Even when the goal isn’t misinformation…

Removing filler words, personal and colloquial terms, replacing words for more appropriate ones according to an “internet search”.

Who would be subjected to that willingly? Why even use a real human in the process at this point?

We used to give our cycles for free to help probing the universe and understanding the building blocks of life

Nowadays it’s about selling them to assholes ruining the world.

Well, kinda glad it ends up not profitable enough for them.

The interviewer in the article had the same reaction I had to this sentence : today, gen AI is very close to be the “cool thing five years ago”. Rejecting it is not rejecting the new thing anymore.

We’ve already seen how it does, we’ve been suffering its consequences, and we hate it.

I’m way before gen Z by the way. If you can’t get young people on your side, maybe you’re not the visionary you want to be, Sam.

Please do. It sounds like the perfect plan for reddit to finally sink to complete irrelevance.

The problem is that you start thinking the robot tentacles are a neat tool and that you’re a centaur.

But before long the tentacles take over and you haven’t even realized you’ve become a reverse centaur.

Source : the ancient wisdom of Spider-Man 2 (2004).

I don’t know, Jewish tradition has golems. They’re basically human-shaped machines that are powered by the words written by their creators.

Though golems are supposed to be mute, and LLM are anything but. Maybe that’s why AI is bad.

I’m sure we can find something written millennia ago that describes precisely why we shouldn’t use gen AI now, we just have to look for it.

So you can’t oppose outsourcing your brain power to a megacorp that will take zero responsibility for any consequence of its spicy autocorrect…

but you can oppose outsourcing your brain power to “the machine” because it’s against your beliefs.

I mean, if it works, yeah sure, call it a religion.

A way less extreme fee was once proposed for steam greenlight, the process that was once supposed to filter new games released on Steam.

Principle was the game studio would present their game (in various states of completion) with a video, a couple screenshots and a blurb, maybe a demo (rare) and Steam users would check their page and vote on them if they wanted it on Steam.

For lots of reasons, it didn’t work. Like now, one problem was that no barrier to entry meant anyone could half-ass a page for their asset-flip or even other people’s work, eclipsing real games that had a hard time being approved (even a few that had a reasonable following outside Steam already).

At some point Steam tried going the fee-to-entry route, even making it a payment to charity so they were not accused of making money from the situation. It was way less than 150k, and it was already seen as rather unfair. Especially for people living in countries where the average wage was lower than in the US.

Greenlight finally imploded because it made nobody happy, and we ended up with the current situation of “anything goes, and the algorithm will recognize real games over the crap”. It’s always been a shit show, and it only became worse with genAI slop.

Regarding the fee you proposed, something that high would make it basically impossible for an indie to get in there without a publisher deal, and even then would make their relationship to said publishers even more one-way, if by chance they get one to absorb the risk. Have you read about small studios being conned out of their work by their publishers? Prepare for more.

They’ve “cracked the code” and need other cynical assholes like them to praise their genius. And possibly buy their training course.

Seriously. The guy’s not even trying to make slop entertainment, he’s gloating because he made an empty slop teaser. Just a shitty mascot that fakes a story beginning and never delivers. Because it’s enough to reap some monetisation.

Wouldn’t be surprised if he was partly faking how well it works too. That’s like 99% of that slop content entrepreneurship.

I’ve never really cared about the foldable fad, so I was not aware that they apparently have terrible durability.

Well okay, I’m sold, please let me spend a freaking fortune for one of those.

In the end LLM are just predicting what is statistically the most likely follow-up for a sentence, according to what they’ve been trained on. That’s how they work.

For this they cut sentences into bits called tokens. The text they are working on is a sequence of tokens, and they predict the next one.

A token could be a whole word (generally if it’s short and usual), often it’s a syllable, and if you’re trying to use uncommon words/names they might even cut them down letter-by-letter in part to try and fit them into something they can swallow.

Of course tokens would be very language dependent, and some languages just use more characters/longer words and sentences than others on average.

You know how scientific this whole shit is when half the system prompt is “Wait, read the instructions again” a dozen times.

And It’s still not actually following this mess in the end.

Debatable. Anyway, the very next guy after the very next guy managed to be infinitely worse than both. What was his name again?