= 27d567bc2d48d9f40e9ccc20ea370b15)
New Year's Predictions Thread
I would like to propose this as a thread for people to write in their predictions for the next year and the next decade, when practical with probabilities attached. I'll probably make some in the comments.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
Comments (426)
My first prediction is that as is usually the case, political and random events will change the way people live far more over the next year than technology will. Given the current state of the financial system, I would place about even odds on politics having more impact than technology over the next decade, but with the caveat that over such a long time scale political and technological events will surely be interwoven.
There's no separation to be had between politics and technology.
The biggest influence on technology is regulation which outlaws, restricts, or places huge financial barriers to entry (as with medical research); another non-trivial influence is politically controlled financing of R&D.
And arguably, the biggest influence on politics that isn't itself political is technology (case in point: modern communications, computer, and the Internet spreading censored information, creating more popular awareness and coordinating protests.)
So I think political and technological events are inseparable over almost any timescale.
I agree that there is little to no separation, but I think a distinction can be made. Namely, there are two different words that mean different things. When predicting what is going to affect people you can probably find a way to split the techno-political mash usefully. This may be as simple as using one word over the other.
It seems pretty vague - do you have any ideas about how this should be measured?
My second prediction is that the largest area of impact from technological change over the next decade will come from increasing communications bandwidth. Supercomputers a hundred times more powerful than those that exist today don't look revolutionary, while ubiquitous ultra-cheap wireless broadband makes storage and processing power less important. Improvements in small scale energy storage, tech transfer from e-paper and lower power computer chips will probably help make portable personal computers more energy efficient, but for always-on augmented reality (and its sister-tech robotics) in areas with ubiquitous broadband computing off-site is the way to go.
Agree on the trend, but I'd put significant odds on some (as yet unexpected) trend being "the largest area of impact" in retrospect.
Latency worries me, though. Bandwidth has been improving a lot faster than latency for a while now. For always-on augmented reality, I think that we're going to need some seriously more power-efficient computing so we can do latency-limited tasks locally. (Also, communication takes energy too -- often more than computation.)
Good news on that, by the way: modern embedded computer architecture and manufacturing techniques are going in the right direction for this. 3D integration will allow shorter wires, making all digital logic much more power efficient. Network-on-chip architectures will make it easier to incorporate special-purpose hardware for image recognition and such. And if you stick the memory right on top of your processor, that goes a long way to speeding it up and cutting down on energy used per operation. If you want to get even more radical, you could try something like bit-serial asynchronous processors (PDF) or something even stranger.
</nerding-out>
Do you have any ideas about how the scale of the impact from various different technological changes should be measured in this context? As far as I know, there is no standard metric for this. So, I am not clear about what you mean.
And distributed to more people. >60% of people will have at least 1 Mb/s internet access by 2020 (75%).
For the next decade:
I'd bet about a 2:3 odds that energy consumption will grow on a par or less than population growth.
Any rise in average standard of living will come from making manufacturing/logistics more efficient, or a redistribution from the very rich to the less well off. There is still scope for increased efficiency by reducing the transport of people and more automation.
I'd take the other side at those odds. Per capita energy expenditures in China are set to skyrocket as rural areas industrialize, and I expect the same of many Second World nations. I don't think increases in efficiency will dwarf that effect quite yet.
I'm basically betting that a short term lack of oil (as evidenced by reduced production in 2008 and high current price), will put a break on that expansion. Or the industrialization of china will only happen in if first would countries reduce their energy consumption to allow it, as they did in 2008.
Data from the BP energy review.
Interesting consideration; but on the other hand, China isn't afraid to build nuclear power plants or burn coal.
An interesting article on china and energy. Nuclear has a lead time (optimistically ) of 3 years, so their prediction of 60-90 GWe won't be too far off. It actually looks like they are planning more wind than nuclear. I'm really curious where they expect the 500 GWe odd of energy they don't mention to come from. All coal? That'll be pretty dirty.
I was probably a little overconfident in my initial bet. I do expect the ratio of energy consumption growth to population growth to trend downwards though.
The article estimates that China's electricity capacity will double from 2008 to 2020; it doesn't seem to list an estimate for electricity production, but I'd think it would trend in much the same way, significantly faster than China's (rapidly falling) population increase. Reading this article makes me even more eager than before to take the "over" at these odds.
I'm rethinking my wager. To give you some information that I found. Which I should have looked at before.
Average energy consumption increase over 15 years to 2008 has been 2.13%. This is very choppy data it varies between 0.09% and 4.5%(2004 then trending downwards). This included a doubling on energy consumption by china in 7 years (2001-2008).
Average population growth is trending downwards and is at 1.1%.
I was probably putting too much weight on my own countries not very well thought out energy policy.
What odds would you give on energy consumption growth rate being lower for the next 10 years than the previous 10 (2.4%)?
Because of the Second World's larger growth rate (and the fact that they occupy a larger part of the total now), I think the odds of energy growth being lower than 2.4% are somewhat worse than even. I'm quite metauncertain; I don't think I'd actually bet unless someone were giving me 3:2 odds to bet the 'over', or 4:1 odds to bet the 'under'.
Wrong. (Well, a little bit right, but wrong in all the ways that matter.) According to the article you linked, they're planning to build about 60-90 GW of nuclear capacity (let's say 80 GW to simplify the arithmetic) and 100 GW of wind. But what we really care about is how much energy they get from those sources per year, and to find that, we have to multiply the peak power generation capacities by the capacity factor for each source.
