Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

There is no A.I.

newyorker.com

Post reply on HN
Print view View original

11–20 of 523 posts

Re: There is no A.I.

#11

Post by quickthrower2 »

monero-xmr wrote: post #7

It is possible that AI can perfectly replicate human intelligence. But because it is purely digital, I cannot kill it - I can delete it, copy it, recreate it, and so on but never kill it. Only biological organisms can die. Therefore no matter what happens, there is a distinction between human intelligence and AI. I can destroy an AI creature and the only penalty should be property damage (if relevant).

Machines running AI can certainly die, that is why chaos monkeys and kubernetes etc. exist. They can be backed up, unlike humans though. Although humans are backed up by virtue of there being 8 billion of them, so if one dies, the world can keep going on pretty much as before, albeit with some sadness for some of the people. This sounds morbid, but hard to avoid when comparing humans to machines!

Re: There is no A.I.

#12

Post by runarberg »

atleastoptimal wrote: post #3
chpatrick wrote: post #2

[flagged]

This is correct. Expect 10000x more articles like this over the next couple years.

But, the author is not wrong.

> In my view, the most accurate way to understand what we are building today is as an innovative form of social collaboration.

This is a very interesting take from a philosophy grandmaster who is well versed in the technology and statistics behind it. If we get 10 000x more articles like this in the next couple of years, we should consider our selves lucky. Grifters have controlled the debate about AI for far too long.

Re: There is no A.I.

#13

Post by hgsgm »

monero-xmr wrote: post #7

It is possible that AI can perfectly replicate human intelligence. But because it is purely digital, I cannot kill it - I can delete it, copy it, recreate it, and so on but never kill it. Only biological organisms can die. Therefore no matter what happens, there is a distinction between human intelligence and AI. I can destroy an AI creature and the only penalty should be property damage (if relevant).

This is "begging the question".

Re: There is no A.I.

#15

Post by sordina »

monero-xmr wrote: post #7

It is possible that AI can perfectly replicate human intelligence. But because it is purely digital, I cannot kill it - I can delete it, copy it, recreate it, and so on but never kill it. Only biological organisms can die. Therefore no matter what happens, there is a distinction between human intelligence and AI. I can destroy an AI creature and the only penalty should be property damage (if relevant).

So if you lost the weights how is that not killing the AI? Is it because it lacks the death experience? If so what about bitrotting the weights incrementally and degrading its inputs?

Re: There is no A.I.

#16

Post by lcnPylGDnU4H9OF »

thejarren wrote: post #8

This article initially seems to be arguing over semantics, saying that AI isn't AGI, which most people would agree with. Then talks about data dignity, which is interesting but a bit of a tangent. My understanding of the concern over AGI, or even an all-powerful "AI" isn't that it's not truly self-aware, but rather that it's motives are not in line with humanity's motives. It doesn't matter if it trains on data by hu…

Data dignity would give us insight into how the AI “thinks” so we can better understand how it’s aligned.

Re: There is no A.I.

#17

Post by majormajor »

thejarren wrote: post #8

This article initially seems to be arguing over semantics, saying that AI isn't AGI, which most people would agree with. Then talks about data dignity, which is interesting but a bit of a tangent. My understanding of the concern over AGI, or even an all-powerful "AI" isn't that it's not truly self-aware, but rather that it's motives are not in line with humanity's motives. It doesn't matter if it trains on data by hu…

My read is that the author doesn't think today's tools are particularly close to AGI that has its own motives, or that it's at all useful to characterize these tools as "having motives" at all.

There are those who disagree with how close we are to "AGI" with today's LLMs, but as an outsider, the author seems pretty well-qualified in his position at MS Research compared to some of the less-informed prognosticators. Even the linked-out discussion with Altman, who the author says is more pessimistic than he is, doesn't appear to be a case of Altman saying these models are close to AGI or dangerous for "alignment" reasons vs "they can be used for bad things."

Worrying about the uses of tools is a VERY different thing than worrying about if the tools are gonna overthrow us, and I think that's what the author is really pushing for here. If we have that conversation we can get closer to real, specific policies, laws, etc to make sure the tools benefit us instead of the other way around.

Re: There is no A.I.

#20

Post by sundarurfriend »

If ChatGPT is "social collaboration", it's an involuntary collaboration between our past selves, with no active control or choice over what we're "collaborating" on. At which point, we're really stretching the meaning of the word collaboration.
Post reply on HN