The same can happen with many things on the internet. Reddit, Youtube etc. It's not specific to facebook.
30 karma · joined December 11, 2015
Exactly. But not because they enjoy tweaking some website settings or enjoy the control over the layout of a page. No, the reason is connection with other people. They write a blog so that they get feedback from people. They post pictures to Instagram and Facebook to get likes and show off.
Connecting the electric guitar and repairing things with screwdrivers is seen by most people as a chore. Sure, techie, geeky tinkerers enjoy messing around with things, the same ones who enjoy configuring linux, for example.
I just haven't seen a single use case that would make sense in an everyday setting. Having a knob that controls the lamp? Well, lamps already have that, without being radio controlled.
If a car dashboard is designed well, then there's no need to remap the buttons. Very few people like to customize their stuff that much. If there is a trend in software (and hardware) it's that they remove options. We have less features and customizations than 10 years ago. Developers realized that people just mess their settings up and they call support. Also, the more options there are, the more bugs there will be. So now instead of supporting a myriad of settings, they just go with a default and leave it at that. Why don't we have the option of vertical browser tab arrangement in Chrome? Well, because they decided it's too complex and unnecessary for most users.
People just want to get their things done. They want to get business things done with other business people and they want to have fun with their friends.
The success of the Internet is all about human communication, sharing experiences (mainly through photos), and gossip.
People don't want to "rewire" their products, they don't want to hack on this kind of thing. Except for geeks and technical people, who can also handle normal apps and interfaces.
The analogy with the Internet breaks down because the Internet is mainly successful because it connects people. That's what people care about, other people. The telephone was a success because it lets you talk to other people.
Unless this thing helps people deal with other people, it's too complex and uninteresting for the masses and probably too simple and tedious for technical people.
It may well be that reading doesn't make you successful, but not-reading makes you not-successful.
It's like playing the lottery doesn't make you win, but not-playing makes you not-win.
Sutskever is a researcher at Google, worked with Hinton in Toronto and Andrew Ng at Stanford.
Karpathy studied in Toronto and at Stanford, worked under Fei-Fei Li, worked at Google. He also has an awesome blog and seems very active and passionate about computer vision and ML.
Kingma also works with deep neural nets, worked under Yann LeCun (who works at Facebook)
Schulman is a PhD Candidate at Berkeley with publications at top conferences.
Zaremba is an PhD student at NYU, intern at Facebook. Impressive publication list and awards.
Abbeel is at Stanford's AI lab.
Bengio is one of the "stars" and celebrated figures of the deep net revival.
Levine is a researcher at Google working on deep nets with many serious papers.
---
Basically these are the main domain experts among them. The list is quite skewed to Google/Facebook, Stanford/Berkeley/Toronto and deep net researchers, working primarily on computer vision.
Correct, I was conflating things. But his unfriendly AI scenario is pretty bad too.
My issue with it is that it's a symptom of an over-confident belief system naturally leading people to such conclusions. And now I don't want to get into refuting it, it's been done numerous times to varying degrees. I think the whole underlying system is problematic, there is no superficial mistake in the reasoning, once you accept a few philosophical standpoints and value judgments that the reader is spoon-fed while reading his blog posts.
The downside is that much of the research is probably held secret for business advantages. The public releases are more of a PR and hiring strategy than anything else in my opinion. By sending papers to conferences, Google's employees can get to know the researchers and attract them to Google.
Others say there's nothing to worry about, Google and Facebook are just today's equivalent of Bell Labs, which gave numerous contributions to computer technologies without causing much harm.
Now, the human brain is definitely a complicated thing to study and understand (by whom? by itself!), but framing it as if the brain was a computer that received a task that it then solved, is the wrong way of thinking about this.
And yes, AI will definitely be used for all sorts of purposes including hostile means. Just like anything else, really. Financial manipulation, spying, intelligent military devices, cracking infrastructure security, etc.
