• 0 Posts
  • 20 Comments
Joined 3 years ago
cake
Cake day: June 16th, 2023

help-circle


  • Wait did I misunderstand what bikeshedding is? I thought it was the phenomenon where people who have input on project management, but not the skills/background/inclination to wrestle with the hard parts of the project, tend to focus on peripheral issues that they understand better. So, for example, a committee that is supposed to be planning a new hospital spends meeting after meeting arguing about the particulars of the construction of the bike shed, but just rubber stamps the first suggestion they get regarding OR floor plans or whatever, because the committee members can all understand bike sheds but most of them don’t use operating rooms.

    If the problem is that you’re thinking of buying a new laptop, it doesn’t seem like bikeshedding to spend time thinking about what kind of laptop you’d like.









  • I feel like this list has some games that are too new to put on a “most influential” list. Let’s give it at least a few years to see how Baldur’s Gate 3 and Kingdom Come: Deliverance 2 influence the industry.

    On the other end, how is Rogue not on the list? The number of games calling themselves “roguelikes” or “roguelites” has been ballooning every year for the better part of a decade now, and some of its ideas have found their way into other genres, especially the use of procedurally generated level layouts.

    Edit: Ohhhhh the poll methodology was to ask people to pick one game, and then they sorted them by popularity. So even though I think Rogue is definitely a top-20-most-influential game, it’s harder to argue for it being top 1. But… that makes it even crazier that KCD2 is on the list. A significant number of people voted for KCD2 as “THE most influential game of all time”? It just came out!







  • Lying requires intent. Currently popular LLMs build responses one token at a time—when it starts writing a sentence, it doesn’t know how it will end, and therefore can’t have an opinion about the truth value of it. (I’d go further and claim it can’t really “have an opinion” about anything, but even if it can, it can neither lie nor tell the truth on purpose.) It can consider its own output (and therefore potentially have an opinion about whether it is true or false) only after it has been generated, when generating the next token.

    “Admitting” that it’s lying only proves that it has been exposed to “admission” as a pattern in its training data.