I read an article once that talk about this the first few layers are like a translation layers then middle is thinking and the output layers translate it back.
It’s pretty interesting actually, you can cut out the middle layers and stack them back to back and that increases the models performance without much memory increase.
If you want to try it out you can get a LLM with no tools and pass it base64 and it will understand what’s in the message
I would love to hand code way more but truth is when you have a small team of developers and they are all in on AI for coding then it’s actually hard to hand code. At a certain point the project ends up being only workable by AIs which is kinda sad. Coding is fun.
The test one I have seen too, one way I have had some luck with is to get an agent to write a plan, then another agent to plan the tests for that plan.
For this agent it is told to think of it as a graph and each “if” or function call, import etc as a new node.
They generate pretty detailed graphs this way. Seems to really help them with the “depth” issue.
Pretty sure the current proposal will already require scanning everything.
To stop the “production” you would need real time scanning of the camera when taking pictures and video and disable the ability to save if csam is detected.
Saving and sharing would still function, albeit in a slight different way.
- Saving, as before.
- Sharing, works, but due to a glitch the recipient got a blank blank image/video. He'll sure reply to let you know, but unfortunately incriminate himself in the process.
- Deleting, as before, no sorry, bugger, I/O error, can't delete this file. But support team has your address and is on their way.
because documenting your api in swagger is too much work?
reply