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

Just A New Fractal Detail In The Big Picture (2015)

edge.org

Post reply on HN
Print view View original

171–180 of 180 posts

Re: Just A New Fractal Detail In The Big Picture (2015)

#171

Post by atemerev »

prof_hobart wrote: post #153
atemerev wrote: post #39

As a curious software engineer, I have a basic familiarity on where heating oil comes from (I remember my first sci-pop book with a section on basics of oil processing), where nuts are grown, basics of food industry supply chains, and every other example covered. My knowledge might be rudimentary and without much detail, but I try to cover all bases, and when I discover some area of human knowledge I genuinely know n…

Curiosity in general is probably a fundamental part of human nature. But you have to make a call at what point you stop trying to understand everything about everything. Thomas Young, the early 19th Century scientist has been famously described as "The Last Man Who Knew Everything" - the sum of human knowledge moves on so vastly that it is simply physically impossible to have an in-depth knowledge of every possible s…

Of course, but there is (loosely defined) hierarchy of knowledge.

I haven't read (until now;)) the article about Four Hu (a specific topic), but I did read on 214 Kangxi radicals (more general/beginner friendly topic), and, of course, on Mandarin and Cantonese in general (I can't speak neither).

This is how it works. Curiosity is insatiable.

Re: Just A New Fractal Detail In The Big Picture (2015)

#172

Post by scotty79 »

Houshalter wrote: post #160
scotty79 wrote: post #156

Earlier quoted context omitted.

> main requirement for manipulating matter is intelligence Rather, it's knowledge. And to get more knowledge you need to manipulate matter. Pure intelligence is useful up to a point but the you need to go out and get more data. That's the limiting factor I think of. The only thing you can do with just intelligence is math (that can have some use later on) and philosophy (that's totally useless). > The best is physici…

>The only thing you can do with just intelligence is math (that can have some use later on) and philosophy (that's totally useless). We already have vast amounts of knowledge on the internet. I can download all of wikipedia in an hour, and fit it on a flash drive. Most of the world's scientific papers and books are digitized and available. The limiting factor is no longer knowledge. It's the ability to absorb knowled…

> The limiting factor is no longer knowledge. It's the ability to absorb knowledge.

I get your point that from all experiments we performed so far and bothered to write down and digitize there might be few tricks left to squeeze out, but vast majority of our current progress comes from new experiments.

I agree that AI could very well write stuxnet or mess with our security so I'm hoping we get bit more tight in that department before we manage to build AI. I'd definitely prefer first artificial consciousness to be built before robots are as popular as cars or smartphones.

We will eventually develop AI and it'll be definitely challenging to orchestrate it running most of our civilization and not killing us in the process.

What I'm not afraid is AI taking what we know so far and secretly turning itself into miracle making god in matter of weeks. We will have few years or decades of existence before AI becomes vastly more powerful than the rest of us and till then we will have ability to align our priorities.

> Human brains run at maybe 100 hz. AI's built out of silicone could work thousands of times faster. Doing the same work, just in much shorter time.

It's not work. It's just thinking. AI can create philosophical theories or build new optimal JavaScript frameworks at blazing speed and still not make any progress in physics that would make it more powerful than us. It needs new data. It needs to run experiments to get anywhere beyond what we achieved so far. If we maintain transparency and caution in what AI is given and allowed to do then we may have safely transition from current world to AI world. Besides, at some point our civilization will encounter others it's better to bring our own AI to the table. Alien one might attach much less sentimental value to us.

> If you went back in time to Blaise Pascal, if you took the right books and plans from the future, you could get them to CPUs in mere years.

No, you couldn't. Do you know how long it takes to build a fab with current infrastructure available? You'd have to bring half or their industry into XX century before you could have a CPU.

> If you can predict prices 1% more accurately than anyone else, then you can make a huge amount of money in the long run.

Yes. The thing with stock market is that you can't. It's not because you are not smarter than other traders. It's because interaction between all the traders creates pretty much perfect chaotic random process that no-one can predict.

> Stock prices aren't random. They are determined by real events, mainly how much profit the company makes.

Same way random generator driven by unknown algorithm is predicted by its seed.

Re: Just A New Fractal Detail In The Big Picture (2015)

#173

Post by mastazi »

steplee wrote: post #110
mastazi wrote: post #85

Earlier quoted context omitted.

> This is a road to madness How so? > If every time you wouldn't understand something you would just plunge into the material and get familiarized you would do nothing else for the rest of your life That's how I've been spending a significant part of my out-of-work time, for most of my adult life. Is there anything wrong with being a curiosity-driven person?

There's no way you could actually do this. Going an hour on a normal day you'd have a life time of material to learn so I agree yes it is madness. You can't learn thousands of years of engineering, math, science, etc.

