@drewdevault I uook forward to seeing the new code

Done, new captransfer code:

https://git.sr.ht/~sircmpwn/helios/tree/master/item/objects/endpoint.ha#L202

@drewdevault wait what's that?

@drewdevault should've gotten a psp

Scaling the Hare community

https://harelang.org/blog/2023-08-28-scaling-the-community/

Really excited about these reforms

@drewdevault scanning your eyeballs to train their pet AI on... for money (or worse)

@drewdevault Any suggestions for specific charities?

@drewdevault I work in research applying machine learning to pathology. We have a pretty big network of pathologists that are really excited at the potential for a machine to handle the tedious, and error prone tasks of evaluating tissue samples; allowing them to focus more on the parts of their job that a machine cannot do. End goal to improve patient outcomes by supporting the pathologists in diagnostics and drug development. Happy to share more of this perspective if it's of interest.

@drewdevault there are many areas where AI will make a change. It won't replace doctors and researchers but will definitively assist them.

For example, trained AI models can outperform radiologist at finding tumors. In the end a radiologist will always be there but they can spot details that will be ready to miss even from a trained radiologist.

It's also very promising for research, being able to process massive amounts of scientific literature and extract very specific details.

@drewdevault Unfortunately healing people and saving lives on a daily basis isn’t something the media are interested in.

@drewdevault Natural selection found a general intelligence (humans), so I don't see why a ridiculous pile of GPUs can't find a general intelligence with gradient descent. Gradient descent is a much more efficient optimization algorithm than the evolutionary algorithms that natural selection used, because the gradients give a lot more information. And we also have done the hard work of creating the hardware for the intelligence already, which natural selection had to do separately.

@drewdevault another aspect that often gets overlooked is that machine "learning", at least the currently often used and hyped variants of it, further entrench the past into the future – something we really do not need currently.

1/3

@drewdevault I wonder if we could create a poison pill license for various content that would force all the AI models to open up if they ingest and regurgitate stuff.

@drewdevault no, wait, they still simp for IP, as long as it's theirs. It's just the little people's who doesn't matter.

@drewdevault I saw one article put it rather nicely. AI is neither artificial nor intelligent. It's not artificial, it regurgitates knowledge that humans created by employing human labor it exploits to categorize that knowledge.

@drewdevault yeah, my existence is political, being "apolitical" is a privilege (and also means accepting the state's ideology)

@drewdevault smh that's nothing, competing services have been sharing API keys for ages without even needing you to ask or want them to :P

@drewdevault sent from Ares?

@drewdevault, I honestly hope it does.

@drewdevault Honestly, Hashicorp ruined itself. This is basically the expected reaction to their actions.

@drewdevault We really need a Hippocratic Oath or similar ethical standard for programmers.

Instead, in my years of studying software engineering at NHL Stenden, I never _once_ had a single ethics course, and Microsoft bought influence over several teachers.

@drewdevault What startles me about the modern tech industry is how many people call themselves 'engineers' and how loosely it's thrown around when ethics are at the centre of any field of engineering and at the core of any engineering program and degree, although these days you no longer need an engineering degree to call yourself an engineer. This may be a factor along with marketing and other teams driving design and decision-making over engineers, but they still have every responsibility.

Tip: work incrementally and make stuff backwards compatible and your life is much easier than it would be with a complete overhaul

@drewdevault hey @sebastian drew is doing it yay!

@drewdevault https://gothub.projectsegfau.lt

Big reason I don't like open core models is not just because they cause nonfree software to proliferate but also because it's a pretty clear promise, especially when combined with a CLA, that they will pull the rug at the first sign of a bad quarter.