Nuclear power has a capacity factor of at least 93% for the newer plant designs that China is building (or even for older plants after operators get experience), so we'll say that their average production is (80 GW) * 0.93 = 74.4 GW average.
Wind power has a capacity factor of around 21% right now. Since we're talking about 2020, i.e. The Future!!, let's assume they get it up to a whopping 30%. Their energy production from wind would come out to (100 GW) * 0.3 = 30 GW average, or less than half of their projected nuclear production.
The average power figures are much more meaningful than the capacity numbers, but the wind salesmen quote whatever numbers make them sound most impressive, and the news media report it. It's as ubiquitous as it is misleading.
Mea culpa. I forgot how misleading some of the energy numbers could be.
A killer application for augmented reality is likely to be the integration of communication channels. Today's, cellular phones annoy people with constant accountability and stress, not to mention spotty coverage, but if a HUD relay over life can display text messages as they are sent and invite fluid shifts to voice conversation. When video is engaged and shared, people could also see what their potential conversation partner is doing prior to requesting attention, giving distributed social life some of the fluidity and contextual awareness of natural social life. These sorts of benefits will motivate the teenagers of 2020 to broadcast much of their lives and to interpret the absence of their friend's data streams as a low intensity request not to call. Archival will at first be a secondary but relatively minor benefit from the technology, but will ultimately widen the divide between public and private life, a disaster for privacy advocates but a boon for academic science (by normalizing the publication of all data). Paranormal beliefs will also tend to decline, as the failure to record paranormal events and the fallibility of memory both become more glaring.
On a AR theme I think there will be a high level language created within ten years for AR that will try to make the following accessible
People will want to mash up different AR services in one "view" so you don't have to switch between them. There needs to be a lingua franca and HTML doesn't seem suited. I'd think it likely that it will be some XML variant.
Aren't these more likely to be done by libraries than languages?
I hope not. Something like JSON is far less verbose.
If AR gets any sort of popularity, even just among early adopters, I guarantee you that there will be several competing tools for doing what you describe, with more coming out every month.
It already has a sort of popularity. There are already startups working in the field.
If you want to keep abreast of the field keep an eye on Bruce Sterling's Blog.
There are already plenty of supposedly "paranormal" events recorded on Youtube, as well as elsewhere. With the increase of recording devices, many more such things will be recorded, and paranormal beliefs will increase.
I think these are great predictions.
Robin Hanson makes a similar prediction in 'Enhancing Our Truth Orientation' (pp. 362-363):
We will end the decade with some mobile energy storage system with an energy density close to or better than fat metabolism.
ETA: I mean in the context of electronics.
The graph you link to says magnesium and diesel already have greater energy density than fat.
So, I think you have to specify how portable, how common or cheap, and maybe whether you are talking about rechargable or not - or the prediction is probably going to be vague - and subject to the criticism that it has already happened.
I meant commonly used for powering portable electronics. I don't assign a high probability to this. It is the upper bound of what I think worth discussing.
Right. TNT does not count as a mobile energy storage system.
I think you're wrong; but it's a really interesting prediction.
The reason I think you're wrong is that the rate of improvement of technologies in a field is more-or-less fixed within a field, because it depends on the economics, not on the science. Moore's Law exists not because there's some magic about semiconductors, but because the market is sized and structured such that you need to sell people a new system every 2 years, and you need to double performance to get people to buy a new system.
This means you can look at the past exponential curve for battery density, and project it into the future with some confidence. I don't know what the exponent per year is; but my gut feeling before checking any data or doing any calculations is that it isn't high enough.
I don't think there is an exponent curve as such for battery tech. Li-ion came in about 2006? And nothing much has improved since then. The trouble with batteries is you can't just shrink components and get some improvement as you do with semi-conductors. Your components are already on the atomic scale. So more fundamental breakthroughs are needed.
The prediction is based mainly on our increasing control of biology and the ability to work on the small scale. If nothing else we'll invent a way to metabolise fat or other carbohydrates to electricity and have small home bioreactors that produce carbs and make nice little cartridges for people to plug into their electronics. Maybe not in 10 years, but some substantial movement is definitely possible in this direction.
What about some of the advances in micro-generators and Fuel-Cells that I have read about?
For instance, I have seen one of those tiny turbine engines running to power an equally tiny generator, and it looked to provide a hellofa lotta power for its size. I know the military is putting them into some applications in the field, so it will probably not be too terribly long before we see them on things like Laptops/tablets or cell phones.
I haven't seen anything recent on these. Any keywords to google? The key thing for a consumer electronics application is ease of getting the fuel. People don't want to have to head out to the shops to get it every few days, which is why rechargeable batteries are the current winner.
I disagree.
I am typing this on a machine I bought 6 years ago. Its CPU speed is still competitive with current hardware. This lack of speedup is not because processor manufacturers chaven't been trying to make processors faster; they have. The reason for the lack of speedup is that it is hard to do. The problem is more to do with the nature of physical reality than the structure and economics of the computer industry.
Consider cars. They do not halve in price every two years. Why not? Because they are designed to move people around, and people are roughly the same size they have always been. But computers move bits around, and bits can be made very small (both in terms of the size of circuitry and the power dissipated); this is the fundamental reason why the computer/communications industry has been able to halve prices / double capabilities every year or two for the last half century.
From looking at the diagram, aren't we starting the decade with such a system (gasoline)?
You are the second person to mistake my intent. I meant in the field of mobile electronics. Take a look at where lithium ion is on this chart.