Of course you can't learn everything, but you can keep your curiosity alive and keep questioning reality, and search for explanations. To me, it's not about learning the whole corpus of human knowledge.

Re: Just A New Fractal Detail In The Big Picture (2015)

#174

Post by xrange »

roymurdock wrote: post #73
Rauchg wrote: post #45

For different versions of this argument see: https://en.wikipedia.org/wiki/AI_effect . Bostrom said: "A lot of cutting edge AI has filtered into general applications, often without being called AI because once something becomes useful enough and common enough it's not labelled AI anymore". Finally, the problem with all these articles is the lack of precise definition for the terminology involved (it reminds me of deb…

A lot of the arguments/heated debates on HN stem from a failure to define key terms. Less so in threads on compsci-related topics, but almost always on threads with econ-related topics. Especially universal basic income. Once you identify this common issue it becomes easy to avoid these frustrating (and frustrated) conversations altogether. Eno has taken a pretty liberal view of "AI" but he makes a good point about s…

>A lot of the arguments/heated debates on HN stem from a failure to define key terms. ... Especially universal basic income.

Can you recommend some definitions that are a good starting place for basic income discussions that we could helpfully point to when discussing that issue?

Re: Just A New Fractal Detail In The Big Picture (2015)

#175

Post by sixQuarks »

mastazi wrote: post #87
sixQuarks wrote: post #10

I don't see the correlation. The author is talking more about systems. The difference with AI is that it will have all of the knowledge within a system and thus be able to make incredible connections between disparate fields of knowledge. On a separate note, I had an interesting thought about AI as I was looking at a cool architectural photo. I would love to design a building, but I don't know a thing about it. It wo…

> The difference with AI is that it will have all of the knowledge within a system If you asked me to come up with 5 different definitions of AI, the concept of "all of the knowledge" wouldn't really be there. I agree on the "incredible connections" part, though.

"all of the knowledge" is a huge part of AI. The bigger the datasets, the better results you have from machine learning, and it's a trivial thing for AI to absorb huge amounts of information. What humans find difficult, computers can do easily and it's usually vice-versa for stuff like pattern recognition, creativity.

Re: Just A New Fractal Detail In The Big Picture (2015)

#176

Post by fuzzfactor »

bogomipz wrote: post #84

I love Eno. Such an influential composer and yet he doesn't consider himself a musician in the least. He's used his own ignorance of music to spectacular effect. This a piece the Telegraph did on him a while ago. It's worth reading: http://www.telegraph.co.uk/music/artists/how-brian-eno-creat...

Also the author of The Microsoft Sound WAV that played upon Windows startup in 1995.

At the time probably the tune played most often per day for a number of years.

Re: Just A New Fractal Detail In The Big Picture (2015)

#177

Post by Houshalter »

scotty79 wrote: post #172
Houshalter wrote: post #160

Earlier quoted context omitted.

>The only thing you can do with just intelligence is math (that can have some use later on) and philosophy (that's totally useless). We already have vast amounts of knowledge on the internet. I can download all of wikipedia in an hour, and fit it on a flash drive. Most of the world's scientific papers and books are digitized and available. The limiting factor is no longer knowledge. It's the ability to absorb knowled…

> The limiting factor is no longer knowledge. It's the ability to absorb knowledge. I get your point that from all experiments we performed so far and bothered to write down and digitize there might be few tricks left to squeeze out, but vast majority of our current progress comes from new experiments. I agree that AI could very well write stuxnet or mess with our security so I'm hoping we get bit more tight in that…

>I get your point that from all experiments we performed so far and bothered to write down and digitize there might be few tricks left to squeeze out, but vast majority of our current progress comes from new experiments.

I don't think so. We already know the laws of physics to a great degree. An AI or even human could design all sorts of amazing things without ever doing a single experiment.

Of course, there's no reason it can't do experiments, also. Once it's free on the internet, it need only contact some random Joe and bribe/threaten/persuade them to do it's bidding.

>No, you couldn't. Do you know how long it takes to build a fab with current infrastructure available? You'd have to bring half or their industry into XX century before you could have a CPU.

Perhaps. This doesn't seem to be true of most technologies though. You could go to 1800 London and show them how to build a modern car or airplane in a year or so. You could introduce everything from antibiotics to radios, centuries before they were actually invented. Ancient romans could have built crude steam engines, and bootstrapped industry in a century, if they had known how.

Building a CPU might require first building multiple other industries to support it, but that can be done. If the AI lays out in painstaking detail, every step necessary to construct every tool, every machine. And yes it would take a lot of labor, but they would be able to do it.

