Retric

Not really making nukes take a ridiculous amount of energy and you only get to use it once where the energy to create ITER gets split across its operating lifespan.

Further fission here requires a fission first stage which represents a huge additional energy input.

The flaw in that argument is a great number of chemicals do not cause cancer.

We don’t disagree on the large scale risk of AI here, I do disagree that cars ‘weren’t hooked up to weapons systems.’

No, prove something else could mean proving something very trivial thus making the proof meaningless on its own.

IE a proof can be true, rigorous, and not at all what was asked for.

Less so when it costs prosecutions. Cops collectively spend quite a long while reading people their Maranda warning for example.

That does not require empathy. Facts are useful independent of your ethical framework.

If your goal was to kill all former pirates you’d want to know their professional background and likely current occupation.

Similarly a ruthless dictator who wants to end piracy should understand all tools at their disposal.

Ignorance is simply a detriment for efficiency. Though obviously if you want to exploit someone else it’s useful for them to be ignorant, which helps explain your own failings.

Cars were quickly tuned into Tanks, APC’s, etc.

The fear around military AI is more about its failure than anything else, but nobody is trying to make fully automated machines which are manufacturing bombs from raw materials. Blowing up the wrong building matters a great deal to the people in the building, but for humanity a building was going to be destroyed either way. So the risk at any one point ends up being quite limited.

Mass adoption of technology always has negative outcomes because any change at scale has negative outcomes. Cars as a technology kill a great number of people and yet few are willing to ban them.

“AI” in the widest definition possible could do while a lot of harm without being a net negative. I’d be way more concerned about the healthcare sector than the military as the military already has significant collateral damage.

You’re rejecting people who self identified as Republicans for not being Republicans.

You can reject the category of Republicans in all situations or zero situations but doing so only some of the time is not a defensible position.

Industry has often asked for regulations that have nothing to do with the claims being made. Zuck was a recent example in the computer industry. https://www.bbc.com/news/world-us-canada-47762091

Here’s the same viewpoint from well outside the HN bubble. https://www.nationalreview.com/corner/be-wary-of-industries-...

If the options are to shoot yourself in the face or wait a year or even a month and have someone else shoot you, I think most people would pick the second.

Companies are continuing because they don’t actually believe they are going to have significant negative consequences happen sooner. It’s marketing fluff around how cutting edge what they are doing is not any kind of realistic threat assessment.

Their stance withAI code is it is theft.

“I don't like their policy, but I accept it as it is because it is not possible right now for me to paint on something else.”

A voluntary donations of your effort are not the same a theft.

A great number of doctors, lawyers, artists, programmers. etc donate their time doing professional level work to worthy causes. Whatever anyone does with AI is not in the same category.

> Yes at constant resources they are significantly better.

There has been some fundamental advancements in LLM training and efficiency. But good luck actually finding non toy models of exactly the same size, hardware, and training one from now and other from 2001 where the newer model is dramatically better at literally everything.

Most of the real world advances are from throwing ever more resources at the problem.

> Gen 2 takes

Diminishing returns are not a question of a single generation. Gen 2, 3, 4, 5… would also need to have the same 3x return on 2x resources or you don’t have an exponential curve.

Not at constant processing power, memory, training, etc.

But that’s beside the point, being arbitrarily bad at everything isn’t a problem. The diminishing returns as you apply the ceiling is problematic for self improving AI.

Self improving AI runs into the same issue as prefect comprehension, you can’t get arbitrarily better at everything.

The idea AI can get better at everything at the same time is a holdover from deeply flawed science fiction not some realistic goal.

Monthly electricity bills are based on usage, and it works well but there’s a limit to how surprising a bill can be. The difference is the relative orders of magnitude you can be charged for these services you can go from 20$/month to 200k/month without warning.

AI isn’t replaced timesheets, payroll, accounting, or tax filing.

It’s exactly that kind of boring plumbing that companies want to outsource which is the exact opposite of what near term AI is good at. So yes there’s many SaaS companies that should be concerned, but suggesting all SaaS is ending anytime soon is pure hype nonsense.

It’s rarely such a simple tradeoff that only two things are being optimized. Modern humans live way longer than wolves for example despite being of similar body mass. However that’s fairly new in our evolutionary history as wild chimps for example cap our at ~63.

Cancer is extremely rare as a cause of death in wild animals. Again is more common but late enough that reproduction has produced multiple offspring.

Energy expenditure isn’t something we’re concerned with but drives a great deal of evolutionary optimization.

People having “large” numbers of children.

The modern trend of most families with 0-2 kids isn’t sustainable. We don’t need to go back to 5+ being the norm but a self stable society needs to be able to offset non parents thus requiring 3+ children families to be quite common.

> thats also not about processing and reacting to signals > not an information processing system in the same sense

I can only assume you have some sort of philosophical objection here, but this is information processing.

Neurons are reacting to a range of chemical signals not just a single neurotransmitter because they don’t map to what CS style neural networks. Understanding the brain requires understanding the nuances not simplified models.

> something bogus like system 3

These systems exist, if the author you’re referring to is unaware or unwilling to discuss major issues with their classification then that’s a hit to their credibility.

Danger isn’t some universal thing, people are adapted to notice fact moving objects heading towards them quickly very fast. A very dangerous snake isn’t in that category of danger and can take a significant chunk of time to notice after you’ve stopped looking in that direction. Noticing social danger is extremely critical and glacial by any of the metrics previously discussed.

Emotions cover a huge range of complex interactions that depend on many different parts of the brain but hormones are inherently slow to diffuse through the brain. Pair bonding between a mother and child needs to happen, it doesn’t need to happen in 50ms.

>I can't answer questions about criminal charges or other news about the President. That's outside America's scope

That it was specifically engineered to give that response. Thus political pressure was applied to the design.

The funny bit was just humorous. There’s likely many ways to get that same information without referencing Biden.

You’re assuming a far more functional system than what was going on.

“And there was no penalty for not paying—there were no regulations allowing the utilities to cut off a customer’s power.”

Collecting more money in a corrupt organization doesn’t means they will automatically get a more effective distribution system.

The amygdala isn’t particularly fast here. It’s got significant connections to higher order centers in the brain and it’s own memory because classifying dangers let alone more complex emotional states can be quite complicated.

If he’s defining System 3+ then that kind of specificity could still work, but from what I understand he doesn’t thus making subconscious a poor fit for the systems described.

We’re talking about biological processes which have physical limitations on how frequently a neuron can fire as well as a signaling system that based on how rapidly neurons fire. Loops anywhere in the system enforce a minimum 5ms latency and more realistically several times that. A typical neuron is firing closer to 0.1 to 2 times a second. Add to that the physical transmission speed and rapid response doesn’t allow for loops.

There’s a reason reflexes for the feet are handled by the spine that’s got nothing to do with total processing power.

System 1 and 2 is a description of physical process in people.

Using conscious vs subconscious processing is not however a meaningful definition. Subconscious processing isn’t necessarily fast. Visual processing and sensory integration can be quite slow without any conscious input.

Slower paths means loops here for humans where the output of a neuron gets feed back into itself. The fastest path = a feed forward neural network without loops.

LLM’s operate strictly feed forward neural networks.

Consciousness is ill defined hogwash when used in such descriptions.

The process of internal refinement without external action however fits.

The ratio between a single pass and multiple passes is unchanged when you throw more processing power at both.

In a human 30ms vs 3-4 seconds is a 1:100 ratio. Single vs multiple passes with an LLM varies but a 1:100 ratio isn’t unrealistic. So with enough compute and the right workload single vs multiple pass LLM could sit in that exact same 30ms vs 3-4 second timeframe.