In an analysis that does not account for any health-care reform bill, the Department of Health and Human Services projected that health care expenditures would double from the 2009 level of $2.2 trillion (16.2% of 2009 GDP) to $4.4 trillion in 2018 (20.3% of projected 2018 GDP). This provides us a baseline from which to predict the cost-control effectiveness of health care reform.
I'm somewhat bullish on the potential of the pilot programs and the excise tax to lower med costs for a given level of health outcomes, although I'm not supremely confident in that. I also think there is a long tail of events or technologies that could unexpectedly increase med expenses (that would do so with or without health-care reform). Furthermore, the current bill will expand coverage for a substantial number of people, as a result of which total expenditures will definitely rise. All things together, here are my (very rough) intuitions:
EDIT: I had this up for a few minutes with different numbers, before I remembered that the individual mandate and subsidies would raise med expenses significantly.
I am 99% confident that AGI comparable to or better than a human, friendly or otherwise, will not be developed in the next ten years.
I am 75% confident that within ten years, the Bayesian paradigm of AGI will be just yet another more or less useful spinoff of the otherwise failed attempt to build AGI.
Shane Legg gives a 10% probability of that here:
http://www.churchofvirus.org/bbs/attachments/agi-prediction.png
My estimate here is a bit bigger - maybe around 15%:
http://alife.co.uk/essays/how_long_before_superintelligence/graphics/pdf_no_xp.png
You seem to be about ten times more confident than us. Is that down to greater knowledge - or overconfidence?
You seem to be about ten times less confident than me. Is that down to greater knowledge - or underconfidence?
I'm not very confident - primarily because we are talking ten years out - and the future fairly rapidly turns into a fog of possibilities which makes it difficult to predict.
Which brings us back to why you seem so confident. What facts, or observations are the ones you find which provide the most compelling evidence that intelligent machines are at least ten years off. Indeed, how do you know that the NSA doesn't have such a machine chained up in its basement right now?
It hasn't worked in sixty years of trying, and I see nothing in the current revival to suggest they have any ideas that are likely to do any better. To be specific, I mean people such as Marcus Hutter, Shane Legg, Steve Omohundro, Ben Goertzel, and so on -- those are the names that come to me off the top of my head. And by their current ideas for AGI I mean Bayesian reasoning, algorithmic information theory, AIXI, Novamente, etc.
I don't think any of these people are stupid or crazy (which is why I don't mention Mentifex in the same breath as them), and I wouldn't try to persuade any of them out of what they are doing unless I had something demonstrably better, but I just don't believe that collection of ideas can be made to work. The fundamental thing that is lacking in AGI research, and always has been, is knowledge of how brains work. The basic ideas that people have tried can be classified as (1) crude imitation of the lowest-level anatomy (neural nets), (2) brute-forced mathematics (automated reasoning, logical or probabilistic), or (3) attempts to code up what it feels like to be a mind (the whole cognitive AI tradition).
My estimates are unaffected by hypothetical possibilities for which there is no evidence, and are protected against that lack of evidence.
Besides, the current state of the world is not suggestive of the presence of AIs in it.
ETA: But this is becoming a digression from the purpose of the thread.
Thanks for sharing. As previously mentioned, we share a generally negative impression of the chances of success in the next ten years.
However, it appears that I give more weight to the possibility that there are researchers within companies, within government organisations, or within other countries who are doing better than you suggest - or that there will be at some time over the next ten years. For example, Voss's estimate (from a year ago) was "8 years" - see: http://www.vimeo.com/3461663
We also appear to differ on our estimates of how important knowledge of how brains work will be. I think there is a good chance that it will not be very important.
Ignorance about NSA projects might not affect our estimates, but perhaps it should affect our confidence in them. An NSA intelligent agent might well remain hidden - on national security grounds. After all, if China's agent found out for sure that America had an agent too, who knows what might happen?
I would guess that the NSA is more interested in quantum computing than in AI.
They are the National Security Agency. Which of those areas presents the biggest potential threat to national security? With a machine intelligence, you could build all the quantum computers you would ever need.
This is my sense as well. I also think there is a substantial limit on what we're likely to learn about the brain given that we can't study brain functionality with large scope, neuron-level definition, in real time given obvious ethical constraints. Does anyone know of any technologies on the horizon that could change this in the next ten years?
http://lesswrong.com/lw/vx/failure_by_analogy/
All that is fine.
But 1) I'm not sure anyone has a good grasp of what the properties we're trying to duplicate are. I'm sure some people think they do and it is possible someone has stumbled on to the answer but I'm not sure there is enough evidence to justify any claims of this sort. How exactly would someone figure out what general intelligence is without ever seeing it in action? The interior experience of being intelligent? Socialization with other intelligences? An analogy to computers?
2) Lets say we do have or can come up with a clear conception of what the AGI project is trying to accomplish without better neuroscience. It isn't then obvious to me that the way to create intelligence will be easy to derive without more neuroscience. Sure, from just from a conception of what flight is it is possible to come up with solutions to the problem of heavier than air flight. But for the most part humans are not this smart. Despite the ridiculous attempts at flight with flapping wings I suspect having birds to study --weigh, measure and see in action-- sped up the process significantly. Same goes for creating intelligence.
(Prediction: .9 probability you have considered both these objections and rejected them for good reason. And .6 you've published something that rebuts at least one of the above. :-)
The NSA does have some scary machines chained in their "Basement," yet I doubt any of them approach AGI. All of them(that I am aware of - so, that would be 2) are geared toward some pretty straightforward real-time data mining, and I am told that the other important gizmos do pretty much the same thing (except with crypto).
