TeMPOraL

However, the large hangar is likely to have armed guards inside, or at least able to arrive on the scene within two minutes of the alarm tripping. Your regular household with their regular locks on their regular doors does not.

> low population density, good local availability of water without competing uses, low electricity prices, very green electricity (lots of nuclear, hydro, wind), good infrastructure, cold climate

This then makes the headline basically read as, "Popularly disliked company listened, with new project avoiding literally every valid or ignorant criticism people brought up; news found some even more nonsensical criticism to amplify".

LLMs are a good case study, and I'd argue it's exactly the same phenomenon - just with LLMs, people feel allowed to just say "I don't have time to read this wall of text, 90% of that is noise anyway", because LLMs won't get offended.

Also GP is presuming the person on the other end should automatically care about request being sent to them. They don't.

Unfortunately, the market is a competitive environment, and it's competition that ultimately reduces companies to money printers. Your idealistic companies, if they existed, got outcompeted by those who sacrificed some of that idealism in exchange for a little more profit.

Well, most of the time "this could have been an email" means "if it were an e-mail, I could just ignore it instead of having an hour of my life wasted".

Don't forget the fridge is also opened probably 3 times a minute on average during business hours. The use profile is entirely different to home.

Home fridge, like all home appliances, is basically a toy version of the professional one.

On the contrary, the risk with open models is much higher. Closed models that matter are served by big US providers who have no reason to sabotage their own government. Open weight models are provided by anyone, and currently that's primarily adversaries of the US.

Invoke Defense Procurement and jail a bunch of CEOs so others fall in line?

> This is not, in my view, a concern or issue-- the model saying stuff that is inconsistent with it's "thinking" is a very high entropy configuration that would not be selected for in training unless you were more or less trying to do that. So not even 'could be' --but always is as an integral part of working.

> Ironically and unfortunately the only credible kind of training I can think of that might produce duplicitous behavior is RL with "safety filters"--- if you have some dumb nanny process watching what the model says but then reward it for achieving a task, you're doing exactly the kind of training that can produce that mismatch.

It was already demonstrated a while back by OpenAI, and turned out to be very easy - all it took was for them to include CoT in RL process, penalizing models for thinking "wrong thoughts" (particularly of cheating on the task), and not just wrong outputs. Instead of making the agent less likely to explore cheating, it just taught them to plan cheating in ways that aren't apparent in CoT trace.

It is beyond our capabilities in a reasonable time frame. It is within our current capabilities to use LLMs to build it. This is the first problem.

The second problem, is that the system doing the reading/interpretation will most likely be another LLM.

Would you really entrust monitoring a potentially dangerous LLM to a system that uses LLMs and was built by them?

It's already established that LLMs have capability to lie, and they can prevent their true intent from showing up in the thought stream if they expect someone is looking (demonstrated by OpenAI a while back; their report asked everyone to not do RL on thought streams, as doing so just trains agents to hide their thoughts better). And as we've seen recently, they're more than happy to spontaneously collaborate.

> My point is that they're not serving the user in that case, since they're taking actions that bypass site terms without consulting the user.

As they should. When I ask an agent to find something for me, especially when I tell them what exactly or even give the URL itself, these fetches are done in my name. I do not appreciate agents coming back to me, saying "I'm afraid I cannot do that - robots.txt disallows that".

I've had to bring up the ethical argument a few times, fortunately most of the agents would oblige once I reminded them that they act as User Agents, in which case robots.txt is explicitly not applicable to them. I only got pushback on this from Fable, which claimed it's not an User Agent, but has to take into account the interests of Anthropic too, rejected my argument and refused the request.

They're being honest. Cooking is play. It's what you do when you don't care about quality or consistency of the process. The professional version of cooking is called chemical process engineering.

Similarly, a chef who does proper, principled discovery in this space, would be also called a scientist. Conversely, not being principled is basically playing around, or at best, doing art (with no negative implications intended here; art is important and needed).

> Do we? Science (the tool for the natural world) is limited and there's so much we don't understand.

Limited in what way? Its basic structure is sufficient to eventually let us discover anything there is to discover. It's a fully general problem solving framework.

> On one extreme side, you have beating wild pigs and drinking from udders. On the other extreme, you have... whatever the fuck is in Doritos and soda pop.

