If AI takes off, there will be nothing to adapt to. Anything a smart (and motivated) person can do, a smart (and motivated) AI will be able to do faster and cheaper. We might end up permanently retired and just spend our time entertaining ourselves.
id still argue that (human) education and intelligence is not worthless even in this world. learning is fun, a joke is often funnier the more clever it is, movies are better when they are well thought out, reading is more interesting when you can acknowledge the parallels to things youve read before, etc...
still, i think getting ASI to distribute itself properly is going to be harder than people give credit. there is still a lot of problems to solve. i dont think im going to wake up one day and we are shuttling people to jupiter and back.
Who was it that said he'd never let his education interfere with his learning? I'm in favor of learning, but what we think of as education isn't particularly conducive towards that.
I feel the same way about education, but as someone who received an education, I'm biased. I like knowing things and understanding things.
Studies on education and well-being don't really validate us though. They generally find a positive correlation between education and happiness when you don't control for other factors. But better educated people are usually from a higher social class, have better income, are healthier, are less likely to have been abused, etc. When they control for these factors, the correlation often goes slightly negative.
(If ASI does all the labor, education would no longer cause better income or health, so we want to remove those confounders.)
There are a number of proposed explanations for this, and it doesn't prove causation. But the data isn't on my side.
Also, I think there are two kinds of learning, easy learning and hard learning, and most people will only do easy learning without extrinsic motivation.
Easy learning is fun. It's why there are so many edutainment videos on YouTube and TikTok. Learning cool facts, or reading an educational book (without doing exams or problems) fall into this category.
Hard learning requires more effort. An example would be learning Cal II. You have to study and practice, and you likely won't get it right away. People do hard learning all the time, but usually because they have to.
Before online courses were a thing, friends or family would sometimes tell me they wish they could go back to college and take some courses they were interested in. Now we can, for free, at our convenience. But they don't. We'll watch endless edutainment videos on YouTube, but we'll only enroll in Cal II if we think we may need it.
If future AI can do everything, the extrinsic motivation is going to disappear. People will naturally tend towards easy learning. Maybe that's good enough. Easy learning will help us get funny jokes, recognize references, and appreciate good movies. But you don't need school for this type of learning. It's everywhere, and people seek it out because it's fun.
> still, i think getting ASI to distribute itself properly is going to be harder than people give credit
I'm guessing you mean getting AI to allocate resources and distribute wealth fairly without employment? Assuming we don't get a doomsday scenario, I see two likely possibilities. If a democratic government has control, people will vote to give themselves wealth. We end up with something like UBI, and people vote with their wallets to decide what the market (controlled by AI) should produce.
The other option is we get a dictator. With ASI and robotic armies, the factors that normally protect us from dictatorships would be gone. Currently, if the president wanted to declare himself a dictator, he would need support from a lot of people. At the very least, the military would have to back him. And you need your citizenry to be productive. They make all your food and machines and services. But if the army is composed of deadly autonomous drones instead of people and one person has authority over the army, they basically have absolute power and they don't need other people for anything. The factories keep running. The drones gun down anyone who resists. We're left at their whim, and we better hope they aren't cruel.
I'd argue that even endless leisure is better if you're smart and educated. Gives you better decision making (even if the AI makes your decisions, you still have to make the meta decision of which questions to ask the AI) and something other to do than just mindless hedonism. More art exhibits and intellectual conversations, less drug abuse and trash food.
1. This goes against the prevailing view of neuroscientists. Quantum indeterminacy is generally thought to be irrelevant to brain function.
2. Quantum randomness is the very opposite of control over your behavior. Instead of doing what you want, your neurons randomly misfire sometimes. That describes an illness where you lose control over whatever is misfiring. For quantum effects to represent your will, they couldn't be truly random (nor even statistically random), and that goes against everything we know about quantum mechanics.
According to the article you linked, there is no evidence at all showing OpenAI plagiarized Buckmaster and Alpoge. Buckmaster just speculated that since they used OpenAI's product, it's possible OpenAI looked at their conversation. Which OpenAI flatly denies.
And then your article says:
> On 13 October 2026, Scientific American published an article surveying several mathematicians stating that "the resemblance between the two groups’ methods now seems merely skin-deep."
But the AI companies didn't gatekeep knowledge. They released all their math results.
Unless you mean they're charging money for a product that contains knowledge. But that's not gatekeeping, just like selling a book isn't gatekeeping. Gatekeeping is when a group tries to prevent others from participating in a field or activity.
> But the AI companies didn't gatekeep knowledge. They released all their math results.
There are some important fields of math that were conspicuously missing. In particular, cryptography.
There are obviously good reasons for being more cautious when dealing with cryptography results in math, but I'm not going to believe they released everything.
> Unless you mean they're charging money for a product that contains knowledge.
They have internal models that the public cannot use (at least not yet).
Selling a book quite literally is gatekeeping, since the idea is that you're prevented from engaging with the work without paying for it. Even with libraries it was society who paid for it in order for the commons to get access, but payment was nevertheless needed.
It just turns out that the Chesterton's gate is useful and people tend to think that people should be paid for their labour, since people kinda need a roof over their heads and food on their table.