I doubt that they have anything in the NSA (or other spooky agencies) that significantly outstrips many of the big names in Enterprise. After all, the Government does go to the same names to buy its supercomputers that everyone else does. It's just the code that would differ.
So: you have a hotline to the NSA, and they tell you about all their secret technology?!? This is one of the most secretive organisations ever! If you genuinely think you know what they are doing, that is probably because they have you totally hoodwinked.
Can you be more specific about what you mean by the Bayesian paradigm of AGI? Is it necessarily a subset of good-old-fashioned symbolic AI? In that case, it's been dead for years. But if not, I can't easily imagine how you're going to enforce Bayes' theorem; or what you're going to enforce it on.
For the next decade: collaborative filtering.
Based on this article on collaborative filtering, we already have it. Every time I buy anything online, I am told what other products people buy who also bought what I bought. It is the central component of the StumbleUpon service.
So, what are you predicting?
I'm predicting that in 2020 you'll look back at this blog comment and say, "Wow, he sure called that one."
I think it's more likely people will say "too vague a prediction".
I don't know. It makes sense; but I thought the same thing in 1999.
There is a lot of interest in using CF to sell people more mass-market things, eg Netflix; less interest in helping people find obscure things from the long tail that they might have a special interest in; still less interest in using CF for social networking.
Better than even odds that in 2020:
GDP per capita at purchasing power parity for Singapore will be more than US$80,000 in 2008 dollars.
GDP per capita for China (PRC), will be more than twice 2009 GDP
Tourism to suborbital space will cost less than $50000.
I estimate 90% odds that Emotiv's EPOC will fail like the Segway did.
I have one of these puppies. It's the most fickle device I've laid my hands on. It's useless for anything except gaining nerd status points. Hey, do you guys want me to post a detailed review? :)
I'd like to see a review, but it isn't a LW thing. It would be nice to have a forum / news structure, so that we could have a section for "Off-topic posts". Heck, it would be nice to sort the posts by topic.
Isn't that what tags are for?
Next Year
Next Decade
I would be very happy to accept a bet with you on those odds if there's a way to sort it out. I'd define major as any attack with more than ten deaths.
Do you have a PayPal account? I'd be willing to wager $50 USD to be paid within 2 weeks of Jan 1st 2011 if you're interested. I can provide my email address. That would rely on mutual trust but I don't know of any websites that can act as trusted intermediaries. Do you know of anything like that?
For $50, trust-based is OK with me.
How about this wording? "10 or more people will be killed on US soil during 2010 as the result of a deliberate attack by a party with a political goal, not overtly the act of any state". And if we hit an edge case where we disagree on whether this has been met, we'll do a poll here on LW and accept the results of the poll. Sound good?
I'd like to change the wording slightly to "on US soil, or on a flight to or from the US" if that's alright with you (even though I think an attack on an aircraft is less likely than an attack not involving aircraft). A poll here sounds like a fair way to resolve any dispute. I expect to still be reading/posting here fairly regularly in a year but I'm also happy to provide my email address if you want.
Do you think this was a terrorist attack? http://en.wikipedia.org/wiki/Fort_Hood_shooting
Excellent question! If such an attack happens this year, I'd say it wasn't a terrorist attack, but if mattnewport felt that it was I'd pay out without making a poll.
Fine with me. My email is paul at ciphergoth.org. How exciting!
Re: "10 or more people will be killed on US soil during 2010 as the result of a deliberate attack by a party with a political goal, not overtly the act of any state".
How come "Pakistan" got dropped? A contributing reason for the claim being unlikely was that it was extremely specific.
What makes your think 2010 is the year? I mean, this has even been floating around lately. And at 99%^h^h^h50% confidence!
That was 99% confidence that the response will be disproportionate to the magnitude of the attack, if an attack takes place, not 99% confidence that there will be an attack. My odds of an attack were 50%. I think an attack is fairly unlikely to be on an aircraft - security is relatively tight on aircraft compared to other possible targets.
I'll agree that if anything happens, or even if something doesn't (is thwarted), the response will be silly and disproportionate. However, I still think you're way too high with 50%.
You must specify disproportionately high, or disproportionately low.
A declaration of war, curtailment of liberties, or other expenditure of resources more than ten times the loss of resources (including life, which is not priceless) it tries prevent.
Is there a standard method for assigning a numerical value to liberties?
The money those people would pay to avoid the loss of liberty, had they the option.
That's a valid measure, but it would require a fairly complicated study to actually get a value for it.
I voted all the betting comments up because I think this is awesome. Does this kind of thing happen often here?
I occasionally offer people bets, but I think this has been the first time for me that the subject of contention is the right shape for betting to be a real possibility.
You display a pessimism much greater than I think is warranted. My predictions for some of your statements:
Next decade:
Next year:
These seem overly optimistic to me. Maybe increase the numbers by 50% to 100% other than 99?
Great example of what I'm talking about. I'd challenge you on most of those actually, if there was a convenient and well structured betting forum, but none of them seem crazy to me.
None of the others do, but this one seems ludicrous to me.
Well under 30% certainly, but I wouldn't give it under 4%. A decade is long and the US is young.
I think a draft is much more likely.
30% probability might be around the point where we start to call things ludicrous. If you talk seriously about things that you think have a 10% chance of happening, you will be beyond the point where most people call it ludicrous, or even crazy; they simply will not understand or believe that that's what you mean.