These are realistic concerns, we shouldn't fall for the Skynet red herring. We can have problems with ethical AI use, even if it's not a self-aware super-human superintelligence.
There are different ideas for what constitutes AI. Expert systems and knowledge-based reasoners? Pattern-recognizer black boxes? Chatbots? AGI?
Over the years the concept of AI shifted. Until recent years "AI" was mostly used for things like A* search, creating algorithms that play turn-based table games for example (see the Russell-Norvig book), symbolic manipulation, ontologies etc., a few years ago it began to also refer to machine learning like neural networks again.
Neural networks are good at what they are designed for. Whether they will lead to the path to human-like artificial intelligence is a speculative question. But symbolic manipulation alone won't be able to handle the messiness of sensory data for sure. I think neural nets are much better suited for open-ended development than hand engineered pipelines that were state of the art until recently (like extracting corner points, describing them with something like SIFT, clustering the descriptors and using an SVM over bag-of-words histograms). Hand engineering seems too restrictive.
When in university, some courses would focus very much on definitions and lists of things and what part of the field covers what things, I could tell there was some pretentious bullshit going on. Now mathematics seems like an exception to this, but actually they don't argue about definitions in this sense. If you define your terms slightly differently a mathematician may be annoyed but he will recognize if your overall work is valid.
It's also true on an individual level. I noticed that people who like to argue whether they are programmers or software developers tend to be less concerned about actually getting something done, vs. people who'd say "call me whatever; you can come and watch what I do and decide what you call it".
It comes across as overcompensation for having little to say otherwise. Good scientific papers also don't dwell too much on how to categorize and break up the related fields. But apparently there are people who enjoy defining terms precisely, like whether they do Data Mining, Data Science, Machine Learning or AI or Statistics or Probability. They are all fluid categories and have significant overlaps. There is just no reason to work towards sharp separation.
In the first case a meaningful question could be "what is the thing that makes these objects similar, what is a description that connects them?" For example if you see that certain animals die from some poison but others are unaffected. Then you may make up the concept of "resistastrong" animals that don't get killed.
You can do it in two very distinct ways. Either by enumeration, or by declaring the set as {the animals that don't die from the poison}. In the first case the definition is fixed and any more animals that are discovered to die from the poison are not accepted into the set. In the second case, the set may grow as more animals are discovered to die from the poison.
The way natural language works is often a hybrid of the two. We may call something a name, but then gradually the meaning can drift. We first have a fixed set. Then we realize some simple description that unites those things. Then we discover more things that fit this description and incorporate them in the set. Then we iterate again to find a better, more compact description of this new set. Maybe this will actually even throw out some of the elements that were previously included, because the description can be made much simpler if you throw out some edge case.
I think an interesting distinction can be made between concepts that are linked more to the actual things covered by them and as we find out more about those things, the definition text gets updated, vs. concepts where entities can move in and out of its coverage as we learn more about them.
In case of life, you may say at the outset that dogs, humans, birds and trees etc. are alive, but rocks, clouds etc. aren't. Then your task essentially becomes like a machine learning algorithm's: you get a training set and then you have to make a model, a decision surface in some feature space that separates the yes from the no examples well enough, while still fulfilling certain smoothness etc. criteria. Of course we never do such things actually, but it's related.
When you get a new example, like a virus or prion, you have to decide on which side it falls.
The interesting question is: what happens if it turns out that one of your training examples were represented with some erroneous features? Or what if we discover some new features that could be relevant?
Do we still label those examples as animate/inanimate and update our decision surface accordingly, even if it makes the surface quite complicated? Or do we relabel them to so that we can keep the decision surface simple? Or do we keep its label and rather relabel some of its neighbors too, to make the decision surface simpler?
These are rhetorical questions, I'm trying to show how arbitrary the whole thing is. There is no "One True" label until we decide it and it's actually like an engineering trade-off decision between matching tradition/intuition and the simplicity of the definition. The former is like the training set error in machine learning, the latter is like regularization.
