Forgot your password?
typodupeerror

Comment Re:HAHAHAHAHA!!! (Score 1) 162

Nope. That is not evidence. That is called "physicalism" and it is basically a religion. You have exactly nothing.

What should we call it when one expects other people to entertain possibilities for which no affirmative evidence exists? Inductive inference is not religion.

You keep yelling physicalism at everyone even when it is made explicitly clear to you no such leaps are being made. The question I have is why? Could it be you are the one clinging to religion?

What you are basically saying is because it is impossible to rule out a magic bunny rabbit that installs the carrot of intelligence into special chosen beings other peoples unwillingness to waste their time considering the possibility of such a rabbit constitutes some sort of a religion.

Comment Re:There is currently no way (Score 1) 162

Anyone who thinks they spotted an LLM reasoning already is foolish because we know for sure they aren't doing that.

I would bet you can't even define what reasoning is much less prove LLMs can't do it.

What I find particularly untenable is if a person were to emit the same words in order to arrive at a conclusion everyone would call it reasoning. Yet when a machine emits the same words for the same purpose it is defined by some as something else. I've heard some refer to this as an elaborate parlor trick. The phrase "not even wrong" comes to mind... what is being asserted seems to be merely begging the question not even in the realm of science to begin with.

Comment Re:Bad ideas and bad laws (Score 1) 110

There is a famous example of an AI being told not to destroy a database, but it does it anyway and then apologizes. The thing is it KNEW it was destroying the database when it did it. This wasn't an accident, it was stupidity.

While everyone can point to AI doing something trivially stupid this doesn't erase demonstrated capabilities. Cherry picking and or anthropomorphizing intelligence isn't all that useful in reasoning about actual system capabilities.

The apologizes mostly take the form of post training bludgeons models are made to say which leads to very jarring experiences for users with the sorry but not sorry. Attention schemes are also not perfect sometimes models gloss over important (to you) instructions as a person would a typo or misstatement.

I personally find it insane anyone would give an LLM access to anything much less production systems.

2) Super intelligent AI's do NOT refer to AIs 'smarter than humans'. Instead it refers to AI's that are substantially more capable at certain specific tasks than humans are. Just as calculators are better at figuring out the square root of something.

I don't accept your definition and I don't think this is most peoples understanding of what ASI means.

They are not artificial 'intelligences' getting more intelligent. They are large language models learning more language skills, more memory, and more processing time.

This is just playing with words.

Comment Re:Can't compete (Score 1) 221

China invading Taiwan and Iran having nukes are the two republican boogey men.

Both are credible possibilities independent of political ideology or personal beliefs.

China has reflected a desire both rhetorically and in military efforts towards a goal of taking Taiwan.

Iran has stated it is actively seeking nukes, Iranians involved with JCPOA have subsequently made public statements of course they were going for a nuke. They have of course been building out all of the technical capabilities necessary to becomes a nuclear weapons state... material processing, delivery vehicles and tens of thousands of centrifuges in hardened underground facilities.

Comment Re: Useful idiot (Score 1) 110

Yep. They would have done a slow collapse. But no, the convicted felon had to misdirect attention away from his child-raping and gave them a strong external enemy and renewed the Iranian regime and gave it new purpose. And it looks like Iran has real power over the world economy and they now know that. Well done!

Iran is a bit of an enigma. Most of the common western sound bites and wisdom are flat wrong. The country is more of a mafia state undergoing an extreme and ever growing legitimacy crisis on top of catastrophic financial picture featuring something on the order of 40% yearly inflation over the last 8 years. When Khamenei was killed and the bombs fell people were literally celebrating and dancing in the streets. You could not be anywhere in any major city and not hear the cheering or see the fireworks.

Worth noting the massacres of Jan 8th and 9th were primarily carried out by foreign terror groups imported by the regime itself. When your government imports foreign terrorists to massacre your own people and harass Iranians in the streets whose flag do you suppose people will be rallying around? When a foreign military attacks your own people who have been severely oppressing and murdering you for decades whose flag do you suppose people will be rallying around?