This comment provides more confirmation for a view I've held for a long time, and which was particularly reinforced by some of the reactions to (the first version of) my Amanda Knox post.
People have trouble distinguishing appropriately among degrees of improbability. This generalizes both underconfidence and overconfidence, and is part of what I regard as a cluster of related errors, including underestimating the size of hypothesis space and failing to judge the strength of evidence properly. (These problems are the reason that judicial systems can't trust people to decide cases without all kinds of artificial-seeming procedures and rules about what kind of evidence is "allowed".)
The reality is that given all the numerous events and decisions we experience on a daily basis and throughout our lives, something with a 10% chance of happening or being true is something that we need to take quite seriously indeed. 10% is, easily, planning-level probability; it should attract a significant amount of our attention. By the same token, something which isn't worth seriously planning on shouldn't be getting more than single digits of probability-percentage, if that.
There is a vast, huge spectrum of degrees of improbability below 1% (never mind 10% or 30%) that careful thinking can allow us to distinguish, even if our evolved intuitions don't. Consider for instance the following ten propositions:
(1) The Republicans will win control of both houses of Congress in the 2010 elections.
(2) It will snow in Los Angeles this winter.
(3) There will be a draft in the U.S. by 2020.
(4) I will be dead in a month.
(5) Amanda Knox (or Raffaele Sollecito) was involved in Meredith Kercher's death.
(6) A U.S. state will make a serious attempt to secede by 2020.
(7) The Copenhagen interpretation of quantum mechanics, as opposed to the many-worlds interpretation, is correct.
(8) A marble statue has waved or will wave at someone due to quantum tunneling.
(9) Jesus of Nazareth rose from the dead.
(10) Christianity is true.
I listed these in (approximately) order of improbability, from most probable to least probable. Now, all of them would be described in ordinary conversation as "extremely improbable". But there are enormous differences in the degrees of improbability among them, and moreover, we have the ability to distinguish these degrees, to a significant extent.
The 10%-30% range is for propositions like (1) ; the 1%-10% range for things like (2) (the last time it snowed in LA was in the 1960s). Around 1% is about right for (3). Propositions (4), (5), and (6) occupy something like the interval from 0.01% to 1% (I find it hard to discriminate in this range, and in particular to judge these three against each other). Propositions (8), (9), and (10), however, are in a completely different category of improbability: double-digit negative exponents, if you're being conservative. We could argue about (7), but it probably belongs somewhere in between (4)-(6) and (8)-(10); maybe around 10^(-10), if you account for post-QM theories somehow turning Copenhagen into something more mundane than it seems now.
So the point is, we, right here, have the tools to make estimates that are a lot more meaningful than "probably yes" or "probably no". I remember reading that we tend to be overconfident on hard things and underconfident on easy things; I think we can afford to be a little more bold on the no-brainers.
It would of course be sacrilegious to place (8) below (9) and (10). Nevertheless even in the case of apparently overwhelming evidence, if you disagree with a mainstream belief 10^(-20) times you will be wrong rather a lot more than once.
Meanwhile, quantum tunnelling is a specific phenomenon which, if possible (very likely) gives fairly clear bounds on just how ridiculously improbable it is for a marble statue to wave. Even possible improbable worlds which make quantum tunnelling more likely still leave (8) less probable than (9) (but perhaps not 10).
I personally place (10) at no less than 10^(-5) and would be comfortable accusing anyone going below 10^(-7) of being confused about probabilities (at least as related to human beliefs).
Like most majoritarian arguments, this throws away information: the relevant reference class is "mainstream beliefs you think are that improbable". (edit: no, I didn't read the whole sentence) In that class, it's not obvious to me that one would certainly be wrong more than once, if one could come up with 10^20 independent mainstream propositions that unlikely and seriously consider them all while never going completely insane. Going completely insane in the time required to consider one proposition seems far more likely than 10^-20, but also seems to cancel out of any decision, so it makes sense to implicitly condition everything on basic sanity.
(Related: Horrible LHC Inconsistency)
(9), and (10) for some definitions of "Christianity", being more likely than (8) seems conceivable due to interventionist simulators (something I really have no idea how to reason about), but not for any other object-level reason I can think of. Can you think of others?
I'd be inclined to accuse anyone going above... something below 10^-7... of being far too modest.
No, that is the reference class intended and described ("apparently overwhelming evidence").
Your prior is wrong (that is, it does not reflect the information that is freely available to you).
Considering normal levels of sanity are sufficient. Failing to account for the known weaknesses in your reasoning is a failure of rationality.
I am comfortable accusing you of being confused about probabilities as related to human beliefs.
I will take a bet on this, if you like. Also, did you perhaps mean "attempt to secede", or are you predicting actual success? I'll take the bet either way.
I don't see that happening -- which one or ones do you think are most likely to leave?
Scotland may well leave the UK (10%), or the UK leave the EU (15%).
US states aren't allowed to secede. Not even Texas. The US government would lose so much prestige from the loss of a state, that they would never allow it. So it would require some kind of armed conflict that no one state could ever win.
By 2020, an Earth-like habitable extrasolar planet is detected. I would take a wager on this one but doubt anyone would give me even odds.
Will anyone give me even odds if the bet is by 2015?
I think I'd give better-than-even odds for either date, and would be shocked if no one else would. How are you defining "Earth-like" and "habitable"?
I think he just meant with liquid water, some type of atmosphere, and approximately earth sized. Given this, my guess is that they find one within the next three years. If he meant "habitable" to human beings without protection, i.e. oxygen atmosphere etc., then this is extremely unlikely (less than 2% chance) that they will find such a thing by 2020.
