities:

Superior Reasoning: The model achieves a 77.1% score on the ARC-AGI-2 benchmark, which tests for novel, non-trained reasoning skills.
Competitive Rankings: It outperforms competitors like Claude Opus 4.6 (44.4% vs. 40.0%) and GPT-5.2 (34.5%) on "Humanity's Last Exam".
Coding and Science: It scores high on specialized benchmarks, including 94.3% on GPQA Diamond (scientific knowledge) and 80.6% on SWE-Bench Verified (coding).
Speed and Efficiency: Gemini 3.1 Pro Preview is noted for being fast at 107 tokens per second, though it can be verbose and has a 128k token context window.
Use Cases: It is designed for complex tasks like data synthesis, visual explanations, and building autonomous agents.

new model go brrr

[Ekco]