Iran's continuous external aggression in the form of the external revolution they have been executing for the last half century earned Iran the title of world leading state sponsor of terror and has lead to destabilization of states throughout the entire region.

After decades Iran's terror network has began to unwind with the fall of Assad, Oct 7th attacks and the failing economic situation in Iran. They are now being successfully starved of resources in which to continue to fund their terrorism throughout the region. This is having huge effects throughout the region some I still find hard to believe. Who would have thought we would ever see a Lebanese president that doesn't just speak out against armed nonstate actors but inks a deal with the Americans and fricking Israel to rid themselves of them. Similarly the anti-corruption push and Sept 30th deadline in Iraq is amazing... The implosion of Hamas in Gaza. Drying up of Iranian support to Houthis. All of it is being systematically eroded.... all of it chaos that could never sustain itself absent these absurd cash flows from oil revenues.

The Iranian regime is increasingly unable to pay even its own forces and foreign terror groups. The Hormuz choke point is a self limiting problem with numerous completed and ongoing efforts to mitigate and workaround it.

Comment Go big or go home (Score 1) 110

I can at least tolerate people who argue for a total ban. I think there is at least a coherent argument to be made yet nobody seems to actually care enough to stake out this position. While you will never stop people from using or improving existing tools the enabling knowledge and industrial base would be starved of the massive capital flows with which to further develop AI specific technology and associated formal education at scale.

Anything short of that is in my view more dangerous than doing nothing. The type of legislation they should be perusing is preventing automated judgement of humans and use by governments and corporations for mass surveillance.

The underlying problem is the enabling knowledge and hardware not what people do. It is a fools errand to think you can guard bags of weights or that any possible safeguards baked into the bags would not be removed with trivial effort. The safeguard nonsense also creates a false sense of security among the population that is not reality based. If technology progresses to the point where someone is able to create an ASI / magic genie it must be assumed others will soon follow.

The immediate threat of AI comes from people not technology. Allowing government to regulate and control it will just aggregate power into the hands of a few government/corporations who of course will seek to leverage their positions. AI is already being used for mass surveillance and control of entire populations.

It is more realistic that you ultimately end up with a situation in which people are held liable for what their agents actually do rather than trying to control development of weights and algorithms. In a future with ASI the counter to a bad ASI is a good ASI... that is really the only thing you can do short of a total ban.

Bernie's law is insane, trying to justify it on the basis of OpenAI's hacking drama and press accounts is a weak argument. The increased bio threats enabled by technology ***persist independent of AI***.

Analogies to nuclear weapons and thinking AI can be controlled/monitored like some sort of CTBT regime is completely divorced from reality.

Comment Re:There is currently no way (Score 1) 162

To blend up LLMs and make them reason. You can train them and refine them but it won't make them any better at reasoning. But they will be able to fool the foolish at reasoning.

In other words anyone who spots an LLM reasoning is foolish? Seems like a rather robustly falsification resistant thing to say.

Comment Re:Same Altman: Snake Oil Salesman (Score 1) 129

Sam Altman has claimed a lot of outlandish things, but he's never said AI would solve all the world's problems, take us to the stars, stop world hunger, fix our governments, or save the planet. YOU made all that stuff up.

Granting your wishes, and solving all the world's problems are two very different things.

In case a magic genie that can grant any wish was not good or specific enough for the reader here is another:

"Although it will happen incrementally, astounding triumphs - fixing the climate, establishing a space colony, and the discovery of all of physics - will eventually become commonplace. With nearly-limitless intelligence and abundant energy - the ability to generate great ideas, and the ability to make them happen - we can do quite a lot." ~Sam Altman

Slashdot Top Deals

In any problem, if you find yourself doing an infinite amount of work, the answer may be obtained by inspection.

Working...