Yes.
I'm not sure we have the technology to make that call even if such a planet does, in fact, lie within range of our telescopes.
We don't. My prediction then is only almost certainly true if we define habitable as a planet in a sun's habitable zone. However, I still think finding a habitable planet, per Unknowns's definition, is likely to happen by 2020.
http://www.npr.org/templates/story/story.php?storyId=101493448
If Kepler does indeed find hundreds of planets in habitable zones, that should get the popular imagination going enough for the successor to Kepler to be very well funded. Kepler Mark II in the air by 2017?
At even odds I would take a loan to make the bet.
For the next decade: Videoconferencing.
Videoconferencing what, exactly?
I've been using it for years. I'm not sure how to correctly expand your sentence, and it shouldn't be subject to interpretation.
I suspect Eliezer is making broad predictions about what is important in the next 10 years. As if someone said smartphone for the next decade in 2000. Not giving too much detail makes it more likely to be true...
coughmakingbeliefspayrentcough
Eliezer seems to be predicting that videoconferencing will become common in the next decade. Yes, some use it now, but it is still not common. I predict that it will not become common until someone uses a utility to modify your appearance so that when you look at the eyes of the person on the screen, your image on the remote end will look like it is looking at the eyes of the person on the other end. This might well be developed in much less than 10 years, however.
I predict a 10% chance that I win my bet with Eliezer in the next decade (the one about a transhuman intelligence being created not by Eliezer, not being deliberately created for Friendliness, and not destroying the world.)
I'll go ahead and claim a 98% chance that, if a transhuman, non-Friendly intelligence is created, it makes things worse. And an 80% chance that this is in a nonrecoverable way.
I kinda hope you're right, but I just don't see how.
This prediction is technically consistent with my prediction (although this doesn't mean that I don't disagree with it anyway.)
I'll put down money on the other side of this prediction provided that we can agree on an objective definition of "transhuman intelligence".
My bet with Eliezer can be found at http://lesswrong.com/lw/wm/disjunctions_antipredictions_etc/.
I said there at the time, "As for what constitutes the AI, since we don't have any measure of superhuman intelligence, it seems to me sufficient that it be clearly more intelligent than any human being." Everyone's agreement that it is clearly more intelligent would be the "objective" standard.
In any case, I am risk averse, so I don't really want to bet on the next decade, which according to my prediction would give me a 90% chance of losing the bet. The bet with Eliezer was indefinite, since I already paid; I am simply counting on it happening within our lifetimes.
I like your side of the original bet because I think the probability that the first superintelligent AI will be only slightly smarter than humans, non-goal-driven, and non-self-improving, and therefore non-Singularity-inducing, is better than 1%. The reason I'm willing to bet against you on the above version is that I think 10% is way overconfident for a 10-year timeframe.
Would a sped-up upload count as super-intelligent in your opinion?
The second estimation in each paragraph is conditional on the first.
By 2020 some kind of CO2 emissions regulation (cap and trade) will be in place in the US(.85). But total CO2 emissions in the US for 2019 will be no less than 95% of total CO2 emissions for 2008 (.9).
Obama wins reelection (.7). The result will be widely attributed to an improving economy (in the media and in polls and whether or not the economy actually improves) (.85)
By 2020 open elections are held for the Iranian presidency (no significant factions excluded from participation) (.5). The president (or some other position selected through open elections) is the highest position in the Iranian state (.5)
"The president (or some other position selected through open elections) is the highest position in the Iranian state (.5)"
Qualify this. Formally, the highest position in the British state is unelected. In terms of political power, the highest position in the British state is elected.
In terms of political power.
I'm 90% confident that the cinematic uncanny valley will be crossed in the next decade. The number applies to movies only, it doesn't apply to humanoid robots (1%) and video game characters (5%).
Edit: After posting this, I thought that my 90% estimate was underconfident, but then I remembered that we started the decade with Jar-Jar Binks and Gollum, and it took us almost ten years to reach the level of Emily and Jake Sully.
Why such a big gulf between your confidence for cinema and your confidence for video games?
Movies are 'pre-computed' so you can use a real human actor as a data source for animations, plus you have enough editing time to spot and iron out any glitches, but in a video game facial animations are generated on-the-fly, so all you can use is a model that perfectly captures human facial behavior. I don't think that it can be realistically imitated by blending between pre-recorded animations like it's done today with mo-cap animations -- e.g. you can't pre-record eye movement for a game character.
As for the robots, they are also real-time, AND they would need muscle / eye / face movement implemented physically (as a machine, not just software), hence the lower confidence level.
The obvious answer would be "offline rendering".
Even if the non-interactivity of pre-rendered video weren't an issue, games as a category can't afford to pre-render more than the occasional cutscene here or there: a typical modern game is much longer than a typical modern movie -- typically by at least one order of magnitude, i.e. 15 to 20 hours of gameplay, and the storyline often branches as well. In terms of dollars grossed per hours rendered, games simply can't afford to keep up. Thus, the rise of real-time hardware 3D rendering in both PC gaming and console gaming.
Is there a reason Avatar doesn't count as crossing the threshold already?
Because the giant blue Na'vi people are not human.
You mean you didn't notice the shots with the simulated humans in Avatar? ;-)
In a way, the uncanny valley has already been crossed-- video game characters in some games are sufficiently humanlike that I hesitate to kill them.
Within ten years either genetic manipulation or embryo selection will have been used on at least 10,000 babies in China to increase the babies’ expected intelligence- 75%.
