nearbuy

I was replying to someone who said we should make lobbying illegal and they don't understand why it's accepted.

"Checks are written" is vague. What do you want to make illegal? Super PACs? I had just said they should be made illegal.

Lobbying is "directly advocating for or against particular legislation or regulations." Writing your representative is lobbying. Fighting bad legislation is lobbying. Any good faith attempt to argue a position on a government policy is lobbying.

Giving money to politicians or their campaigns is not lobbying, and it is already illegal for lobbyists to do so.

What could and should be made illegal is allowing unlimited political campaign donations via Super PACs. Political donations aren't lobbying.

It's worth being clear about what you actually want to make illegal because you probably don't want to ban anyone from arguing a political position.

That's an editorialized headline. What they actually wrote was that it could be used to "generate misleading news articles, impersonate others online, automate the production of abusive or faked content to post on social media, [or] automate the production of spam/phishing content" and that they are aware other researchers have the ability to reproduce and open source their results, but this would give the community some time to decide how to proceed.

They were correct.

I don't mean that Rand wrote more. I mean that her idea was different and nearly opposite. This is a short story about an artist learning to reframe their frustration with customers wanting utility over artistry as a positive. The similarity to Rand is in the first few sentences. The point is entirely different.

If you judge stories to be the same based on this level of similarity, then The Fountainhead is just the same as a dozen older stories with the artist vs the philistine theme. It was common before Rand. As T. S. Eliot said, "Immature poets imitate; mature poets steal".

There is a similar theme in both of an artistic person not wanting to compromise their vision to suit common tastes. But this goes in a completely different direction than Rand.

I'd say the same thing about two thirds of the iTunes top 100. Different people love different songs I guess.

The lyrics of the one you linked are fairly strong compared to other songs on the top 100 list.

If my math is right, assuming a mix of around 70% cached tokens, 20% input tokens, and 10% output tokens, it breaks even with the old pricing at around 130k tokens per message, or about 13k output tokens per message.

With the hidden reasoning tokens and tool calls, I have no idea how many tokens I typically use per message. I would guess maybe a quarter of that, which would make the new pricing cheaper.

> The reasoning part is not diferente from the part that goes in answer.

Exactly. And this instruction isn't telling it to skip the reasoning. That part is unaffected. The instruction is only for the user-visible output.

By the time the reasoning models get to writing the output you see, they've already decided what they are going to say. The answer is based on whatever it decided while reasoning. It doesn't matter whether you tell it to put the answer first or the explanation first. It already knows both by the time it starts outputting either.

You're basically hoping that adding more CoT in the output after reasoning will improve the answer quality. It won't. It's already done way more CoT while reasoning, and its answer is already decided by then.

It won't matter. By the time it's done reasoning, it has already decided what it wants to say.

Reasoning tokens are just regular output tokens the model generates before answering. The UI just doesn't show the reasoning. Conceptually, the output is something like:

  <reasoning>
    Lots of text here
  </reasoning>
  <answer>
    Part you see here. Usually much shorter.
  </answer>

> Answer is always line 1. Reasoning comes after, never before.

This doesn't stop it from reasoning before answering. This only affects the user-facing output, not the reasoning tokens. It has already reasoned by the time it shows the answer, and it just shows the answer above any explanation.

> See this definition of AGI (link shamelessly stolen from someone else in this comment section) from before the latest AI hypecycle started warping things.

Every definition on that page, both theoretical and operational, match my definition and not yours. Notice that none of them would exclude an AGI with an IQ around 90, provided it's intelligence is general.

> Approximately that an unaided agent must, with no outside assistance, be able to solve ~90% of the most difficult tasks that we throw at it with a ~90% success rate. It's not a precise definition but that's approximately where I stand on the matter.

This isn't your definition. How hard are these "most difficult tasks"? Can 50% of humans solve them? 10%? If it were literally the most difficult problems, they would be the ones 0% of humans have solved.

> Said race as a class would presumably be capable of meeting or exceeding my above criteria

Some might, but this hypothetical alien race does not. Do you still consider them to have general intelligence if they can merely do everything a 95 IQ person could?

