Welp here we fuckin go. GPT-4 is launched.

Back in the run-up to this release, four whole days ago, a bunch of people were talking over on the birdsite about their reactions to the pronouncements of those who'd gotten beta access, and MSFT's own discussions of what they believed it was going to be like.

Everybody in the gpt-4 trend was talking about how much faster it's reported to be, how many more things it can supposedly be applied to, all while blithely glossing over the fact that MSFT itself said that it is *Still Very Often Wrong* (

And that? That is the problem with our technoculture, in a nutshell.

And today, as seen in the image below, we have have these just… incredible scenes, live from OpenAI's GPT-4, where, when Chris Nicholson asked GPT-4 to give him a syllabus and mnemonics for learning Spanish pronunciation, not only did GPT-4 *Get Most Of The Pronunciations Wrong*, but— and I genuinely do not know which of these is worse— *the New York Times Reprinted The Exchange Without Calling This Out*.

I just… Holy shit y'all.

And then add to ALL of that the fact that MSFT just laid off their entire "Ethical A.I." team, in favour of their increasing the workload and making increasingly nebulous the remit of their "Office of Responsible A.I." And they did this right before this the GPT-4 release announcement went wide, specifically citing the pace of development as the reason they were laying their ethicists off (

Let me say that again:

Microsoft is laying off their whole "ethical A.I." team specifically BECAUSE they are increasing the range of integrations of their "A.I." tool throughout their *Widely-Used-By-The-U.S.-Government* product ecosystem.

…Which tells me they woefully misunderstand both "Ethics" and "A.I."

I mean this is damn near beyond parody at this point. If I were going to craft a list of shit to absolutely NOT DO in a time like this, it would look pretty much exactly like what OpenAI and MSFT are doing, right now.

Things are about to get real bad, real fast, and you can feel free to quote me on that.

But, again, and 'twas ever thus, we don't even really know let alone agree on what we're asking "A.I." tech to do and be, so it's honestly no surprise that we can't adequately design it to meet our un- or ill-defined ends.

Show thread

@Wolven Like this is the point in the scifi movie where the audience is going "nooooooo, don't do it"....

@fifilamoura @Wolven Between Microsoft firing that ethics team and Altman tweeting that the industry needs more regulation, the vibe I'm getting this week is "stop me before I kill again"

@misc @fifilamoura @Wolven the issue here is... The audience is cheering you to keep going (without realising the audience is the potential next victim)

@Wolven I guess poets' jobs will be safe from chatGPT because this thing has no fucking idea how to rhyme

@Wolven "grassy ass" is kind of cracking me up because there is a very specific set of people who do pronounce it that way and mean it, at the very least, as a microaggression. The idea that the chatbot thinks that is how its actually pronounced is so far beyond "its a bad machine" or "its a learning model beset by bad influences" but stuff like that does kind of show it becoming a twisted mockery of its bad influences. Not an automaton capable of action, but more like a dick pick on a cell phone, sculpted in marble. As you put it, beyond parody.

@Montaagge @Wolven Oh well, seems harmless, it's not like Duolingo is building new features on top of this. *puts finger to earpiece*

@misc @Wolven the good ole "I want to be the most obnoxious piece of shit in Cancun" course. I'm pretty sure they already offered that one tho.

@Montaagge @Wolven Maybe it's the fact that my first language is a language almost no-one speaks. But I honestly appreciate any attempt visitors make to speak my language. Personally, I like trying to pronounce words right, and it's awesome when people do. But seeing this as a microaggression seems to me, as a non-English native speaker, like an extreme case of over-sensitivity.

@Schouten_B @Wolven you're right, that probably is because you have no idea what you are talking about.

@Montaagge @Wolven Ah yes, self confidence, of the superior self-righteous people couples with passive aggressive insult. That's how you know you're talking to someone on the alt-right or alt-left. 🙂

@Schouten_B @Wolven lol, what the hell is the alt left? I've been an anarchist since 2006

@Schouten_B @Wolven "everyone who disagrees with me is a dangerous radical" grow up kid

@Schouten_B @Montaagge Hey. I genuinely don't care which of you does it, but the next one to reply to this shit needs to remove me from this conversation. Bye.

@Wolven I was thinking today about how a bunch of anti intellectuals burnt down the Library of Alexandria and how unleashing these poorly programmed AI's into the internet is much the same, except rather than burning it down, they're going to bury it in bullshit.

@samhainnight @Wolven After 20 years of human-generated BS, this is just cake frosting.

@samhainnight @Wolven

“AI can run round the world before the truth has got its boots on.”

@samhainnight @Wolven The AI doesn't know what's valuable for human yet - if we don't tell it we're safe.