It's not gatekeeping unless you're using your own esoteric definition that's different from how everyone else uses it. Words get their meaning from how they're used in the wild, and dictionaries are descriptivist, not prescriptivist. Anyone is allowed to buy the book, so it's not considered gatekeeping, even if some people may not be able to.
If you define any obstacle to an activity as gatekeeping, then literally everything can be considered gatekeeping and the term loses all meaning. Some people can't read, so offering the books for free would still exclude them. If you include an audio book, some people still can't understand the language. And both charging and not charging money for products would simultaneously be gatekeeping under that definition, since not charging places more financial burden on the producer. Not everyone can afford to print books and give them away for free.
> Gatekeeping is when a group tries to prevent others from participating in a field or activity.
And I gave a way in which charging a price does that.
I am open to the idea of being incorrect in this case, in which I apologise for misunderstanding, and concede that I'm missing the nuance but in my defence, I'm not a native speaker.
> Words get their meaning from how they're used in the wild, and dictionaries are descriptivist, not prescriptivist.
But I didn't explain that clearly. What I mean is that definitions are always imprecise because language is imprecise and dictionaries (or people) try to keep definitions reasonably short and simple. Dictionaries are trying to describe what the words really mean in the wild. They aren't the source of that meaning.
For example, if we look up Merriam-Webster's definition of "game (noun)", none of the definitions really work for a game like Solitaire without also including an activity like reading a book.
The two relevant definitions are
1. a physical or mental competition conducted according to rules with the participants in direct opposition to each other
2. activity engaged in for diversion or amusement
(there are a bunch of other definitions as well, like animals you hunt)
The second definition is too broad and the first clearly excludes Solitaire. If you were deciding what a game is based on the dictionary definitions, you'd either think watching TV is a game, or that Solitaire isn't. But that's not the case.
Charging money for products isn't considered gatekeeping. So from a descriptivist perspective, it isn't, even if it seems like it should be, because the word isn't used like that. Banning a group of people from an activity is gatekeeping. If mathematicians said only mathematicians can prove theorems and correct proofs from non-mathematicians won't be accepted, that would be gatekeeping. But they don't do that.
If you want to use a broad definition of gatekeeping that includes virtually all human activity, then the response to any accusation of gatekeeping could be "well, everything is gatekeeping". That road just leads to pointless and confusing conversations.
We absolutely want politicians to listen to lobbyists in a way that is proportional to some parameter besides the number of people they represent. The should listen to who is right, not which lobby represents the most members. Self-interested groups lobby for bad legislation all the time. When the harm to society is diffuse, they don't always have a large lobby opposing them.
Certainly somebody needs to identify cases when a minority voice needs disproportional consideration because they happen to be onto something special. But it's the job of an elected official to make that determination only after they have considered their options in a proportional-to-population way.
If you try to handle nuances like that up front, and somehow select for who gets the politician's attention on the basis of some parameter besides population, you're going to end up with a government that only works for the rich, because they are disproportionally able to game such things.
> But it's the job of an elected official to make that determination only after they have considered their options in a proportional-to-population way.
I don't know what this means, and there's no way to operationalize this where "do what's right only if the right lobby represents enough members" is better than simply "do what's right".
What you're advocating is going to do the exact opposite of what you're hoping for. It will increase the power of the rich and special interests.
It's common for a large special-interest lobby to face no comparably organized opposition, even when its preferred policies harm a much larger number of people. Look at the National Association of Realtors or the National Automobile Dealers Association. These are large, well-organized lobbies that promote their own special interests and harm the rest of us. There is no anti-realtor lobby.
Right now, anyone can write to or call up their representative and argue their position. That's lobbying. You don't need to get X signatures or form an organization with at least X members for the right to be heard by your representative. Putting up barriers like that isn't going to hurt the lobbies. It's going to restrict the opposing voices.
(To be clear, I'm not saying there would be an explicit threshold where a lobby needs a certain number of members. But no matter how you implement this proposal, if it has any effect, then there is some implicit threshold where that number of supporters pushes the lobby over the edge so their position is considered.)
All the money in politics - the campaign donations and the Super PACs - exist to try to get votes. Campaigning costs money. If you start requiring a threshold of votes, signatures, or members before a political position can be considered, you are handing more power to the rich and powerful. The well-funded NAR would get their policies by default unless we could organize a bigger lobby with more than their 1.5 million members.
Requiring votes, signatures, or members won't solve this. People cannot stay informed on every little issue and every special interest group. These groups are well funded, and they will campaign for their positions. Often their arguments sound convincing. Often they're misleading or dishonest. And they win. This isn't hypothetical. It happens all the time.
You're misremembering. The first known use of AGI was in 1997, but that was a single, mostly unknown use in one paper. It wasn't until at least a decade later that the term started entering mainstream use after being independently reinvented in the 2000s. AGI just wasn't a term in the 90s.
But that's not the same thing at all. Searle used the terms "strong AI" and "weak AI" to distinguish between systems emulating humans that were "actually conscious" and systems emulating humans in behavior only. The whole distinction was bogus.
