Futurist prediction methods and accuracy
danluu.com
danluu.com
- influencing investor or consumer confidence. Even if you don't have high confidence in what you're saying in the long term, it may be necessary in the near term to make statements that project confidence
- building a brand as a thought-leader, futurist, whatever. In part this depends on saying something interesting that people will want to read or hear about. People can enjoy this kind of prediction without believing it. A book review for Kurzweil's The Singularity is Near said "exhilarating speculation is great fun to read, but needs to be taken with a huge dose of salt." Buckminster Fuller was wrong about many things but people still enjoy his work because it's interesting.
https://en.wikipedia.org/wiki/The_Singularity_Is_Near#Review...
- influencing policy in your preferred direction. No one can predict the number of severe zoonotic diseases that will have epidemics in the next 30 years, but some experts think they'll happen more frequently as the climate crisis progresses. We might all be better off if experts understate the uncertainty in their projections when they make their case to policy-makers.
Frustrating, but true. I was recently asked at work to stop adding statements on certainty, to be more concise. I suppose they were in essence delegating the decision of whether something matters to me. That's in some ways a reasonable request, because if I make a nuanced presentation, the audience doesn't have the education/knowledge to evaluate the nuances.
To your point about people enjoying the work, they're still allowed to. Anyone can read what Dan wrote, and say "I enjoy this work despite the fact that (almost) none of it is true." I do this with Marvel movies, for instance. I do suspect that many readers did not realize how phenomenally bad Kurzweil's projections are, and would be disappointed by realizing that they are.
Nobody needs to be reminded that people who make public predictions might want to deceive investors or customers, to be famous and make money, or to deceive voters or legislators into pursuing their favored policies. People need to be reminded that their current psychics (and pundits for that matter) have a long history of only getting the things right that everybody got right, and even getting a few of those wrong for a bit of contrarian attention.
No, my assumption is that the author is interested in understanding what contributes to good vs bad predictions, and how credible long range predictions are.
The author points out factors such as whether the predictor appears to learn from past prediction errors, whether they have a deep expertise in the area they're making a prediction about, whether they depend on extrapolating exponential trends etc. I am guessing that "Do they sell a lot of popular press books?" and "Do they receive performance based compensation from an organization invested in the domain under discussion?" are at least as important explanatory factors as the ones the author explicitly discusses.
I can score a 100% accuracy rate by predicting things that everyone knows are going to happen (humans will still exist in 2025, the sun will rise in 2050, etc.). A futurist is not useful when they are merely accurate, but when they are accurate about events that no one else expected.
It is inappropriate to compare someone like Kurzweil to Caplan. Caplan is just trying to be correct. Kurzweil is trying to make predictions about black-swan events -- the things people don't see coming. These are different ballgames, and the ballgame of predicting unexpected events is inherently more difficult.
https://web.archive.org/web/20170225013846/https://en.wikipe...
Interestingly enough, in the latest version of this wikipedia page, all these "predictions" are removed.
LOL, Zuck's metaverse would like a word...
To be indistinguishable, though, not quite there yet. Then again, the valve index is still visually nearly, if not the, top of the line, and 3 years old, and it sounds like valve, meta, apple, and some more smaller players are still working on a true next gen, where "gen" is something that feels wholly, indisputably superior to previous offerings. Vive is also still trying but also not doing anything revolutionary. We are however simultaneously on the brink of consumer-level extremely high fidelity, high speed, cable-matching wireless video, and being able to strap sufficient computer power into the headset itself, and no-extra-equipment, compromise-free tracking. VR is absolutely a guaranteed long term growth market
At the time of writing (the quietest period in the week for online gaming worldwide), it is currently 54th place in terms of people playing on steam right now.
None of the 53 games above it require/recommend a dedicated peripheral to play.
VR won't hit 1 million simultaneous users in anything until VR displays can render high quality text on a quad (think a screen or piece of paper) and/or provide haptic feedback.
VR is still held back by today's technology for the most interesting uses, but is still acceptable enough for a social space to thrive.
It’s all early days but I think OP is very correct. 20k is comparatively impressive but absolutely shows how limited in scope a “runaway” success looks like. Think of it this way. Traditional console/PC games every year have several AAA titles that sell millions of copies each and have millions of people playing online. Compared with VR where Beat Saber is basically the #1 game for several years. That’s a massive marketplace to sell into with a long track record of winning formula (that’s why gameplay and business modes aren’t innovated on too much). That kind of stuff hasn’t happened for VR yet. Note that I’m not saying it won’t.
Also I wouldn’t look at Steam numbers here - Quest, which is the number 1 VR headset in the world, doesn’t make its analytics visible there [1]. VRChat is top 10 but Beat Saber is more popular I think. So the conclusion is that VRChat is the #1 VR game for PCVR which is a comparatively stagnant market and doesn’t speak to the broader ecosystem.
[1] https://www.oculus.com/experiences/quest/section/31585546237...
I cannot tell which one is real, can you?