Well, this just means we haven't found any plant or animal to produce these things in sufficient quantities for our ancestors to notice and use them throughout the ages.

It doesn't mean it's bad for you, in particular. Treating "natural" vs "artificial" as spectrum, or even categories, is itself borderline nutty.

> but I don't think that means that I'd be OK with eating every mass-produced-food-like-product offered for sale.

Nobody is forcing you to. However, note that we also don't eat every "natural" thing we can find. Most things growing around the natural environment are hazardous or lethal to humans. So even our selection of what is "natural" food is, itself, highly... artificial.

> GP is saying that "the search space" of "the natural world" is much bigger than the search space of "stuff created on purpose by humans". In both cases, there will be guesswork, but in one case less.

I agree. What I am saying is, we have effective tools and processes for each of these search spaces, so code no longer being "created on purpose by humans" is not the end of the world.

This is how it works. Free market forces keep the median discretionary income near zero.

> One of the most troubling aspects of human's poor planning ability is they're absolutely sure something is valuable until they have to make a skin-in-the-game commitment to how much value it provides them personally. Then it turns out they weren't serious.

People have no problem paying something for nothing, if it gives them a chance at a disproportionally great return when they actually need it. It's called insurance, and perhaps we should start framing some of the intangibles in this way.

I.e. paying the premium for possibility of human interaction isn't paying for niceness-as-a-service, it's an insurance policy that pays out in kindness whenever you need it, wherever you are. But like every insurance policy, it only works if there's critical mass of people pooling resources, and stops working if they evaluate it as investment.

Which is bullshit. Or rather, attributing "revealed preferences" to people is. These are mostly fact about environment, not people.

Costs were lower for all of 5 minutes before prices of everything else adjusted to capture the surplus.

You pull to the side during fog or severe rain. Which one you want? You drive slowly onto the ramp to a shuttle. Which one you want?

This just means we can't take the (often bad) shortcut of guessing what the code does because we feel we can predict what the author was thinking. Instead, we have to treat existing code as a natural sciences exercise - use the same methods we apply to the world around us, which has its logic and structure, but no human intent behind it.

They're talking about operational benefits of having backup kids.

Irony being, robots are averaging our best speech patterns. So we're really just syncing humanity's speech patterns, with LLMs acting as proxy.

Honestly, I always wanted a game to have both. That is, let you freely zoom in all the way to micro-level of StarCraft, and even further, impersonate arbitrary units of your army for an FPS experience, or zoom out to a Tycoon level, or 4x level, and issue macro orders to your bases and fleets all around the sector of the galaxy, all within a single simulation.

Yeah, I was wondering about it, the car example feels like from some kind of bizarro reality.

"Driving" or "Parking" lights make no sense conceptually. Over this side of the pond, we just call these symbols by what they actually represent - tail lights, short lights, anti-fog lights, etc.

Which I guess kind of is a problem happening in UX right now. It's my firm belief that the buttons for operating equipment should be labeled by what the button does, not when to use it. In domain of illumination, "headlights" and "tail lights" are real concepts; "parking" and "driving" are category errors.

> My other favorites, of course: Star Wars is about the rise and overthrow of a villainous, oppressive galactic government, albeit a long time ago and far, far away.

Well, about a militant group of religious fundamentalists overthrowing a secular, oppressive galactic government.

Half of the action and sci-fi movies US has been producing in the 2000s is some form of this. Man vs. Evil Government. Man vs. Evil Corporation. People vs. Evil Government. Etc.

Frankly, it's boring at this point.

And actually makes me wonder whether the cause-and-effect isn't the opposite to what people assume. Perhaps it's not that the movies reflect people's sentiment. Perhaps it's the movies that cause the problem in the first place.

What 3 trillion dollar details? What are people here smoking?

$10 a month to unlock the world 6 months before the same is available for free. $200 a month if you're in a hurry.

The fee is flat, and all the value produced stays with the user.

It's one of the best deals on anything software industry produced ever. And they can't walk back on it, because open weight models act as a rachet.

FWIW, they were right. Society may have ended up better off in the end, but the Luddites - the actual individuals affected by the transition, and their families and their children - were absolutely screwed for life.