Andrew ZonenbergMar 14

How much do you think someone would have to pay redhat and canonical to introduce a no-slop policy including all upstream projects (llvm, systemd, kernel, etc)?

Like if you offered them $100M to permanently ban all of their devs from using LLMs, not merge any LLM generated upstream commits, etc. would they do it? $200M?

Show threadF4GRX SébastienMar 14
@azonenberg wanna crowdfund it?
Show threadAndrew Zonenberg
@f4grx i suspect mozilla alone would cost more than the community could afford. the rot runs deep
Mar 14 at 8:56amWeb
Show threadF4GRX SébastienMar 14
@azonenberg sure thing 😭
Show threadSchrödinger's Catgirl (Joyce)Mar 14
@f4grx @azonenberg it sounds easier and cheaper to hard fork certain projects, and offer a livable wage to some core developers, don't you think?
Show threadF4GRX SébastienMar 14
@sleepyowl @azonenberg I agree.
Show threadF4GRX SébastienMar 14
@sleepyowl @azonenberg but a few project wont be enough if we want a slop free linux distro. we need to fork the whole stack at this point :(
Show threadAndrew ZonenbergMar 14

@f4grx @sleepyowl it could go deeper.

How much slop do you think there is at cadence, synopsys, or TSMC by this point?

Show threadF4GRX SébastienMar 14
@azonenberg @sleepyowl oh no
Show threadAndrew ZonenbergMar 14
@f4grx @sleepyowl I legitimately dont know. i see people on linkedin playing with llms for rtl design but I dont know how much has been adopted in industry... given how conservative and risk-averse silicon design people are I expect more inertia
Show threadkrzysdzMar 14
@azonenberg @f4grx @sleepyowl when Cadence took part in an event organised by the Silesian University of Technology, their representative talked A LOT about AI. It wasn't just about designing hardware for AI (I think it was only briefly mentioned), but largely about using AI for RTL design.
Their roll-up banner (photo: https://www.facebook.com/photo/?fbid=1331467572105130&set=pcb.752895021113346) in job requirements listed "Podstawowa znajomość technologii AI będzie sporym atutem!" (eng. "Basic knowledge of AI technology will be a big advantage!')
Facebook

Show threadkrzysdzMar 14
@azonenberg @f4grx @sleepyowl It sounded like the management of Cadence branch in Katowice really wants their employees to use LLMs/"agents" for development of their IP cores.
Show threadOzzelot Mar 14
@azonenberg
We'll return to Core 2 Duo and go from there
@f4grx @sleepyowl
Show threadAMSMar 14
@azonenberg @f4grx @sleepyowl Most the AI for RTL presentations I've seen are pushing on slopping out the testbench and saying now we can get more test coverage in the same time.
Show threadAndrew ZonenbergMar 14

@sleepyowl @f4grx that's fundamentally the problem, most of these projects are dependent on corporate support to pay for developers and the corporate support has decided thou shalt use slop.

I don't know of another source of funding with deep enough pockets that isn't infected by slop.