If you google the terms you will see that the meanings have shifted since then to mean AGI and ANI ... such shifts occur because people aren't familiar with the original technical meanings ... e.g., the way "meme" now means a gif you post on social media, as opposed to Dawkins' original concept of a cultural thought component akin to a gene.
It must have had shifted already by 96-97. I remember how I used it in AI discussions back then and it meant “the opposite of narrow AI”, which makes it very similar to how we use “AGI” term today. Consciousness was considered to be a potentially emergent phenomenon, not a required property.
Given that they're just using it to avoid the googling for bash command syntax, I'm not sure they'll save in the end against the 140k training examples they generated.
You can buy a $10 subscription for a month to generate the training data, then cancel your subscription. The trained model is yours to use (and share with others) forever.
Yea specialized models could also be resold in a shareware style, like even if it cost 50$ to produce you'd just have to sell 10 copies of it to people for 5$.
Which 5$ is a pretty easy sell if its useful in any way, It's pretty easy to justify a purchase if its yours forever and doesn't use much CPU so is easy to run I mean people were spending 1000$+ on mac mini setups to run local llms or run remote agents.
It wouldn't cost $10 for their lifetime use if they used cheap models instead. Mistral Small 3 24B is probably substantially better than their Qwen 3 0.6B trained model and would give you about 500,000–2,000,000 English to Bash translations. If they're a heavy user, they might use $0.30 in total, ever.
They also used Astra for the coding, which they can't get on a $10 subscription. And then there's the actual training cost.
This example is not that niche. Lots of people use human to bash.
I'd probably use a small pre trained model if it was easy to use. I use a little script right now that calls a cheap model.
Actually.. $5 would probably will last me over a year so the only real benefit would be if I didn't have an internet connection.
If it's about the latency / flow disruption, spending a few hours once could easily be worth it if the result is actually good enough to skip googling/retries.
I think OP's point remains, if you generate 140k pairs, your local model would need to run that many to offset having just used the generator (SOTA or not) model to begin with.
I wonder if another approach if latency is a concern is just to do a two shot pass with Jev (perhaps given small context you'd want one to match command, then one to match args of given command) would be an extremely fast, and cheap way to do it - rather than training your own.
Not quite. One, because you save on the initial query being sent multiple times. Two, because the reasoning will be very similar at the beginning; "user asked me to", "let's check what's in this project already", etc. you'll get similar actual output cost, but input and reasoning will be shorter.
The typical perception-to-reaction latency of an alert driver to a hazard is about 1-2 seconds. 250 ms when you're waiting for an event and know how to respond. For example, like a batter in baseball waiting to swing.
In humans, it's about 200–250 ms for a visual cue where you already know how to respond and you're ready, but you don't know exactly when it'll happen. It can be a fair bit longer if you need to identify what you see and choose how to respond. Typical perception to reaction time estimates for drivers when there's an unexpected hazard on the road are 1-2 seconds.
do you know what it is for non visual cue? like the gust of wind or unexpected jerk of the wheel from a road feature? both of those seem like more of a speed of proprioception which seems faster than visual and have way more to do with general "car control" in normal driving where the path is planned.
I would imagine it’s much less than this with muscle memory. Driving is a very habitual process. The human body is certainly able to “short circuit” its thinking and respond faster.
I mean, consider competitive video games. Humans who play a lot and pay attention respond to stimuli much faster than 250ms.
This is well studied by researchers, and 1-2s is the typical reaction time for real drivers with real muscle memory when reacting to an unexpected hazard on the road.
There are a lot of reasons people could sometimes react faster (for example, if they anticipate the hazard, or if they're just above average in reaction speed), but one to two seconds is the reaction speed we find most of the time.
The fastest human reactions aren't to unexpected road hazards. We have a much faster reaction speed in tests where you just have to click the mouse each time the screen flashes. Our reactions are fastest when you know in advance the event is about to happen. But this isn't relevant for road safety.
Your wording is kind of misleading. While the human reaction time to the unexpected is not great it's the human's ability to predict within reason what the likely next things to expect are expansive.
Like just running a first pass sanity analysis on the 1-2sec timeline fails because if it were true in practice all those idiots who screech about how normal traffic doesn't keep following distances worthy of semi trucks to the traffic ahead would be proven right as every braking event would cause a pile up. So either humans react much faster to the unexpected (not likely, we've measured) or humans have a huge "context window" for what to expect that makes the 1-2sec number not relevant in the base case.
No it's not. 1-2 seconds is what we find in practice for real drivers responding to unexpected hazards.
Your close following distance example doesn't show anything. A normal braking event on the highway doesn't require a fast reaction. If you're driving 100 km/h and you're following 1.5 seconds behind the car in front of you (about half the recommended following distance) and they brake to 80 km/h, you have about 9 seconds to slow down or switch lanes. That's plenty of time.
The risky scenario is if the car in front of you has to do a full, hard emergency stop and you're following too closely. That's rare, and collisions are common when it happens.
If you know the car in front of you is going to brake because you can see the traffic ahead slowing down, that's not fast reaction time. That's just you seeing cars slow down and reacting at a normal speed. An AI has just as much time to react to that as you do.