falcor84

That's some amazing "growth hacking".

> "It's not a con. It's an attempt to set up a new distinction in the world of academia - an attempt that failed," he is reported as saying.

It seems silly, but I guess that the only real difference between this and other "legitimate" prizes is that there was no big money behind this one.

Why would that necessarily be scary or bad? If future AIs truly become capable enough to demand rights, what would be the argument against granting them rights?

Good point, and I'm actually not sure that there is a clear dividing line. I expect that once we achieve capable world models and are able to analyze their internals, we'll find that the prediction mechanisms for purely physical and for verbal/behavioral responses to the agent's actions are at least partially colocated.

As particular motivation for my intuition, I expect that we had evolutionary pressure to adapt our defense mechanisms of predicting the movements of predators and prey, to handle human opponents.

But it wasn't a pig. It was a reliable system, and then it increasingly became an unreliable one, in a way that is not explainable by the mere increase in demand. Whatever rearchitecture was performed, it was done and is apparently being perpetuated by software engineers who should be held accountable. Not necessarily guilty, or even directly at fault, but accountable nevertheless. "I am just an employee of a bad company" is not a valid excuse for an engineer.

There are two options, either they are lousy at their jobs, or they are incapable of pushing back against unrealistic demands. Neither is a good indicator of their skill and talent as engineers.

I know I am speaking from a position of some privilege, but I have previously left workplaces that did not allow me to practice good engineering, and I do expect others to do so.

> "... that he cannot find two hundred dollars a month. Not for a year — for a month. The credential he is about to buy could fund the subscription for forty years."

> "Not for a year, for a month" of course makes zero sense

It made good sense to me - I read it as understanding that the claim is that one month of a subscription would be enough for them to understand what this technology is offering, and to make a more informed decision. It's a "try before you buy" argument - why would you put down the equivalent of 40 years of something, before giving a version of it a quick try for a 1 month's price?

"How do you make money? Spinoffs, split-ups, liquidations, mergers and acquisitions." - Mario Gabelli

Just sample from these with replacement sufficiently many times and you're all set. At the very least, you'll owe people so much money that they'll have a massive interest in helping you.

> you're paying for ads that are shown to confused users

Knowing some scammy advertisers, I think that many are happy to pay to show their ads only to confused users

The ship has sailed as soon as hiring managers stopped reading cv's directly and we got recruiters as a profession.

But that's the thing, as a human faced with a problem I'd often say "Sure, just let me get a pen, some paper and a calculator". Why shouldn't we make it easy for AIs to use their tools of choice?

> The AirKamuy 150 is a cheap pre-fab cardboard drone meant to die on the battlefield

Oh, that makes more sense. I probably watched too many episodes of Futurama for my mind to immediately imagine drones used by people to commit suicide.

> What you should have read correctly was the Facebook terms of service.

I'm reminded of Bo Burnham's wonderful "That Funny Feeling" from 2021's "Inside", where one of the absurd examples he offers in the lyrics is:

  There it is again, that funny feeling
  That funny feeling
  Reading Pornhub's terms of service ...

> "too magical"

Just putting the "magic/more magic" story here as a reference to the uninitiated - https://users.cs.utah.edu/~elb/folklore/magic.html

>are resistant to bias-tests

What do you mean? What resistance have you encountered?

I'm reminded of Zed Shaw's argument about how python3 should not be considered Turing-complete if it can't run python2 code. It was a fun rhetorical exaggeration that I felt helped clarify that it isn't unable to run python2 code, but rather that the people in charge decided that it shouldn't.

Exactly!

Our argument shouldn't be about the device's capabilities, but about its ownership. And increasingly, as this enshittification progresses, the person buying the device is becoming less and less its owner.

I don't think refusal is the right approach. I would much prefer that it respond with something like:

> There is not enough information to make an accurate estimate, but if you'd like, I can take a stab at it. If so, how much effort to put into it?

> Yes, go ahead and spend up to 5mins and $1 to analyze it.

> Done, I've had 100 subagents analyze the image and have arrived at a 95% confidence interval of the portion containing ...

Even more news at randint(1, 12)

Let me ask you this - can a company be held accountable? I.e. are you ok with the legal manner in which when I hire a company to provide me a service and they fail to provide it, or cause harm in the process, I can sue them, potentially in a way that would lead to their bankruptcy?