I think learning about CS and programming would be beneficial to philosophers as well as the other way around. Due to the required unambiguity of computer programs, many times programmers and CS people had to tackle issues like this. Philosophy is largely written in natural language, which results in a lot of ambiguity. For example, read the source code of a simple Quine (a program that prints out its own source code). You'll see code, code written in quote marks, things in quote marks escaped and nested into other other quote marks etc. I mean CS has developed concepts like currying, or variable scoping, which are very much related to philosophical issues. For example when you create and event handler, and say onMouseClick = function(){print x;} what do you mean by x, the current value of it, or the value when the event happens? Do you evaluate it now, or postpone it until the event? In natural language both sound the same, "you print x". But do you mean x as a symbol or do you actually mean the thing pointed at? Like asking "Do you think the president of the US will die in 2050?" can mean whether Obama will die in 2050, or the president incumbent in 2050.
The similarity of this to the "life" issue is whether we freeze the current meaning of "life", or we allow it to change.
So I think a lot of this confusion is just the result of not having to face these distinctions in natural language, while CS people have sorted out many of these things conceptually, in things like reification, reflection, virtualization etc. Of course it's not new, the "use-mention distinction" is well-known from earlier, but programming makes it really straightforward and obvious.
Of course, but I'd qualify this a bit. If you are in a life situation where you feel alone or in need of a community, you feel you're not included enough, or feel "smarter" and more reflective than your environment, then spending time at LW can amplify your smugness and stroke your ego to the point of not noticing what you are turning into. I saw many people there with milder psychological problems. Really, a psychologist would probably have an interesting time analyzing the "life advice/coaching/coping" parts of that forum.
However, if you're living a balanced, good life, reading this stuff won't matter much. It's similar in many ideologies. If you're in a receptive situation, then Scientology, or Reddit's Redpill, or basically any elitist, enlightening type of subculture along the lines of "you now understand what it all really is about" can be harmful.
One must keep the ability to laugh at oneself, to be able to humorously see a kind of futility in what one does but dance the dance anyway. Clenching too hard, making one more nested iteration in the prisoner's dilemma or the chicken game won't help.
Also don't take stuff more seriously than a certain ceiling.
For me, reading about Zen ideas helped me get out of this narrow, "rationalized", neatly ordered, algorithmic, packaged-and-labeled way of thinking. Metaphorically it's kind of the difference between a probabilistic machine learning system vs. a symbolic knowledge system where everything is defined precisely and unambiguously and every rule is laid out etc.
I'm still materialist and atheist, but I think being too deeply involved in any ideology is harmful (be it Marxism or Fascism or Scientology or LW). Yeah, you can reflect upon whether it's possible to live truly independent of ideologies without making or finding a new one for yourself. But there are certain indicators, like when you feel you're getting distanced away from the people around you physically, when you start feeling superior for belonging to the in-group etc., it usually means you're just not noticing some aspects and are obsessing over something. Sure one can say that if innovators thought like this and always stayed in line of the mainstream, we never would have gotten Ford or Jobs etc. And there is truth to this, but it kind of sounds like "X dropped out of college and went on to be successful therefore dropping out of college is a good idea". No, generally the good idea is to be humble and positive about the people around you, be open and reflective but not obsessive and be able to relax.
So if you eat these ideas at the right time, you won't find anything suspicious about him lacking any reputation, any qualifications. He's trying to sell you the idea that he's so much above the ridiculous ideas of the mainstream that he doesn't need to go among them an get tested by them, he tries to sell you that this is just signalling, the way a really rich person doesn't need to show off expensive items, or a big dog doesn't need to bark as much.
