Esme Povirk @madewokherd@computerfairi.es

Computers have been beating the best human Go players since 2016. The Go world champion retired in part because AI is “an entity that cannot be defeated.”

But a human just trounced one of the world’s best Go AIs 14 games to 1: https://arstechnica.com/information-technology/2023/02/man-beats-machine-at-go-in-human-victory-over-ai

I think this news story is more interesting than it might first appear (without knowing details, so grain of salt). It isn’t just a gaming curiosity; it points to a fundamental flaw with “deep learning” approaches in general.
1/

I’ve had it.

The reason you feel you can’t effectively communicate over zoom is that you confuse “captive audience” with “engaged”.

Just cos someone is looking at you doesn’t mean they are paying attention. They’re just being polite.

You earn attention. You’re not owed it.

Because a friend found this useful yesterday:
Before tackling a problem, figure out whether it's the sort of problem where when you've solved 80% of the problem, you've solved 80% of the problem, or the kind where solving 80% means you've solved 0% of the problem.

This is especially important in security and privacy because that last 5% might be impossible.

Yes, it really sucks when an instance shuts down. Yes, it really sucks navigating interpersonal conflicts. Yes, it really sucks discovering that some admins are shitty, actually.

But, consider this: this isn't just about a migration from Twitter to Mastodon - or to fedi, for that matter.

Rather, we're all collectively having to relearn how to build and maintain communities and their infrastructure, after two decades of total reliance on corporations to (fail to) do so.

That *is* going to be a messy process, involving a lot of conflict as people figure things out. For many, it's learning these things the hard way, through trial and error.

It may be painful, but it's very much worth it in the long run. And fedi just happens to be the environment that people learn it in.

The season 1 trailer is here! Listen now - https://audio.jumpleads.zone/trailer/

Listen to Dolly, do it on purpose

Just absolutely floored that the state-of-the-art in keeping language models on the rails is "give it a really firm talking-to about staying on script" and that this approach has been publicly and embarrassingly proven not to work multiple times, and that they still just keep trying with more elaborate and forceful pep talks

Sometimes you see negativity directed against people who “collect hobbies”, that is: people who have an interest for a short while, get the stuff they need to try it, and then seemingly abandon those supplies.
This might be tied to a shame about wasting something (money, potential, time). But as a fellow collector I want to say: don’t think like that! Every interest migh cycle around and when it does you get much more fun out of it if you already have the supplies.
#art #specialinterest

Hey, how's that New Year's resolution coming along?

On a completely unrelated note, here is a completely random reminder that you can make improvements to yourself and your life whenever you'd like.

If you've failed before, then that is just practice for the next time. Learn from that failure, and try again.

You've got this. You can make you the you you want you to be.

game development is seeing an ask for "two characters carry a box together" and proposing that the box carries the characters and only the box moves as an alternative and vastly simpler solution.

PSA: If you don't have ADHD, you may be unaware how many of us keep going on & off our meds right now due to med shortages. It's not just Adderall anymore, there's no Concerta or Ritalin to be had either.

So please be patient with your scatterbrained coworkers and friends; we're trying. :) #ADHD

So OpenAI just released a detector of AI-generated text, I assume because of concerns in education / homework.

https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text/

Maybe this is good?

No, it's very bad.

They claim 26% true positives, 9% false positives. Assume 10% of submitted homework is chatgpt generated, you get the classic counterintuitive outcome of poor predictive power: if a homework is flagged, there's a 3:1 chance it's *human* generated.

This is going to cause a lot of harm. It should be immediately recalled.

@dolari It also gives us one of the greatest out of context images of all times.

ever think of the songs we’re missing out of because no one will sing about open relationships or polyamory

I activated my Facebook to reconnect with old friends after disappearing for years and my god that is the saddest website. The vibe is like those rows of old folks playing slot machines.

I’ve started muting people who use asterisks in place of letters in certain words in order to get past filters.

(Most often these words are the names of public figures or companies. This is completely unnecessary. Just be specific and let our filters do their work.)

#MastoTips