That’s definitely true, but it would be dishonest if SIAI wasn’t up front about its views on the Singularity.
There’s enough alternative terminology (intelligence explosion, hard takeoff, etc.) that they could (and often do (e.g. the AI Risk paper)) manage to talk about the Singularity just fine without talking about the “Singularity”. I don’t think anyone’s suggesting that they not be upfront about their actual positions on those issues; some people would just prefer to avoid the “Singularity” terminology now that it’s turned into a bloated mutant futurist meme complex. (So really, the name change suggestions are about being less accidentally misleading; they wouldn’t have to spend as much time explaining that they aren’t Ray Kurzweil. And hey, if they abandon the “artificial intelligence” terminology, maybe they won’t have to spend so much time explaining that they aren’t building Skynet or Terminators or HAL or the Matrix, either! ..but probably not.)
I have the impression that the Singularity Summit and other association with “big-tent” singularitarianism provide some PR and fund-raising advantages. Even if SIAI research focuses on AI and Friendliness, Thiel probably prefers that they remain on speaking terms with futurism and transhumanism more generally.
Big money is probably the the best argument. But you also have to ask who you need to convince to mitigate risks from AI, e.g. who you want to implement your formal definition of friendliness. The answer to this question is likely not Thiel but some AGI researcher working on an academic project, some big corporation like IBM or a government. In most cases having Singularity in your name or talking about transhumanism will make them delete your e-Mails. This has nothing to do with being dishonest but social engineering, being aware of public perception and using an euphemism if possible.
If the SI were known for two things—both organizing the annual Singularity Summit and also sponsoring a peer-reviewed Electronic Journal of Friendly AI Research (EJFAIR), then I suspect it would have the best of both worlds. Neither activity is so disreputable as to tarnish the good reputation derived from the other (within relevant subcultures).
To my mind, the real threat to SI (or SIAI) credibility is the perception that it is an isolated intellectual subculture, speaking only to itself, and not engaged in an open and critical dialog with other AI researchers. Choosing the right euphemism for use in the name of the organization is not the most important task in this PR battle.
Choosing the right euphemism for use in the name of the organization is not the most important task in this PR battle.
Absolutely, I was just bringing it up because Jack asked about a possible name change. It may also be the case that the term Singularity (technological) will gain general acceptance in future, as current progress towards an academic analysis suggests.
It may also be the case that the term Singularity (technological) will gain general acceptance in future, as current progress towards an academic analysis suggests.
It would be unprecedented terminology.
Evolutionary changes have been called “revolutions” (industrial), “explosions” (cambrian), “takeovers” (genetic), “catastrophe” (oxygen), or “genesis” (abiogenesis).
There aren’t any “singularities” on record, though.
I prefer “Technology Explosion”. I think it is more descriptive and more accurate. There are also the terms “Digital Revolution” and “Memesis”—which I approve of.
Well, presumably if there was more than one of it, it wouldn’t be a singularity.
Note that that isn’t what the people who use the “singularity” term to describe the hypothetical discontinity in the middle of black holes seem to think.
Somebody should point out that TheOtherDave’s comment was a joke. (At least, I’m 95% confident it was so intended, and also I found it amusing.)
The Singularity is more analogous to the event horizon of a black hole; that used to be called the Schwarzschild singularity, but since its singular behaviour was an artifact of co-ordinate systems (as was first clearly shown by David Finkelstein), this terminology has fallen out of use. One imagines that eventually historical Singularities (which are really just prediction horizons) cease to be so called after they are past.
The singularity at the center of a black hole is presumably equally illusory; general relativity simply breaks down there, giving nonsensical answers (infinities). Singularities are in the map, not the territory. (Am I the first to say that?)
Well, it was intended to be amusing, yes. (In retrospect, perhaps “if there was more than one of it, it would be a multiplicity” might have worked better. Perhaps not.)
That said, I do think there’s an actual point buried in there somewhere. The sorts of people who use the term “Singularity” to describe an event are deliberately using a term that isn’t commonly used to describe transition points, precisely in order to convey the inadequacy of former transition points as a reference class for thinking about that event. If “singularity” were a word in common usage to refer to other transition points in history, singularitarians would choose a different word.