> Your attempt to compare to individual humans is an error. AGI applies on the class level

According only to you. LLMs are benchmarked individually. No one runs a benchmark where Claude gets half the questions right, GPT gets the other half right, and it's reported as a combined perfect score as a class. Instead the each score 50%. (Not that I think current AIs can solve the harder benchmark problems. The point is they are measured individually.)

No one else ascribes general intelligence only to a class. You can talk to one average person (or alien), give them some tests, and determine they have general intelligence. This is how everyone else uses the term.

First, what is your definition exactly? That it must be better than the median human intelligence?

You're trying to define a term in a way that's completely detached from how anyone uses it. If we discover an alien race with an IQ of 95, people aren't going to say they don't have general intelligence.

We haven't defined an exact cutoff for what counts as general intelligence, but it has to include regular people with an IQ in the 70s that don't have a serious mental disability. If an AI can do every single cognitive task as well as a stupid person, it would have to qualify as having general intelligence if the stupid person qualified. It doesn't matter if the AI beats the median person 0% of the time, as long as it beats someone who is considered to have general intelligence at the task.

It has nothing to do with circular reasoning or my personal opinions.

You can choose to define general intelligence in a way that excludes regular people if you like, but then you'd be using a weird definition that differs from how 99.9% of people define it. Humans have general intelligence by any common definition.

They would score much worse on the private set than the public set. And they haven't done this for any of the other ARC-AGI benchmarks, so why would they do it for this one?

> Not true.

It's certainly true. By definition. If the bar for general intelligence is being smarter than the median human, 50% of people won't reach the threshold for general intelligence. (And if the bar is beating the median in every cognitive test, then a much smaller fraction of people would qualify.)

People don't have a consistent definition of AGI, and the definitions have changed over the past couple years, but I think most people have settled on it meaning at least as smart as humans in every cognitive area. But that has to be compared to dumb people, not median. We don't want to say that regular people don't have general intelligence.

That's their claim for how much less economic activity Iran (not just the government) had in total due to sanctions, not how much the government "spent on the nuclear weapons program" that could have instead been spent on desalination plants.

Even dumb humans are considered to have general intelligence. If the bar is having to outdo the median human, then 50% of humans don't have general intelligence.

It can't be 2T USD. That's about 60 times the cost of the Manhattan project in today's dollars. It could maybe be 2T Iranian rials.

> This gives it the ability to approximate multiplication problems it hasn't seen before.

More than approximate. It straight up knows the algorithms and will do arbitrarily long multiplications correctly. (Within reason. Obviously it couldn't do a multiplication so large the reasoning tokens would exceed its context window.)

Having ChatGPT 5.4 do 1566168165163321561 * 115616131811365737 without tools, after multiplying out a lot of coefficients, it eventually answered 181074305022287409585376614708755457, which is correct.

At this point, it's less misleading to say it knows the algorithm.

The author is partly right here. If those values are ints, you'll get something like this:

  sar eax, 0x1f
  and eax, 7
  add eax, edx
  sar eax, 3

You get 4 instructions instead of one because value >> 3 rounds towards negative infinity and value / 8 rounds towards zero.

And while this wouldn't apply to C++, in languages with checked arithmetic, the left shift won't necessarily set the overflow flag, so the compiler often can't use it.

Wikipedia says it "had the highest opening for a non-concert documentary since the $10.7 million opening for Chimpanzee (2012)".

It did well by documentary standards, poorly compared to its budget, and the stories about empty theaters are mostly in areas with very weak Trump support. Those stories spread mainly because they makes us feel good.

If someone has heart rate of 150 during an interview, it's most likely anxiety.

It's probably not great for this. Resting heart rate is correlated with fitness, but candidates can be nervous during interviews which can drastically change their heart rate. You'd probably be able to make better guesses about their health through looking at them (do they look sickly or overweight?) and interviewing them.

It's also not a specific enough signal for the types of health conditions that require taking time off work. Assuming they could get an accurate resting heart rate, most of the time it'll just indicate whether the candidate does a lot of exercise.