From the top it looks like humans can protect humans. That's a new plot in a superher movie and Batman is upset he's gonna lose his franchise.

@Wolven Re: "I genuinely do not know which of these is worse”

The New York Times is very much worse here, if they did not call out that all of those pronunciations are incorrect.

@peterbutler Yeah? Maybe? but the GPT integration is going into literally everything microsoft builds, and is apparently being used as the backbone of some new Duolingo tools, so I honestly just don't know that even them stepping it up would have dinged this. It's still real damn bad tho.

@Wolven I expect higher standards of ethics from journalists, (opposed to snake oil salesfolks. ;)

@Wolven What is the bad part? I guess it's bad for Microsoft when their products fail their everyday business users in a million hilarious ways, but that's about it.

@craig no, the bad part is the ongoing and accelerating crisis of knowledge and trust being precipitated by openai ans meft deploying these tools to a public who will uncritically trust them, up to and including wothing government agencies.

@Wolven Most of the things I see about GPT etc. are about its failings. But the fact that the human at the NYT lauded its Spanish pronunciation tips without checking suggests to me that the decline in trust is well-founded.

@Wolven I thought today about what the future might look like when we get all our info from ChatGPT, and how it would destabilize and continue to create a rift between totally normal and logical people.
I shrugged it off as just me overreacting before I saw this, but it's supporting the idea that in the future, we're just going to accept whatever the algorithm tells us is true.

Did they train it on Peggy Hill's spanish lessons?

@Wolven The mnemonics are hilariously bad. How is it possible that the New York Times did not notice that?

@erik @Wolven Even if they would, they don't care anymore for reasonable reporting.

@Wolven People need to understand the difference between a language model and a knowledge model and that you use them differently. Right now people think it's an everything model.

@droidboy yep, and when openai and MSFT constantly sell themselves as that and no one in the press calls them on the difference, but in fact feeds the hype, which more people then buy? Yeah, problem.

@droidboy @Wolven I still haven't seen an answer to the question "What's the difference between a language model and a Markov chain text generator?".

@Wolven Don't worry, nobody expects US citizens abroad to speak any foreign languages and if they do, people are so surprised they won't call out mispronounciations. Maybe that is why the NYT did not call out those mistakes. ;)

@Wolven The spanish pronunciation is so funny! Apparently there are no hispanics working at the New York times, or they have an accent like Arnold Schwarzenegger! Then it mostly fits 😂

@Wolven We went to college to learn to think and become creative, to not have to be mechanics or factory workers or do menial office work, to not be replaced by robots or automation. Why do I feel like I've seen the grim reaper? Or seen then load the A-bomb into the Enola Gay?

#author #writingCommunity #writers #gpt4 #ai #BoostingIsSharing

@sfwrtr @Wolven the thing I ponder is when automation wipes most workers obsolete. The rich will have to create a market since those workers won’t have funds to buy their stuff. That’s when the consumer ceases to be the product. What’s the ultimate consumer-supplier relationship which can be sustained look like?

@Wolven "It has learned to be more precise."

Yeah. Back in 'O'-level chemistry, I leaned the distinction between "precise" and "accurate".


For fuck's sake, let's just give it access to the nukes and the drones with guns and get this over with. It's not the Apocalypse that's killing me, it's the waiting around!

@Wolven This is some "Brad Pitt speaks Italian in Inglorious Basterds" levels of bullshit.

@Wolven hype gonna hype. I am kinda tired to see same thing over and over again.

@Wolven Probably important/relevant to note here that most of the E&S team was not laid off but its employees moved into individual product teams. Although the impact of that may be similar, hard to tell.

@Schouten_B What I've heard from those who were on said team was that most of the team was laid off, and then of the seven left many were laid off, and some were redistributed, after they were all essentially told they were disposable. So it's not great.

@Wolven Appreciate the heads up. #KhanAcademy announced their new AI tutor this week. They're using this same AI. Yeah, we'll be avoiding that, #SalKhan. *sigh* #MSFT #Microsoft #GPT4 #grassyass

@Wolven grassy ass is usually a derogatory way of saying gracias. People I have known that are racist towards the Spanish/Mexican culture say it that way. I don’t even speak Spanish, but I know all of those are wrong. Should we coin #grassyassgate for when GPT4 implodes?

@Wolven, hey. If they want to hasten the collapse of the digital industry by dumping key infrastructure onto a predominantly faulty mechanism then at this point I can't wait for the 'find put' section of all this 'fuck around'

Sign in to participate in the conversation

We come here in search of a place to express our thoughts outside of the direct control and surveillance of unaccountable, mega-corporations. There is no common theme that binds us other than these being the bonds we've chosen rather than those that have been chosen for us.