The book “The Major Transitions in Evolution” fairly cleraly lays out many of the previous major transitions. Evolution has taken multiple major steps forwards before—for example, with the last genetic takeover, or with the origin of sex.
In my view, what we are seeing now is a modern genetic takeover. If so, we do have a useful reference class.
The genetic takeover/substrate migration is a useful reference class, but (although I agree with you that in some weak sense the intelligence explosion is already underway) minds that can rewrite themselves in place will be qualitatively different from anything that has happened before.
Sure. Memes were new. Technology was new. Machine intelligence is just the applicationn of technology to minds. The digital revolution has replaced most of our analogue media—with one major exception.
Also, thanks for linking to that Moravec paper (on your website). I hadn’t seen it before, and as I enjoy his writing a great deal, I’ll be reading it posthaste (after I finish these damned logos).
The Singularity is more analogous to the event horizon of a black hole; that used to be called the Schwartzschild singularity, but since its singular behaviour was an artifact of co-ordinate systems (as was first clearly shown by David Finkelstein), this terminology has fallen out of use.
Calling an “event horizon” a “singularity” would be pretty bonkers.
No—the commonly-cited justification is that so our model of the spacetime supposedly breaks down in the middle of black holes, so the model of the future breaks down when contemplating smarter-than-human intelligences.
Personally, I think that is nonsense. Some people are proposing busted models, is all. If your model is broken, that is hardly a cause for celebration—you should instead return to the drawing board, and build a better model.
The event horizon is a co-ordinate singularity—like the North Pole. (Wikipedia has an article on the types of ‘mathematical’ singularity. I haven’t read it, so I don’t know how good it is.) Yes, in some sense it was a mistake to call the event horizon a ‘singularity’; that’s why they stopped.
Otherwise, we appear to agree. As you say, ‘If you divide by zero in your model, that just shows that your model is broken.’ I share your interest in good terminology; but it looks to me like it’s probably far too late to get rid of ‘The Singularity’. So I just adopt it, but with a certain air of amused detachment. Sure, it’s just a co-ordinate singularity at most.
I wonder if Ulam was thinking of the Schwarzschild singularity in 1958; that’s the same year Finkelstein’s co-ordinates were introduced, so he may or may not have absorbed the illusory nature of the horizon at that point, but probably he wasn’t thinking of the singularity at the centre of the hole. But I’m not much of an historian, and would welcome corrections from those better informed.
I share your interest in good terminology; but it looks to me like it’s probably far too late to get rid of ‘The Singularity’.
I expect marketers to play along—but probably not scientists—or at least not serious ones in the relevant fields.
There are already a few exceptions, though. For example, Robin Hanson is apparently trying to rechristen the agricultural and industrial revolutions as being “singularites”. However, I think most historians will go the other way around. The “industrial singularity” will just sound like crazy talk to most people.
The singularity at the center of a black hole is presumably equally illusory; general relativity simply breaks down there, giving nonsensical answers (infinities). Singularities are in the map, not the territory. (Am I the first to say that?)
You might be the first to say it that way although the idea has certainly been expressed before. However, we don’t actually know this for sure. We don’t what happens at the singularity in a black hole. It seems likely that are models really are giving something wrong, and there are a handful of suggestions about what might be actually happening, but none of them are completely mathematically satisfying, and the lack of observational evidence is also an issue. We might be able to answer this if we a) had a theory of quantum gravity or b) had some naked singularities we could observe. Neither looks very likely at the moment, with the first probably more promising.
So, I haven’t read that paper by Krauss et al. that you suggest, but they and you seem to be saying that the event horizon must have some special local properties, which seems implausible to me. The event horizon is just a perfectly normal region of spacetime which happens to be such that trajectories on one side of the region will never leave its interior. We could already be inside the event horizon of a black hole, and there would be no way to know it. It seems wrong to consider event horizons to be boundaries of the Universe/Reality. Perhaps I am missing something, and I will look into these arguments in more detail at some point. I am very interested in this kind of thing, and in particular questions of discreteness and infinities in physics (but I cannot yet claim to be an expert).