I don't know any commercial uses of such tech today. I'm not saying they don't exist. I just don't know of them.

I had said I don't think it's very useful for "manipulating and profiling people over video chat", so I wouldn't really expect there to be a commercial product for that. Probably it's used in fitness or heart rate monitoring apps for people that don't have a fitness tracker device and prefer not to manually count their pulse.

Here is the tech demonstrated in 2007: https://pubmed.ncbi.nlm.nih.gov/17074525/

The core algorithm is really simple. You find a patch of skin. Take the average color of the pixels in that patch. The color will become more reddish each pulse. Do an FFT and take the strongest peak in the plausible heart rate range. You could prototype this in a few hundred lines of python.

If this were useful for police or hiring managers, someone could have use the tech to make an app for them within the past 19 years.

Of course, companies have a history of trying to market a lot of BS metrics (e.g. graphology, MBTI) to hiring managers, so I wouldn't be that surprised to see a company claim they can predict employee success using pulse. Whether it works is another story.

This tech (detecting pulse from regular video) has been around almost 20 years now, and this doesn't seemed to have happened yet.

You see this type of thing in spy movies, but I'm not sure it's that useful in real life. You're basically taking one piece of data a polygraph uses, but without the most important component (skin conductance). Polygraph accuracy isn't that great to begin with. You can profile and manipulate people more effectively based on their reactions and behaviour, and their pulse will be much harder to interpret.

That's pretty close to what they have. They just named them Instant, Thinking (Standard), and Thinking (Extended), and they're discrete presets instead of a slider.

Just because bus lanes are beneficial where feasible doesn't mean optimizing the number of bus stops isn't also beneficial.

Clearly it's possible for there to be too many bus stops. You wouldn't put a bus stop every 30 ft. So it's also possible that some existing bus routes have too many stops, and spacing them out more would help.

> "The emergency lights and markings of the victims’ vehicles would have been clearly visible to the soldiers at the time of the attacks." -> speculative. The soldiers argued they were wearing night vision equipment and did not see either the markings or the emergency lights. This is at least plausible (as someone who has used thermal night vision equipment).

Is it plausible?

There were four ambulances and a fire truck with flashing lights on the roofs and the report says the soldiers had a clear view from the elevated sandbank. Night vision would obscure the markings, but lights still show in both thermal and image-intensified NVG. Even if they weren't sure they were ambulances, they should still be wondering about the emergency lights. And if they weren't sure, did no soldier look even briefly without night vision? This occurred during twilight, about half an hour before sunrise.

If they could see so little that they couldn't recognize 4 ambulances and a fire truck with emergency lights, and the aid workers never fired shot, why did they open fire?

It doesn't explain well why they initially said the vehicles were acting suspiciously by driving with their lights off and only changed their story after video emerged. And it doesn't explain why they shot at the "clearly marked UN vehicle" when it arrived well after sunrise.

If you didn't forget about the RLHF, your comment is oddly pedantic, confusing and misleading. "Correct and satisfying answers" is roughly the loss function for RLHF, assuming the humans favor satisfying answers, and using "loss function" loosely, as you yourself do, by gesturing at what the loss function is meant to do rather than formally describing an actual function. The comment you responded to didn't say this was the only loss function during all stages of training. Just that "When your loss function is X", then Y happens.

You could have just acknowledged they are roughly correct about RLHF, but brought up issues caused by pretraining.

> And I doubt RLHF gets rid of this ability.

The commenter you were replying to is worried the RLHF causes lying.

You're kidding yourself if you think you can make TNT from the 3 sentences Wikipedia has on the two-step process with no chemistry background. (And even moreso if you attempt the industrial process instead.) This isn't nearly as simple as making nitroglycerin. TNT is a much trickier process. You're more likely to get yourself injured than end up with a useable explosive. There's no procedure written there.

> If ChatGPT gave detailed plan, the chances of attack would reduce.

So you think helping a terrorist plan how to kill people somehow makes things safer? That's some mental gymnastics...