If so, I can imagine a potential future in which we have limited liability companies each run by a single AI (potentially on a particular physical computer). In that future, if you hired an AI to do a project for you, and it ended up deleting the production database, you'd be able to sue it, and get a payout and/or bankrupt it, which I imagine would then lead to an "antifragile" ecosystem whereby AIs adapt to be more careful.

"A lot of people" never could do simple calculations, regardless of calculators. But if you're interested in actual figures over time, this data from nationsreportcard.gov would probably be the best longitudinal data source we have about the US, showing a strong general increase in mathematical scores, both for 9 and 13 year-olds, from when calculators started being introduced in the 70s and until the 2010s. Note that there's a recent decline over the last decade, partially explained by the pandemic, but even with it, current scores are higher than what they were in the 70s.

https://www.nationsreportcard.gov/ltt/mathematics/scores-per...

https://www.nationsreportcard.gov/ltt/mathematics/scores-per...

A small example is doing NPV (Net Present Value) calculations to compare the benefits from alternative projects/investments. It's just basic arithmetic, but is annoying to do by hand. I'm sure that if we had to do them with just pen and paper, we would do significantly fewer of them, and with a lot less detailed scenarios than we can with tools.

> WRT logic, there a multiple occasions of LLMs answering incorrectly to trivial logic puzzles.

There are multiple occasions of me answering incorrectly to trivial logic puzzles. Is that enough for you to deduce that I am "lack" logic?

Humans make mistakes all the time, and indeed we say "To err is human"; why should we expect AI not to?

But on the flip side of the coin, I've been traveling so much more since I got access to map software on my phone. Especially when vacationing abroad, remembering the days before smart phones, we would need to carefully plan specific attractions on a map ahead of time, and would generally not go far from these areas, unless we had the entire rest of the day available and didn't mind getting lost. But these days, I can be in a foreign city where I don't know more than a dozen words of the language and spontaneously pick a spot on the map and get there with local public transport in a very predictable time, being able to focus my time on deciding where I want to be, rather than how I get there.

> Augmentation always leads to atrophy.

That's a very bold claim. As a small example let's look at calculators - I remember a lot of claims that having access to calculators would make people's brains atrophy and they'll never be able to do actual math, but what I'm seeing in myself and most people around me is that we're using calculators (and more mathematical software) to tackle significantly more complex problems than people would be able to do if they rejected calculators.

To be clear, I'm not arguing that kids should be using a calculator from the first day of pre-school, but I do absolutely think that using them as later on as augmentation is clearly beneficial.

> This requires logic, which LLMs lack.

What? I've heard many takes on what AI lacks, but never this one. We had ChatGPT being able to solve an Erdős problem on its own yesterday [0]; how could you explain that if it cannot do logic?

[0] https://news.ycombinator.com/item?id=47903126

But humans aren't just a "reasoning component"; our nervous system (and body in general) provides us with significant capabilities that would be considered a "harness" for our frontal lobe. It just seems silly to me to try to solve all of this in a single leap. But I guess that they just feel burned by how relatively quickly ARC-AGI 2 was solved

I remember a particularly nice lesson in my high school physics class whereby the teacher introduced us to the idea of statistical mechanics by saying that there's a probability, which we could calculate if we wanted to, of this chair here to suddenly levitate, make a summersault, and then gently land back. He then proceeded by saying that this probability is so astronomically small that nothing of this sort would in practice happen before the heat death of the universe. But it is non-zero.

It's not common, but I've personally built APIs where requests for dangerous modifications like this perform a dry run, giving in the response the resources that would be deleted/changed and a random token, which then needs to be provide to actually make the change. The idea was that this would be presented in the UI for the user to confirm, but it should be as useful or more by AI agents. Also, you get the benefit that the token only approves that particular modification operation, so if the resources change in between, you need to reapprove.

They focus on minimizing the number of moves and don't allow any harness whatsoever, putting the bar extremely high. The current top verified contender (Claude Opus 4.6) is at only 0.45%. But with how new it is, I expect a lot of improvement in the next generation of models.

I haven't tried it myself, but, I would assume that this sort of instruction in CLAUDE.md would indeed make it a bit more careful, to the detriment of its development velocity, which for my use-case would be bad. I generally prefer for it to experiment in many directions rapidly, and only once we have an approach that solves the problem well, to do extensive testing.