I mean, yeah, you call the organization whatever the hell Thiel wants you to call it. But I don’t see in particular why those connections couldn’t be maintained alongside a name change that made the organization more palatable to non-futurists.
That’s definitely true, but it would be dishonest if SIAI wasn’t up front about its views on the Singularity.
There’s enough alternative terminology (intelligence explosion, hard takeoff, etc.) that they could (and often do (e.g. the AI Risk paper)) manage to talk about the Singularity just fine without talking about the “Singularity”. I don’t think anyone’s suggesting that they not be upfront about their actual positions on those issues; some people would just prefer to avoid the “Singularity” terminology now that it’s turned into a bloated mutant futurist meme complex. (So really, the name change suggestions are about being less accidentally misleading; they wouldn’t have to spend as much time explaining that they aren’t Ray Kurzweil. And hey, if they abandon the “artificial intelligence” terminology, maybe they won’t have to spend so much time explaining that they aren’t building Skynet or Terminators or HAL or the Matrix, either! ..but probably not.)
I have the impression that the Singularity Summit and other association with “big-tent” singularitarianism provide some PR and fund-raising advantages. Even if SIAI research focuses on AI and Friendliness, Thiel probably prefers that they remain on speaking terms with futurism and transhumanism more generally.
Big money is probably the the best argument. But you also have to ask who you need to convince to mitigate risks from AI, e.g. who you want to implement your formal definition of friendliness. The answer to this question is likely not Thiel but some AGI researcher working on an academic project, some big corporation like IBM or a government. In most cases having Singularity in your name or talking about transhumanism will make them delete your e-Mails. This has nothing to do with being dishonest but social engineering, being aware of public perception and using an euphemism if possible.
If the SI were known for two things—both organizing the annual Singularity Summit and also sponsoring a peer-reviewed Electronic Journal of Friendly AI Research (EJFAIR), then I suspect it would have the best of both worlds. Neither activity is so disreputable as to tarnish the good reputation derived from the other (within relevant subcultures).
To my mind, the real threat to SI (or SIAI) credibility is the perception that it is an isolated intellectual subculture, speaking only to itself, and not engaged in an open and critical dialog with other AI researchers. Choosing the right euphemism for use in the name of the organization is not the most important task in this PR battle.
Absolutely, I was just bringing it up because Jack asked about a possible name change. It may also be the case that the term Singularity (technological) will gain general acceptance in future, as current progress towards an academic analysis suggests.
It would be unprecedented terminology.
Evolutionary changes have been called “revolutions” (industrial), “explosions” (cambrian), “takeovers” (genetic), “catastrophe” (oxygen), or “genesis” (abiogenesis).
There aren’t any “singularities” on record, though.
I prefer “Technology Explosion”. I think it is more descriptive and more accurate. There are also the terms “Digital Revolution” and “Memesis”—which I approve of.
Well, presumably if there was more than one of it, it wouldn’t be a singularity.
Note that that isn’t what the people who use the “singularity” term to describe the hypothetical discontinity in the middle of black holes seem to think.
Somebody should point out that TheOtherDave’s comment was a joke. (At least, I’m 95% confident it was so intended, and also I found it amusing.)
The Singularity is more analogous to the event horizon of a black hole; that used to be called the Schwarzschild singularity, but since its singular behaviour was an artifact of co-ordinate systems (as was first clearly shown by David Finkelstein), this terminology has fallen out of use. One imagines that eventually historical Singularities (which are really just prediction horizons) cease to be so called after they are past.
The singularity at the center of a black hole is presumably equally illusory; general relativity simply breaks down there, giving nonsensical answers (infinities). Singularities are in the map, not the territory. (Am I the first to say that?)
Well, it was intended to be amusing, yes. (In retrospect, perhaps “if there was more than one of it, it would be a multiplicity” might have worked better. Perhaps not.)
That said, I do think there’s an actual point buried in there somewhere. The sorts of people who use the term “Singularity” to describe an event are deliberately using a term that isn’t commonly used to describe transition points, precisely in order to convey the inadequacy of former transition points as a reference class for thinking about that event. If “singularity” were a word in common usage to refer to other transition points in history, singularitarians would choose a different word.