Within ten years either genetic manipulation or embryo selection will have been used on at least 50% of Chinese babies to increase the babies’ expected intelligence- 15%.
Within ten years the SAT testing service will require students to take a blood test to prove they are not on cognitive enhancing drugs. – 40%
All of the major candidates for the 2016 presidential election will have had samples of their DNA taken and analyzed (perhaps without the candidates’ permission.) The results of the analysis for each candidate will be widely disseminated and will influence many peoples' voting decisions - 70%
While president, Obama will announce support for a VAT tax - 70%.
While president, Obama will announce support for means testing Social Security - 70%
Within ten years the U.S. repudiates its debt either officially or with an inflation rate of over 100% for one year - 20%.
Within five years the Israeli economy will have been devastated because many believe there is a high probability that an atomic bomb will someday be used against Israel – 30%
Within ten years there will be another $200 billion+ Wall Street Bailout - 80%
I'd take the other side on any of these if we can find a way to make it precise.
"While president, Obama will announce support for means testing Social Security - 70%"
I'd be wiling to take those odds, with some refinements.
China is the 2nd biggest economy in 2020 (99%). Note I'm counting the EU as lots of countries, not as one big economy. Counting the EU together, China will be the 3rd biggest.
Pirate Parties will have been in government for a time in at least one country by 2020 (90%)
Pirate Parties will win >=10 seats in the European parliament in 2014 (75%), and <=30 seats (75%).
The Conservatives will win a majority the next UK general election (60%), there will be no overal majority (37%), or any other outcome (3%).
Do you have bets on Intrade or Betfair for those guesses? It's probably better for you to bet directly than for me to do arbitrage on you :-) They have around 68% Conservative victory, 26% no overall majority, and around 6% Labour victory.
Betfair
By the end of 2013: Either the Iranian regime is overthrown by popular revolution, or there is an overt airstrike against Iran by either the US or Israel, or Israel is attacked by an Iranian nuclear weapon (70%).
Essentially seconding mattnewport: the price of gold reaches $3000USD, or inflation of the US dollar exceeds 12% in one year (65%).
The current lull in the increase of the speed at which CPUs perform sequential operations comes to an end, yielding a consumer CPU that performs sequential integer arithmetic operations 4x as quickly as a modern 3GHz Xeon (80%).
Android-descended smartphones outnumber iPhone-descended smartphones (60%).
The number of IMAX theaters in the US triples (40%).
When you say sequential integer operations, do you mean integer operations that really are sequential? In other words, the instructions can't be performed in parallel because of data dependencies? If not, then this is already possible with a sufficiently wide superscalar processor or really big SIMD units.
But let's assume you really mean sequential integer operations. The only pipeline stage in this example that can't work on several instructions at once is the execute stage, so I'm assuming that's where the bottleneck is here. This means that the speed is limited by the clock frequency. So, here are two ways to achieve your prediction:
Crank up the clock! Find a way to get it up to 12 GHz without burning up.
Make the execute stage capable of running much faster than the rest of the processor does. This is natural for asynchronous processors; in normal operation the integer functional units will be sitting idle most of the time waiting for input, and the bulk of the time and complexity will be in fetching the instructions, decoding them, scheduling them, and in memory access and I/O. But in your contrived scenario, the integer math units could just go hog wild and the rest of the processor would keep them fed. This can be done with current semiconductor technology, I'm pretty sure.
So, either way, kind of an ambitious prediction. I like it.
Next 10 years:
Nativism discredited (80%)
Traditional economics discredited (80%)
Cognitivism/computationalism discredited (70%)
Generative linguistics discredited (60%)
To elaborate somewhat: By #1 I mean that in the fields of biology, psychology and neuroscience the idea that behaviours or ideas or patterns of thought can be "innate" will be marginalised and not accepted by mainstream researchers.
By #2 I mean that, not only will behavioural economics provide accounts of deviations from traditional economic models, but mainstream economists will accept that these models need to be discarded completely and replaced from the ground-up with psychologically-plausible models.
By #3 I mean the idea that the brain can be thought of as a computer and the "mind" as its algorithms will be marginalised. I give this lower odds than nativism being discredited only because the cognitivist tradition has managed to sustain itself through belligerence rather than evidence and is therefore likely to be more persistent and pernicious. Nativism, on the other hand, has persisted because of the difficulty of experimentally demonstrating that certain behaviours are learned rather than innate (as well as belligerence).
By #4 I mean that traditional linguistics, and especially generative grammar, will be marginalised. This one has long puzzled me since the generative grammarians based their ideas on intuition and explicitly deny a role for data or experiment (or the need to reconcile their beliefs with biology). The main problem has been the absence of a viable alternative research program. This is beginning to change.
Can you unpack what you mean by innate. I think babies would have a hard time surviving if sucking things wasn't a behaviour that was with them from their genes.
And more generally, the distinction innate/learned is overly simplistic in a lot of contexts; rather, there are adaptations that determine the way organism develops depending on its environment. The standard reference I know of is
J. Tooby & L. Cosmides (1992). `The psychological foundations of culture'. In J. Barkow, L. Cosmides, & J. Tooby (eds.), The adapted mind: Evolutionary psychology and the generation of culture. Oxford University Press, New York.
If we could agree on a suitable judging mechanism, I would bet up to $10,000 against you on #1 and on #3 at those odds (or even at substantially different odds). I also disagree on the latter claim in #2, but that's not as much of a slam dunk for me as the others.
