Only the best and brightest are working for Anthropic.
-
@Mabande
You don't get a PhD without rigor.We don't need to attack a philosopher's credentials or skill here. The problem isn't the philosophy, it's the engineers who are providing inaccurate technical explanations to the philosopher. This is why we need to infuse the humanities into our engineering programs - we need technical philosophers to call this bullshit out.
(And we have them - see the DAIR group, for example.)
@adardis@Mabande
since this thread is making the rounds again, y'all who read this far get a link to my own amateur-but-carefully-thought-through math and philosophy on why Anthropic and others are provably full of shit: https://thoughtlost.org/nosingularity/ -
@rollspelosofen @Mabande I'm sorry I have to differ.
LLMs don't reason.
And there a difference between obsequiousness and being polite.@marjolica @Mabande By ”logical reasoning” I was referring to the common ”if… else…” reasoning statements that all modern software use. And I was using polite very broadly here.
-
@petealexharris @celesteh @adardis The IRL Darth Vader wannabe got to have his cool soundtrack...
-
@marjolica @Mabande By ”logical reasoning” I was referring to the common ”if… else…” reasoning statements that all modern software use. And I was using polite very broadly here.
@rollspelosofen @Mabande I don't think LLMs use if .. else logic, though decision trees are a different form of AI and largely work that way (I've written them myself in the past).
LLMs just include a whole set of statements in parallel, including your prompt, their billions of weights based on data they were trained on and imperative guard rails and meta instructions set by their trainers. Of course as a stochastic parrot they don't always even prioritise their guard rails and often let information in their training data override what you tell them to do in the prompt. -
@petealexharris @adardis @Mabande
People with philosophy undergraduate degrees have very high mid career income, but the more advanced the degree, the lower the income gets (this is also true for music). The reason is that most don't work as philophers.
If somebody is offering super high wages to somebody not already famous, that should be a clue that something is amiss. But, I mean, who *wants* to see the problem.
(Nobody ever uses composers for evil, so its very easy for me to say all of this, with my music PhD.)
@celesteh @petealexharris @adardis @Mabande smash cut to the composer who was horrified to find out that his cheerful tunes were being used for torture …
-
@rollspelosofen @Mabande I don't think LLMs use if .. else logic, though decision trees are a different form of AI and largely work that way (I've written them myself in the past).
LLMs just include a whole set of statements in parallel, including your prompt, their billions of weights based on data they were trained on and imperative guard rails and meta instructions set by their trainers. Of course as a stochastic parrot they don't always even prioritise their guard rails and often let information in their training data override what you tell them to do in the prompt.@marjolica @Mabande Okay.
-
@Mabande Have a look at her Wikipedia page. If it's accurate, she's a very rigorous philosopher.
I'm tempted to say "no bullshit there" but it's possible to be rigorous and to spout bullshit.
My instinct says she has/they have other reasons for saying the things they do.
-
@petealexharris @celesteh @adardis The IRL Darth Vader wannabe got to have his cool soundtrack...
@Mabande @petealexharris @adardis sadly no
-
@Mabande Whenever I read this, I have to seriously reconsider if Mao's idea of sending the youth to the country to learn the life of the peasants was so wrong…
@Illuminatus The idea? Kinda good; The implementation? Not that good
-
@fgbjr @joshg @Mabande @adardis
I got my Ph. D. in chemistry after graduating near the top of my undergraduate class at a tough school.
Linus Pauling knew more about chemistry on a bad day than I ever learned.
He knew more about chemistry than many if not most of the professors I knew.
He was still very much wrong about vitamin C and its role in human health.
There are lots and lots of very smart people out there who get specific things demonstrably wrong despite their very apparent talents.
-
@Mabande I'm struck by the extent to which the responses focus on the person, rather than her arguments/ideas.
1. I see intelligence & consciousness (appropriately defined...) as something that humans have.
2. I see us as large, complex machines, subject to the laws of physics.
3. thus, I see intelligence & consciousness as emergent properties of ion channel mechanics, voltage differences...
4. I have not heard a strong argument as to why these properties can only emerge from C-based, not Si- -
@Mabande I'm struck by the extent to which the responses focus on the person, rather than her arguments/ideas.
1. I see intelligence & consciousness (appropriately defined...) as something that humans have.
2. I see us as large, complex machines, subject to the laws of physics.
3. thus, I see intelligence & consciousness as emergent properties of ion channel mechanics, voltage differences...
4. I have not heard a strong argument as to why these properties can only emerge from C-based, not Si-5. We've not seen Si-based entities perform like this before: passing Turing tests... That qualifies them, for me, as a new kind of entity.