Yes, I think you’re right, concerning contemporary usage; but it’s not clear to me that either Ulam or Vinge had that connotation in mind.
ETA: Or von Neumann.
The book “The Major Transitions in Evolution” fairly cleraly lays out many of the previous major transitions. Evolution has taken multiple major steps forwards before—for example, with the last genetic takeover, or with the origin of sex.
In my view, what we are seeing now is a modern genetic takeover. If so, we do have a useful reference class.
The genetic takeover/substrate migration is a useful reference class, but (although I agree with you that in some weak sense the intelligence explosion is already underway) minds that can rewrite themselves in place will be qualitatively different from anything that has happened before.
Sure. Memes were new. Technology was new. Machine intelligence is just the applicationn of technology to minds. The digital revolution has replaced most of our analogue media—with one major exception.
Also, thanks for linking to that Moravec paper (on your website). I hadn’t seen it before, and as I enjoy his writing a great deal, I’ll be reading it posthaste (after I finish these damned logos).
Calling an “event horizon” a “singularity” would be pretty bonkers.
No—the commonly-cited justification is that so our model of the spacetime supposedly breaks down in the middle of black holes, so the model of the future breaks down when contemplating smarter-than-human intelligences.
Personally, I think that is nonsense. Some people are proposing busted models, is all. If your model is broken, that is hardly a cause for celebration—you should instead return to the drawing board, and build a better model.
The event horizon is a co-ordinate singularity—like the North Pole. (Wikipedia has an article on the types of ‘mathematical’ singularity. I haven’t read it, so I don’t know how good it is.) Yes, in some sense it was a mistake to call the event horizon a ‘singularity’; that’s why they stopped.
Otherwise, we appear to agree. As you say, ‘If you divide by zero in your model, that just shows that your model is broken.’ I share your interest in good terminology; but it looks to me like it’s probably far too late to get rid of ‘The Singularity’. So I just adopt it, but with a certain air of amused detachment. Sure, it’s just a co-ordinate singularity at most.
I wonder if Ulam was thinking of the Schwarzschild singularity in 1958; that’s the same year Finkelstein’s co-ordinates were introduced, so he may or may not have absorbed the illusory nature of the horizon at that point, but probably he wasn’t thinking of the singularity at the centre of the hole. But I’m not much of an historian, and would welcome corrections from those better informed.
I expect marketers to play along—but probably not scientists—or at least not serious ones in the relevant fields.
There are already a few exceptions, though. For example, Robin Hanson is apparently trying to rechristen the agricultural and industrial revolutions as being “singularites”. However, I think most historians will go the other way around. The “industrial singularity” will just sound like crazy talk to most people.
You might be the first to say it that way although the idea has certainly been expressed before. However, we don’t actually know this for sure. We don’t what happens at the singularity in a black hole. It seems likely that are models really are giving something wrong, and there are a handful of suggestions about what might be actually happening, but none of them are completely mathematically satisfying, and the lack of observational evidence is also an issue. We might be able to answer this if we a) had a theory of quantum gravity or b) had some naked singularities we could observe. Neither looks very likely at the moment, with the first probably more promising.
My take on the issue: http://finitenature.com/no_singularity/
So, I haven’t read that paper by Krauss et al. that you suggest, but they and you seem to be saying that the event horizon must have some special local properties, which seems implausible to me. The event horizon is just a perfectly normal region of spacetime which happens to be such that trajectories on one side of the region will never leave its interior. We could already be inside the event horizon of a black hole, and there would be no way to know it. It seems wrong to consider event horizons to be boundaries of the Universe/Reality. Perhaps I am missing something, and I will look into these arguments in more detail at some point. I am very interested in this kind of thing, and in particular questions of discreteness and infinities in physics (but I cannot yet claim to be an expert).
Am inclined to say: “but that’s why it’s called an event horizon.”
I mean, yeah, you call the organization whatever the hell Thiel wants you to call it. But I don’t see in particular why those connections couldn’t be maintained alongside a name change that made the organization more palatable to non-futurists.