It seems impossible to us, because we can't imagine that kind of complexity. Humans are terrible at managing complex systems. We overlook or forget details, we don't account for possible mistakes, etc. No single human even knows every step necessary to build a pencil, because we specialize so much. An AI could be aware of every detail, of every step in the process, and manage it at terrifying efficiency.

Re: Just A New Fractal Detail In The Big Picture (2015)

#178

Post by vec »

xrange wrote: post #174
roymurdock wrote: post #73

Earlier quoted context omitted.

A lot of the arguments/heated debates on HN stem from a failure to define key terms. Less so in threads on compsci-related topics, but almost always on threads with econ-related topics. Especially universal basic income. Once you identify this common issue it becomes easy to avoid these frustrating (and frustrated) conversations altogether. Eno has taken a pretty liberal view of "AI" but he makes a good point about s…

>A lot of the arguments/heated debates on HN stem from a failure to define key terms. ... Especially universal basic income. Can you recommend some definitions that are a good starting place for basic income discussions that we could helpfully point to when discussing that issue?

There are actually two slightly different ideas that are called UBI.

The libertarian variant is a relatively small guaranteed income (usually roughly the current minimum wage), is usually paid for with a VAT or sales tax, and replaces most or all of the current welfare programs. The main objectives are to provide a safety net for the poorest citizens in a fairer and less bureaucratic manner and to somewhat equalize the bargaining power between low-wage workers and their employers, usually as an alternative to unionization.

The socialist variant is significantly larger (living wage), is usually paid for with graduated income or capital gains taxes, and supplements rather than replaces the existing welfare programs. Its main goals are to provide a direct mechanism for income redistribution and to facilitate non-profitable professions (artist, full-time volunteer, professional student, stay at home caretaker, etc.) for those that want or need them.

There's obviously a lot of overlap between the two, and most proponents support both sets of objectives to some degree. That said, most of the contentious arguments seem to derive from critics either not understanding the difference between the two proposals or strawmanning their interlocutors into one extreme or the other.

Re: Just A New Fractal Detail In The Big Picture (2015)

#179

Post by JoBrad »

getgoingnow wrote: post #124
JoBrad wrote: post #63

Earlier quoted context omitted.

It's generally accepted that any intelligence that isn't from a living being is artificial. Global Civilization isn't living, per se.

Could you offer an abstract definition of "life" and "intelligence" (or maybe even 'thinking' or 'consciousness') ? For example, if you're trying to figure out whether something is 'moving', you would define the term in such a way that it could apply to planets or animals or other things/systems. If you define 'motion' = 'humans putting one leg in front of another', then of course non-human animals or anything else c…

No, I couldn't. It's terribly difficult to define, and the more you try, the harder it becomes. I think it's a bit easier if you try to generally define it in terms of what we think it isn't: the average layperson would consider any intelligence created by humans to be artificial, such as "artificial cells" or other "AI" like Watson. That's what I meant by "generally accepted" but I realize that my comment reflected very little of the thoughts I had when I made it.

Using that definition, Global Civilization, regardless of how complex it has become, is nearly entirely created by humans. Ignoring the "human" part of Oxford's definition for civilization[1], you could say that certain groups of animals do have a form of civilization: some dolphins and killer whales, for example, teach their children very specialized hunting and recreational methods that others within their species do not know. But no other animal engages in anything at the scale of Global Civilization. In that sense, Global Civilization is an artificial construct (intelligence, if you like).

I'm not arguing that it isn't intelligence. It's almost like a virus, really, in that it can't do anything by itself (it requires humans), but it certainly has life-like characteristics when powered by a "host."

[1] http://www.oxforddictionaries.com/us/definition/american_eng...

Re: Just A New Fractal Detail In The Big Picture (2015)

#180

Post by mastazi »

sixQuarks wrote: post #175
mastazi wrote: post #87

Earlier quoted context omitted.

> The difference with AI is that it will have all of the knowledge within a system If you asked me to come up with 5 different definitions of AI, the concept of "all of the knowledge" wouldn't really be there. I agree on the "incredible connections" part, though.

"all of the knowledge" is a huge part of AI. The bigger the datasets, the better results you have from machine learning, and it's a trivial thing for AI to absorb huge amounts of information. What humans find difficult, computers can do easily and it's usually vice-versa for stuff like pattern recognition, creativity.

> The bigger the datasets, the better results you have

Wouldn't you agree that the ability to achieve relatively good results with relatively small datasets is one of the main reasons for the current ML boom?

> it's a trivial thing for AI to absorb huge amounts of information

Of course, I agree on that. But I also note that "traditional" (non-ML) algorithms can digest big data too, so I don't see that ability as the defining characteristic of the current wave of ML-related technologies.

Post reply on HN