They're different. The frontier is jagged/spiky/whatever.
For me, that's an invitation to explore, not attack.
-
@adardis @Mabande @petealexharris
that doesn't make it OK for philosophers to become propagandists.
-
5. We've not seen Si-based entities perform like this before: passing Turing tests... That qualifies them, for me, as a new kind of entity.
They're different. The frontier is jagged/spiky/whatever.
For me, that's an invitation to explore, not attack.
@rowat_c I haven't seen a strong position against the possibility of Si-based intelligent lifeforms in the scientific community (one of the baseline assumptions in the exobiology field is AFAIK that Si-based life is probable), what's dismissed is that these LLM:s are in any meaningful way conscious or even showing signs of intelligence.
The Turing Test doesn't measure intelligence, just the ability to mimic human output, and have been passed by chatbots as far back as at least ELIZA, 60 yrs ago. -
@rowat_c I haven't seen a strong position against the possibility of Si-based intelligent lifeforms in the scientific community (one of the baseline assumptions in the exobiology field is AFAIK that Si-based life is probable), what's dismissed is that these LLM:s are in any meaningful way conscious or even showing signs of intelligence.
The Turing Test doesn't measure intelligence, just the ability to mimic human output, and have been passed by chatbots as far back as at least ELIZA, 60 yrs ago.Thanks @Mabande !
My preferred approach to both terms is against a definition. Otherwise it seems to me we lose the possibility of making falsifiable statements, leaving us to sling "yes it is!" and "no it isn't!" back and forth.
When an old definition is applied to a new domain, it often needs to be adjusted. Thus, an LLM maxing out an IQ test doesn't lead us to conclude it has a 200+ IQ: it's memorised the training data.
1/n
-
Thanks @Mabande !
My preferred approach to both terms is against a definition. Otherwise it seems to me we lose the possibility of making falsifiable statements, leaving us to sling "yes it is!" and "no it isn't!" back and forth.
When an old definition is applied to a new domain, it often needs to be adjusted. Thus, an LLM maxing out an IQ test doesn't lead us to conclude it has a 200+ IQ: it's memorised the training data.
1/n
@Mabande However, if we say that a key ingredient to C-based mathematicians solving problems is their intelligence (rather than their pencil or paper supply, or Hagoromo chalk, coffee or blood sugar), then it feels like we're using sleight of hand if we don't attribute success in the same way for Si-based mathematical problem solvers.
2/n
-
@Mabande However, if we say that a key ingredient to C-based mathematicians solving problems is their intelligence (rather than their pencil or paper supply, or Hagoromo chalk, coffee or blood sugar), then it feels like we're using sleight of hand if we don't attribute success in the same way for Si-based mathematical problem solvers.
2/n
@Mabande thus, Anthropic's J-space paper seems sound: neuroscientists have a working/operational definition of (a form of?) consciousness that yields testable predictions. (Else, not science?) They find that Claude exhibits some of those predicted properties.
They don't take a stance on (the hard problem of) consciousness, merely whether Claude exhibits behaviour predicted by one definition of consciousness.
To me, it seems that, to disagree, we need to show a flaw in their derivations.
3/3
-
@Mabande thus, Anthropic's J-space paper seems sound: neuroscientists have a working/operational definition of (a form of?) consciousness that yields testable predictions. (Else, not science?) They find that Claude exhibits some of those predicted properties.
They don't take a stance on (the hard problem of) consciousness, merely whether Claude exhibits behaviour predicted by one definition of consciousness.
To me, it seems that, to disagree, we need to show a flaw in their derivations.
3/3
@rowat_c Sorry, but it's useless to try to have a conversation if one of us redefines or throws out well-established definitions on the fly.
It simply makes it so we can't agree on anything, even the basics of what the conversation's about. -
@rowat_c Sorry, but it's useless to try to have a conversation if one of us redefines or throws out well-established definitions on the fly.
It simply makes it so we can't agree on anything, even the basics of what the conversation's about.@Mabande I think we agree: if we want to make non-tautological statements about C- or Si-based intelligence or consciousness, we need to operationally define those terms: "X has property Y if it exhibits behaviour Z".
-
@Mabande I think we agree: if we want to make non-tautological statements about C- or Si-based intelligence or consciousness, we need to operationally define those terms: "X has property Y if it exhibits behaviour Z".
@rowat_c Problem is when there's a vast discrepancy re. what "exhibiting behaviour" means.
Successful mimicry of a property (eg the baseline for passing the Turing test) is not the same as having the property.