Another thing about the anti-inductive meme is that seeing the world thought these glasses makes you kind of paranoid. Because how can you know if the thing you conclude shouldn't be the exact opposite? You know, maybe the system figured out what you will think, it simulated you and now you are supposed to outsmart that system, etc. When in reality, the overwhelming majority of everyday things are straightforward and inductive and rarely go deeper than 1 or 2 levels of "simulating the other". But his focus on the superintelligent being makes you think subconsciously that this is the nature of the current world as well, and a lack of qualifications should actually be seen as a proof of skill, not as a counterevidence.
Maybe you say nobody would think that, but I did, some years ago. I got sucked into his writings and I was gullible. Since then I have learned a lot from other sources and can see much better how he rips off previous ideas, repackages them and creates a crazy ideology.
But Climate science is a lot less about an Apocalypse. Mainstream climate science isn't about a total wiping out of humanity but a few degrees of change in temperature, extreme weather, more deserts, floods etc., which are terrible catastrophes but nowhere near on the scale that Yudkowsky likes to dream up (evil robots destroying everything and torturing and extorting and creating myriads of simulations of simulations of you being tortured in the worst ways possible in various hypothetical, counterfactual scenarios to acausally motivate you to serve these evil overlords etc.)
Also, climate science doesn't ask you to "obviously" dismiss reputable fields and past sources. It's just a very bad analogy. The best defense of Singulitarianism is the "but what if you're wrong" kind of betting. Because their imagined doom scenario is so extremely unimaginably bad, they argue that they can get away with little proof, since multiplying the huge disaster with a little probability still gives a large expected risk. They like to say "shut up and multiply" as a slogan for this, i.e. multiply probability and outcome to get the expectation. He lays the groundwork and path for this such that if you gradually get into his system, he can take you in the woods without you noticing. It's no less than a panic-inducing mind virus.
And yeah, you can say I didn't disprove him with this. I don't want to. My main point is: don't start with his writings and don't recommend them to people who are uneducated in these topics. You'll do them a favor by recommending reputable sources.
So I won't attempt that here. Instead, I'll tell my (random internet guy's) feelings and vague ideas: You know when you meet a stranger and you feel something is off, but you don't really know what, but something's funny? Martial art and personal defense coaches usually recommend that you listen to this feeling and act accordingly.
Singulitarianism is basically an apocalyptic doom-religion. It follows the same scheme. For example, I just read a reddit thread (https://www.reddit.com/r/DebateReligion/comments/3vsoe7/chri...) the other day and the post reminded me of Singulitarianism a lot. There is an apocalypse, some external information that you should rely on instead of you intuitive faculties because they are biased and unreliable, but if you follow the path of the leader you'll be saved.
Also, its followers react to criticism similarly to religious people.
And another problem is that nobody serious has ever rolled up their sleeves, went through and criticized the whole mess of their Bible (called Sequences), a massive collection of writings, interlinked thousand-fold in an intricate complex network. Simply because the people who could do this don't care much about some random guy posting stuff to his blog.
Fortunately since the Basilisk story came out and we could see Yudkowsky's reaction, it was an eye-opener for many people. Still his movement is quite well-spread and gets linked to often by CS people, mathematics students etc.
And I won't start to refute individual claims. I admit this. The thing is not about the individual claims but the way they are linked and the kind of narrative it builds. The untold assumptions lurking in the background etc. The way he tries to take credit for age-old ideas by renaming them and not caring about the origin of them, since the history of philosophy is irrelevant anyway, we should just read his blog posts and related stuff.
My advice for novices is to start with the mainstream. You can of course criticize the mainstream once you have sufficient knowledge. Read textbooks, classical works, go to universities, take interesting courses etc. Don't teach yourself this stuff from a random guy's blog who has no qualifications on anything he speaks about (which isn't a disproof but a red flag).
I've been vague and I stand by it. Yes I might have been irrational in my comment or led by feelings and emotions. I don't care. I just had to put this out here. I don't aim my paragraphs as ultimate refutation but as a warning. I felt I had to write this for the benefit of people who may be new to this man's world.