A few thoughts:
It would be valuable to do an outside view sanity check: historically, how frequently have research programs of similar prestige been discredited?
There are all the standard problems with authority---lots of folks insist that they're in the mainstream and that opposing views have been discredited. Clearly nativism &c. have been discredited in your mind; when do they get canonically discredited? Sometimes I almost think that everyone would be better off if everyone just directly talked about how the world really is rather than swiping at the integrity of each other's research programs, but I'm probably just being naive.
Re 3, my domain knowledge is somewhat weak, so everyone ignore me if my very words are confused, but I'm not sure what would count as a refutation or the mind being an algorithm. Surely (surely?) most would agree that the brain is not literally a computer as we ordinarily think of computers, but I understand algorithm in the broadest sense to refer to some systematic mechanism for accomplishing a task. Thought isn't ontologically fundamental; the brain systematically accomplishes something; why shouldn't we speak of abstracting away an algorithm from that? Maybe I've just made computationalism an empty tautology, but I don't ... think so.
I don't think the innate/learned dichotomy is fundamental; it's both, everyone knows that's it's both, everyone knows that everyone knows that it's both. Like that old analogy, a rectangle's area is a product of length and width. What specific questions of fact are people confused about?
I think these research programs represent something without a clear historical precedent. Traditional economics and generative linguistics, for example, could be compared to pre-scientific disciplines that were overthrown by scientific disciplines. But both exhibit a high degree of formal and institutional sophistication. I don't think pre-Copernican astronomy had the same level of sophistication. Economics also has data (although so did geocentric astronomy) whereas the generative tradition in linguistics considers data misleading and prefers intuitive judgement. What neither has is a systematic experimental research program or a desire to integrate with the natural sciences.
Cognitivism is essentially Cartesian philosophy with a computer analogy and experiments. In practice it just becomes experimental psychology with some extra jargon. Nativism, too, comes from Cartesian philosophy (Chomsky was quite explicit about this). While cognitivism has experiments it has an interpretation that isn't founded in experiment (the type of computer the brain is supposed to be and the algorithms it could be said to run is not addressed) and an opposition to integration with the natural sciences (the so-called "autonomy of psychology" thesis).
These research programs are similar to pre-scientific research programs but have managed to persist in a world where you have to attempt to "look scientific" in order to secure research grants and they reflect this fact.
You point to many problems and I wouldn't take any bets because of these. It would be too difficult to judge who had won. On the nature/nurture debate: Empiricism evolved into constructivism/interactionism (i.e., the developing organism interacting with the environment with genes driving development), which is the dominate view in biology, and it's not obvious what, precisely, modern Nativists believe. But it is obvious that they still exist since naive nativist talk persists almost everywhere else. It's similarly difficult to figure out what computationalists mean by their analogies and the degree to which they intend them to be analogies vs. literal propositions. This is probably why the natural sciences tend not to base research programs on analogies. What is clear is that they have a particular style of interpreting their results in terms of representations and sequential processing that is clearly at odds with biology and display no interest in addressing the issue.
First, this is the genetic fallacy. Secondly, I don't take Chomsky's authority seriously.
The experimental evidence that, say, Steven Pinker presents in How the Mind Works for innate mental traits and for the computational perspective are sound, and have nothing to do with Cartesian dualism.
The point is that the views have their origins in philosophy rather than experiment. We're not dealing with a research program developed from a set of compelling experimental results but a research program that has inherited a set of assumptions from a non-empirical source. This is more obviously the case with computationalism, where advocates have shown almost no interest in establishing the foundational assumptions of their discipline experimentally, and some claim that to do so would be irrelevant. But it's also true for nativism where almost no thought is given to how nativist mechanisms would be realised biologically.
Regarding 3, there's no way to find evidence against it (or for it, for that matter). You can't look at a given system and measure its sentience. The closest to that anyone's ever attempted is to try and test intelligence, but that assumes cognitivism/computationalism, among other things.
I agree with orthonormal, except that I don't have $10,000 to bet.
One word: subcultures.
I think we'll see an expansion to most of the First World of the trend we see in cities like San Francisco, where the Internet has allowed people to organize niche cultures (steampunk, furries, pyromaniacs, etc.) like never before. I think that, by and large, people would prefer to seek out a smaller culture based on a common idiosyncratic interest if it were an option, not least because rising in status there is often easier than getting noticed in the local mainstream culture. I think that the main reason the mainstream culture is presently so large, therefore, is because it's hard for a juggling enthusiast in Des Moines to find like-minded people.
I expect that over the next 10 years, more and more niche cultures will arise and begin to sprout their own characteristics, with the measurable effect that cultural products will have to be targeted more narrowly. I expect that the most popular books, music, etc. of the late 2010s will sell fewer copies in the US than the most popular books, music, etc. of the Aughts, but that total consumption of media will go up substantially as a thousand niche bands, niche fiction markets, etc. become the norm. I expect that high schoolers in 2020 will spend less social time with their classmates and more time with the groups they met through the Internet.
And I expect that the next generation of hipsters will find a way to be irritatingly disdainful of a thousand cultures at once.
You forgot us!
So it's possible that, if we had a really huge, dense, wired city with excellent transportation, we would find a significant subculture of steampunk furries, or vampire gothic lolita hip-hop dance squads? Actually, this sounds like a lot like Tokyo.
It's easy, really. Practice this phrase: "Man, what weirdos." You just have to selectively overlook the weirdness of your own subculture while recognizing and stigmatizing it in others. It's an elegant approach.