Thread 107135870

>*writes up its own tools to shore up its weaknesses*
in this case because it sucks at reading images it can write its own tools to analyze them and analyze the bounding boxes from it
This is a form of AGI even if you don't like it

>>107135870 (OP)
I double dog dare you to define intelligence

It doesn't write anything it copies code

>>107135870 (OP)
claude models can't follow prompts or even complete basic tasks, you swap from their models to gpt5-mini or something and the success ratio goes through the roof, I have no idea why people even consider the shit that anthropic is pushing out worth looking at

Code With Claude

C:\print"Hello world";
goto C:\

>>107135870 (OP)
Umm no sweaty akctshually it's just a stochastic parrot markov chain next word prediction autocomplete that cannot think any more than a 70 IQ nigerian can.

>>107135870 (OP)
>This is a form of AGI
>AGI
>Artificial General Intelligence
>Not general
>Not intelligent
>This is a form of AGI
I like how words simply don't mean anything to LLM slop consoomers. It's quite appropriate.

>>107137740
>I double dog dare you to define intelligence
The capacity to autonomously model and predict reality. In other news, "AI" doesn't exist.

>>107137806
>cannot think any more than a 70 IQ nigerian can.
A 70 IQ Nigerian is more capable of thinking than your favorite LLM just by virtue of being able to think at all, no matter how basic and primitive his thoughts happen to be.

>>107135870 (OP)
> agi
> literally no intelligence, let alone general

>>107138207
The ostich knew nothing good was comming out of being herded into the fire squad

>>107137778
You can paste up to around a 20K LOC codebase into Claude for analysis. GPT you can't post more than like 2 pages at a time.

>>107135870 (OP)
>>107138223
/thread

>>107138347
You can if you upload files i think

>>107138347
rag, tool calls to read and search code, there are multiple approaches, latter is not going to work with claude because it's bottom tier garbage