SamC

@SamC@lemmy.nz

• ⁨Comment⁩ on ⁨Anyone knows a good lightweight self-hosted alternative to GitHub?⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  Most of the Web GUIs are designed for interaction/collaboration between multiple people, and are massive overkill for one person. Tools like gitk/git gui are more than enough to see what’s going on graphically.

  If you want to install all the other stuff, that’s completely up to you, but a lot of people don’t seem to realise that the Web GUI stuff and command line are completely separate things, and you don’t have to install both of them.

• ⁨Comment⁩ on ⁨Anyone knows a good lightweight self-hosted alternative to GitHub?⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  If you don’t need the web gui stuff (and you shouldn’t for personal use) you can set up a git server using gitolite. Very easy to manage

• ⁨Comment⁩ on ⁨Sadly, even after Picard season 3 this mystery is still unsolved.⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  Modern TV is too tidy, with everything tied to one or two storylines and everything being wrapped up tidily with maybe one or two cliffhangers. It makes fictional settings cough Star Wars cough seem small and insuler.

  Interesting point!

  I think a big part of this is because of the internet. Nowadays, if the writing isn’t 100% polished you will get people screaming loudly about inconsistencies, plot holes, or hooks left hanging and I’m sure this has an impact on the showrunners.

  That said, I think you can have highly professional/polished writing, and still make the universe seem big and complex. The high quality dramas (Sopranos, etc.) have shown this. Not something Star Trek has every really been good at though.

• ⁨Comment⁩ on ⁨The Batshit Crazy Story Of The Day Elon Musk Decided To Personally Rip Servers Out Of A Sacramento Data Center⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  Yeah, that’s basically why I got out of IT. Too many managers/clients refusing to listen to warnings about what would happen when they did X, then blaming the techies when things went to shit.

  Because they are the “boss”, they have 0 accountability. Worst case for them is a golden handshake, and failing upwards where the cycle starts again.

• ⁨Comment⁩ on ⁨Is it dumb to create a wiki in this day and age?⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  I think choice of software (wiki or otherwise) is the least of your worries. The problem is not so much with fake data, it’s with the interpretation of the data. That’s where the bias (and sometimes manipulation) comes in. Even if you managed to moderate it well enough so that all the data was “objective”, you couldn’t stop subjectivity being a part of the interpretation.

  As an example, in most countries, certain minority groups are over-represented in prison populations. e.g. in the US, black people disproportionally end up in prison. That is an objective fact (so far as it goes).

  But based on that fact, you could interpret it as either:

  • Black people are just inherently more likely to commit crimes
  • There are systemic biases that mean black people are imprisoned more often

  How do you decide which is right when both are based on the data? (One is clearly racist, but still based partially on facts)

• ⁨Comment⁩ on ⁨When you notice Lemmy is quieter than usual, thean have a look at the Lemmy.world status⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  Or find a smaller instance. If everyone joins a bigger instance it’s just moving the problem

• ⁨Comment⁩ on ⁨Top AI Companies Pledge to Watermark AI Content for Safety⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  LLMs choose words based on probabilities, i.e. given the word “blue”, it will have a list of words and probabilities that those words should follow “blue”. So “sky” would be a high probability, “car” might also be quite high, as well as a long list of other words. The LLM chooses the words not by selecting whatever has the highest probability, but with a degree of randomness. This has been found to make the text sound more natural.

  To watermark, you essentially make this randomness happen in a predefined way, at least for cases where many different words could fit. So (to use a flawed example), you might make it so that “blue” is followed by “car” rather than “sky”. You do this throughout the text, and in a way that doesn’t affect the meaning of the text. It is then possible to write a simple algorithm to detect whether this text was written by an AI, because of the probability of different words appearing in particular sequences. Because its spread throughout the text, it’s quite difficult to remove the watermark completely (although not impossible).

  Here’s an article that explains it better than I can: kdnuggets.com/…/watermarking-help-mitigate-potent…

• ⁨Comment⁩ on ⁨⁩ ⁨⁨1⁩ ⁨year⁩ ago⁩:

  You have to be careful with any TLD. People outside the US have found themselves subject to US law because they had a US controlled domain name.

  Some ccTLDs are fine, some are not, but you have to think carefully when you buy it.