Why is it always transphobia with these kinds of people... is it because they feel racism is too risky? So they need a different outlet for hating on people based on what they believe is "science"?

that “90%” of the audience was “on [Tucker’s] side” but had been unwilling to admit it.

Why are transphobes always this full of shit. Polls always disagree with this, and it cant be 'people are afraid to speak out out of fear', as there seem to be very little consequences to just being a lowkey transphobe. Hell even raging transphobes who stalk people, assault kids, and call for violence in what can only be called a life destroying obsession with trans people get defended by the people in power. There is very little reason for people to lie on polls over this.

Also dislike the focus on 'trans women' as these whole shitty arguments forget trans men exist. (This seems to be a general problem sadly, trans men, same as intersex people, or detransitioners tend to be only brought up like gotcha arguments, and not much to discuss the different problems they also face. Obv, I'm also doing that here, so I'm not immune)

I am extremely, extremely sickos.png for having all these weird little fucks find out that all this nonsense is heavily subsidised by VCs

replacing ChatGPT with a script that replies "Great, I'll finish that for you by ${Date.Now() + 86400}"

My guess would be because most people don’t understand what you all actually do so gen AI output looks to them like their impression of the work you do. Just look at the game studios replacing concept artists with Midjourney, not grasping what concept art even is for and screwing up everyone’s workflow as a result.

I’m neither a programmer nor an artist I can sorta understand how people get fooled. Show me a snippet of nonsense code or image and I’ll nod along if you say it’s good. But then as a writer (even if only hobbyist) I am able to see how godawful gen AI writing is whereas some non-writers won’t, and so I extrapolate from that since it’s not good at the thing I have domain expertise in, it probably isn’t good at the things I don’t understand.

>50 min read  
>”why company has perverse incentives”
>no mention of capitalism

rationalism.mpeg

Others were alarmed and advocated internally against scaling large language models. But these were not AGI safety researchers, but critical AI researchers, like Dr. Timnit Gebru.

Here we see rationalists approaching dangerously close to self-awareness and recognizing their whole concept of "AI safety" as marketing copy.

shoutouts to @dgerard@awful.systems and @mawhrin@circumstances.run for reminding me that Steve Klabnik being really shitty isn’t a temporary state of affairs

Oh Abundance, the repackaging of Reaganomics by liberals to court conservatives, is having a conference and has booked right wing speakers? Nobody could have predicted this

your periodic reminder: don't listen to Garry Tan, unless maybe he's doing a tequila review

Putting this into the current context of LLMs... Given how Eliezer still repeats the "diamondoid bacteria" line in his AI-doom scenarios, even multiple decades after Drexler has both been thoroughly debunked and slightly contributed to inspiring real science, I bet memes of LLM-AGI doom and utopia will last long after the LLM bubble pops.

Indeed great piece, good to document the older history of that stuff as well.

Does somebody have a rundown or something on DJB? All of the tantrum throwing has me confused over what his deal is.

I love it giving the temperature in Europe. Down to a decimal, even.

The blatant covering for the confabulated zip code is some peak boosterism. It knows what an address looks like and that some kind of postal code has to go there, and while it was pretty close I would still expect that to get returned to sender. Pretty close isn't good enough.

Procreate, a digital painting software for iPads, has not just avoided AI, they even have an anti-GenAI statement on their website.

Gonna cheat a little bit and put one-woman consultancy firm/personal blog deadSimpleTech up as an example. The sole member is Iris Meredith, whose involvement begins and ends at publicly lambasting AI's continued shittiness.

Where the fuck has that guy been for 20 years? I've seen that happen many times with junior programmers during my 20 years of experience.

The trigger for activating the backdoor in Gayfemboy is the character string "meowmeow".

Whisper meow meow to your femboy to get access to his backdoor... is this malware the blackhat equivalent of a shitpost?

Such sofisticated methods, and then they drop crypto miners.

Ultra trivialization is the blood of tech bros. Everything is so easy and everyone else is just so clueless. Boy met computer through hello world and thought that’s the pinnacle of computer science.

How long til some bright eyed influencer dev tries to take on JPL and MISRA? It’s always some middleware garbage these startups make and never a QNX competitor.

“The biggest problem is that it used to take about 50 pages to prove the system works, like a scalpel, but today… that could take 10,000 pages,” Kaminski explained. “And you need to make 10,000 pages every time you make a new version. It’s just really, really difficult to make so much proof as products become more software and AI-driven.”

Methinks that scalpel is a tad simpler than insulin pump, and considering that sometimes, there are unpatchable vulns found in the latter, i'd really prefer not to see vibecoded driver for these kind of things. Would match nicely with vibe-lawyered compliance docs

it's almost as funny as when one time chatbot told vibecoder to learn to code

Out: 'getting paid in exposure'

In: 'when you are done, just send your invoice to chatgpt'

ah yes, that great mark of certainty and product security, when you have to unleash pitbulls to patrol the completely not dangerous park that everyone can totally feel at ease in

(and of course I bet the damn play is a resource exhaustion attack on critics, isn’t it)

really has turned into Scientology