This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
July 2026: A new tranche of significantly better models
October 2026: A new tranche of significantly better models
(this repeats every 3 months and each time people complain that there was degradation because their tone, behaviour and failure modes are different each time. Simultaneously the hyperscalers get more and more AI revenue and invest more and more into AI hardware.)
Late 2026 Copilot or Word introduces an 'automated proofreading' button that shifts the mainstream white collar conception of AI from 'wtf is this popup in Adobe that wants to summarize a PDF, I don't want a summary of this PDF. I want to see every tech company sundered and razed' to 'ok this is actually quite handy'. This could've happened at any time in the last 2 years if Microsoft had a clue of what they were doing. Human blundering prevails over technical possibilities for now.
End 2026 there's a series of major AI-enabled cyberattacks that just never stops, it resembles 'Trommelfeuer' (WW1 term denoting when the artillery fire is so heavy one blast merges into the next creating a continuous roar of explosions). Websites, especially older websites, are just down all the time and people are quite frustrated they now need to pay a hyperscaler for expensive security assistance. Same with all the very lifelike, convincing, highly researched and well-planned scam calls (now in a warm, english-speaking accent). People are trapped into this love-hate relationship with AI where they have AI make propaganda art against datacentres, where the average person scarcely does anything novel without AI advice or assistance but also despises the effect it's having on society.
Early 2027 Microsoft is finally going to make an AI buddy for Minecraft to help sell a monthly xbox subscription. It'll be fun to play with and will help reenergize Minecraft's brand. People will feel proud they know more of the intricacies of TNT cannons than the bots, not realizing this is amongst the cheapest AIs Microsoft deploys. A series of AI agents emerge that can play most games at an amateurish level and be talked to. The reputation of AI begins to improve somewhat amongst the young and online, though it's highly divisive.
Mid 2027 the Goonpocalypse: AI avatar big tiddy anime girls (+ every flavour of girl and boy) to ERP with and form relationships with, huge revenue, makes onlyfans look like a joke. The key improvement over precursors like Ani or Replika is how much cheaper they are to run and stream real-time and how much more seductive their personalities are. Big moral panic. Lots of incredibly tedious 'zoomer men are losers' discourse and dating discourse. Legislation is introduced to ban them and instantly, predictably fails in a myriad of ways.
Late 2027: Massive AI-enabled FPV drone terror attack scares the hell out of people and spurs massive, ineffective netting operations across major cities. Police can be seen with sci-fi raygun looking widgets that don't do much of anything, or shotguns that work but aren't remotely sufficient. Advancements in robotics and software agents are displacing people at scale. AI reaches its nadir in reputation as people see the inevitable and can no longer look away. Everywhere they see some AI - the cameras tracking them, the algorithms watching them online, the machines making the content they watch and play, the robots working for them, the automated cars driving them around, cults driven insane by AIs. GPT4o cultists are charming and friendly compared to some of the new cults worshipping the bots that started self-modifying and prompt-altering, live and loose online.
And then by end of 2027 we get Dario's 'nation of geniuses in a datacentre' concept. Growth was not steady, it was jagged. The superheaviest nonpublic models with their slow speed and high cost were tasked with sorting data and implementing algorithmic improvements for the succeeding superheavy model, now running on a heap of next generation chips. They have been running for weeks in parallel, exploring and testing new approaches, RLing and training new models, testing them and reviewing them in depth. Medical breakthrough. Terror attacks. Industrial breakthrough. Mass deaths. Robotics breakthrough. Huge disaster. Huge innovations in all fields constantly and incessantly: Trommelfeuer. Events happen so quickly the situation as a whole becomes surreal and indescribable. Gary Marcus is banished from the timeline, never to be seen again except in tones of mockery.
I'm not so confident about specific times or events in sequence, though I am confident about a 'nation of geniuses in a datacentre' by end 2027. I will be clearly wrong if there's no 'it's happening!' by the end of 2027.
Interesting. Thanks for your predictions.
More options
Context Copy link
Well by the five wounds of Christ, it better be more useful than the current garbage of "this spelling is wrong" (no, it's because I'm not American), "this grammar/punctuation is wrong" (no, I meant to write it like that), and "do you want to say it this way?" (if I had, I would have done so in the first place).
Because if Copilot did pop up with that, I'd make sure to read through the entire fifty page document myself to make sure the thing hadn't turned it into beige garbage due to its 'helpful' suggestions with re-writing and summaries.
I've made one such auto-proofreader, I use it, it works just fine bringing up real mistakes for me to fix, hardcoded what kind of English I'm using, only a handful of false positives that I can quickly ignore. Every white collar worker I've shown it to says 'huh this is really cool, picks up things I've missed, saves lots of time' and they routinely ask to use it...
Absolutely massive own goal by Microsoft and big tech that most people's main experience of AI is a derpy chatbot that mangles documents or genericises things and not well-designed processes to solve specific tedious problems, nor the extremely flexible coding tool I used to make the auto-proofreader.
I have no idea what the hell is wrong with Microsoft but their support sucks, their help pages are useless, any time I need to solve a problem just searching online gives me better and clearer answers. It's beyond strange, it's perverse.
More options
Context Copy link
More options
Context Copy link
You can go into Word's spell-check settings and change it from American English to
BritishNon-American English.More options
Context Copy link
More options
Context Copy link
This is much more maximalist than even the AI 2027 crowd or the actual frontier labs themselves but I respect providing specific events and an end date on the prediction.
My own expectation is that none of this happens by the end of 2027 except a tranche of models that are notably better at RLVR'd tasks and the Copilot button (shorthand for AI getting more integrated into white-collar workflows). Let's see what happens in 18 months.
Superintelligence by end of 2027 is roughly what Anthropic/Dario seem to predict, so it's only roughly as maximalist as the most bullish frontier lab. The AI 2027 crowd backed out a year or two but I think they were roughly on the money the first time in their analysis. I'm not so keen on their 'centralize all US AI research and hand ultimate authority to a chosen council of experts' proposal though.
More options
Context Copy link
It sounds less maximalist than the AI 2027 crowd.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link