site banner
PaperclipPerfector  5d ago (text post)   18620 thread views

Culture War Roundup for the week of August 17, 2026

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.

2
Top Bottom Old Controversial Comments
Jump in the discussion.

No email address required.

By this I mean a system with basic vision, basic sound processing, basic movement control, and basic language abilities, with all of these things being essentially learnt rather than preprogrammed.

Seems like we are still missing a few of these bullets, even if we add his 50%?

No, we have such systems, for a certain notion of "basic". It can be built as a simple omnimodal Transformer. Or something like this. It was just underrated how much can be done with language or images alone, without robotic embodiment – these criteria were informed by the assumption that human-level intelligence is more tightly coupled to our mode of existence, and sample efficiency would suffer catastrophically if we just, like, pretrained on a large web corpus.

Liang Wenfeng explicitly says that he won't bother with embodiment because there's a more fundamental problem within the current paradigm:

With the current generation of AI technology, if you can describe a problem very clearly and give it complete context and instructions, it already surpasses humans. But there is a condition, a prerequisite: you give it the complete context and the complete instructions. And this prerequisite is very hard to meet.
For example, today’s meeting: we have a very long and rich context. Perhaps each of us has decades of context. AI doesn’t have that. What AI can have today is the ability, within a limited context, to do better than humans. But it still cannot replace humans. What’s missing is continuous learning. Because humans can continue learning.
… But if AI had continuous learning — if it could, like your employee, come into the company and learn for two months — then it could replace anyone. So the next step is still missing “learning to learn.” We can understand AI development as a staircase. Last year’s step was CoT — the chain of thought. We found that by using chain-of-thought reasoning, we could push intelligence higher. By letting AI think for itself, the ceiling rises and AI can do more. We crossed one step. This year’s step is Agent. We found that with agents, even more things can be done: the range of capabilities expands, and the upper bound of intelligence rises. Why is it a staircase? Because every later step builds on earlier ones. Agent uses CoT, and CoT uses the previous step — the language model. So no step was wasted.
Thus, the development of AI, the direction of intelligence, is traceable. This year’s step is Agent, but Agent will also eventually reach the end of its step. When it solves all solvable problems, it still won’t replace your employees — it will just have reached its ceiling.
Like CoT: after CoT reached its ceiling, it already surpassed the best humans at Olympiad math and programming. But it stopped there; that technology didn’t reach AGI. See, the direction of intelligence is traceable.
… Where we stand now, at the Agent step, we can see the next bottleneck: continuous learning. The next problem to solve is how to make continuous learning work. This is visible, and relatively clear. It’s the obstacle right in front of us. You must cross it, and there must be a way to cross it, but it will take time. After continuous learning, we may arrive at a singularity. That singularity is: when the model can continuously learn, it can do everything humans can do. It will be able to develop its own next version, conduct research itself, and create the next, more advanced AI model. So it will reach a singularity — self-iteration. But this singularity is not truly a singularity; it’s also a gradual process.
The process may be a long, gradual shift, not a sudden mutation. But habitually, people call it a singularity, because earlier prophets thought there would be a singularity. But actually, it’s not a singularity — it’s a continuous process. And after this step, I think the next is embodied intelligence.
This is our speculation — our view of the timeline: first solve learning-to-learn, then reach the self-iterating intelligence singularity, and only then embodied intelligence. After embodied intelligence, AI enters the real world: it can do housework for you, care for the elderly. We think this is an ideal roadmap. Everyone has different views; there is no right or wrong. We just think this roadmap is the easiest. The reason is that each step requires very little that is new. With this roadmap, we don’t have to work overtime. But if the roadmap were reversed — say, embodied intelligence first — then you’d be doing very hard, exhausting labor. We don’t want a roadmap like that. We want to do it the easy way.
If we first solve continuous learning, then the self-iterating singularity, then embodied intelligence, the path is easy. Because later you can use earlier technologies to help develop later ones. After the singularity, embodied intelligence no longer needs to be built by humans — the model itself will produce it. So this answers the question about our long-term goal.

We've made 1 million token context easy and cheap. With LLMs simply greping over a codebase, writing their own memos and using RLM-like tricks it's not hard to have them operate over many millions of tokens continuously. Then there are techniques to compress a given context into a higher-density "cartridge" prefix. Given strong priors from pretraining, this can functionally substitute for most of true continuous learning, in the sense that agents will be able to do long-range tasks well beyond the pretraining distribution. I'm pretty optimistic about the trajectory here. Robots will continue to develop in parallel for a while but that's just because this is "the easy way". We could merge it already.

Or something like this.

Sure, but does that thing talk? It's somewhat reasonable to argue that we have those things individually, but your guy seems to be expecting all of them as a unit?

Also 2011 + (8*1.5) would be 2023, at which point it would have been a much less reasonable argument...

The most promising systems are literally post-trained VLMs, it's really not hard to get them to talk again.

Man, do you really want to quibble about failing to nail 2026 or 2027 from 2011? We could have trained GPT-2 in 2005 if researchers had a bit more taste. Had Americans been less lazy, they wouldn't have needed Alex Krizhevsky to figure out how to use gaming GPUs for training AlexNet in 2012 (prior art was Romanian, in 2011, by the way). Timelines of exponential progress are very sensitive to the exact exponent.

What is this place?

This website is a place for people who want to move past shady thinking and test their ideas in a court of people who don't all share the same biases. Our goal is to optimize for light, not heat; this is a group effort, and all commentators are asked to do their part.

The weekly Culture War threads host the most controversial topics and are the most visible aspect of The Motte. However, many other topics are appropriate here. We encourage people to post anything related to science, politics, or philosophy; if in doubt, post!

Check out The Vault for an archive of old quality posts. You are encouraged to crosspost these elsewhere.


Why are you called The Motte?

A motte is a stone keep on a raised earthwork common in early medieval fortifications. More pertinently, it's an element in a rhetorical move called a "Motte-and-Bailey", originally identified by philosopher Nicholas Shackel. It describes the tendency in discourse for people to move from a controversial but high value claim to a defensible but less exciting one upon any resistance to the former. He likens this to the medieval fortification, where a desirable land (the bailey) is abandoned when in danger for the more easily defended motte. In Shackel's words, "The Motte represents the defensible but undesired propositions to which one retreats when hard pressed."

On The Motte, always attempt to remain inside your defensible territory, even if you are not being pressed.


New post guidelines

If you're posting something that isn't related to the culture war, we encourage you to post a thread for it. A submission statement is highly appreciated, but isn't necessary for text posts or links to largely-text posts such as blogs or news articles; if we're unsure of the value of your post, we might remove it until you add a submission statement. A submission statement is required for non-text sources (videos, podcasts, images).

Culture war posts go in the culture war thread; all links must either include a submission statement or significant commentary. Bare links without those will be removed.

If in doubt, please post it!


Rules


Recommended Posts And Communities

Recommended Realtime Chats