TBH, I never tried VR and when I intended to try VR as I hear good things about Half Life: Alyx I was not able to find a VR centre close to me that has not failed yet. The few I found, all gone.
I also have been reading from VR advocates about how it can actually be nice for working because you can have very large view field, as if using multiple large monitors in a distraction-free environment.
However for me VR has become this picture of Zuk and I bet I'm not alone because it is so iconic and icons assume the functions of the actual thing in the absence of it. Maybe they shouldn't have been promoting Metaverse to people who don't have the equipment and can't give it a fair go.
Think of VR as a really good arcade on-rails shooter game and that's the gist of Half Life Alyx. It's a good game. But it also left me feeling very disappointed.
* You won't move much. You'll peek in and out of cover, which is fun, but you won't actually "move" around all that much for most encounters. People really like ducking in and out of cover to shoot at things, and again, that is genuinely fun.
* Some enemies break this mold and require you to move; doing extremely telegraphed attacks that pretty much just require you to move to anywhere that isn't exactly where you're standing at that moment. Specifically just enough that you need to use the controller inputs to move rather than stepping physically. If you stop to think about why the game is designed this way, it's because people are bad at moving irl, and this is about all they can handle.
* In general all enemies have this feel of an on rails shooter though, and you have no way to interact with them besides throwing grenades and shooting. You can't melee and a lot of enemies stand just a little too far to seem realistic so that they don't do weird overlaps with your camera.
* There's some physical interaction stuff that is good. Stuff that in other games would have been "Press X to do Y" is now a physical machine you need to move around with your hands, especially reloading. But it's also very limited because most of the game is a static mesh that you can't really do anything with.
So... yeah, I enjoyed it but it made me pessimistic for the future. I think they worked hard to deal with VR's limitations but they are same damn high walls.
The downside of allowing these interactions is that they feel very unreal and “physics sim”-ish, almost dreamlike due to weightlessness. For a game with a serious story, it might be immersion-breaking. But I had an amazing time just experiencing it as its own thing.
You can physically hit enemies but it's a very limited system. If you play with it for a bit you'll probably come to the conclusion that "fighting" is impractical and the best option is to hold forward movement and stab people with your arm outstretched. Because your input movement is wayyyy more maneuverable than your physical movement. Trying to two hand a large weapon is difficult because your hands don't lock to the object irl but they try to in game and you rarely get satisfying bonks with the sweet spot of a weapon. You can grapple enemies, but it's a stretch to say you're grappling. They simply become rag dolls that don't do anything once you grab them. You can also grab them with one hand and punch them with the other but it takes FOREVER.
And lastly, while I grinded through the nausea and didn't feel like puking by the end of it, it still was vaguely unpleasant to climb and jump. It's a good example of physical presence in video games. it'll be hard to do better than that but it still kind of sucks
They all forget to mention that you need to put a big goofy computer on your head
I wonder what was up with this one.
Computers are now capable of learning and creating new knowledge entirely on their own and with no human help. By scanning the enormous content of the Internet, some computers "know" literally every single piece of public information (every scientific discovery, every book and movie, every public statement, etc.) generated by human beings.
I don't know what the deletion of predictions means, but Wikipedia editors in general seem to dislike checklists.
But I was surprised to find that the article did not say much about the bets themselves, nor analyze in any great detail what exactly Caplan is doing. Indeed, it's not even possible based on the article AFAICT to check the 100% figure (23/23) independently.
Hopping through some links, I found this list:
https://docs.google.com/document/d/1qShKedFJptpxfTHl9MBtHARA...
What's interesting is how many of Caplan's bets favor a continuation of the status quo against an opponent believing in a more extreme outcome.
The open bets are also interesting. Consider this one:
> You pay me $x today. If any European country that was not Communist in 1988 has a civil war leading to 10,000 or more fatalities between today and December 31, 2045, I will immediately pay you $11x.” (initial terms, final terms here) 09/24/2015. Status: open.
That said, the language prevents me from figuring out which side Caplan bet on. This is a problem with many of the terms, but fortunately the winner is given as Caplan. Not so in this case.
Given Caplan's apparently conservative bent, I suspect he's "me".
Whenever someone mentions the singularity, this is what I imagine.
Edit: I’m reading some of the criticisms of Kurzweil’s claims in the article. It is clear that he was wrong about growth rates in many areas and about life expectancy growing unbounded. Perhaps my skepticism kept me from taking these claims at face value at any point. Perhaps it is dangerous for someone to be making these claims to the public. I am not sure.
Kurzweil is to futurism what Carrot Top is to comedy.
>In 2001, he uses this method to plot a graph and then predicts unbounded life expectancy by 2011
and then Dan cites this sentence:
> In the eighteenth century, we added a few days every year to human longevity; during the nineteenth century we added a couple of weeks each year; and now we’re adding almost a half a year every year.
Like, let's lay aside the predictions. Between 1980 and 2000 US life expectancy increased by 3 years. So let's just shelve the question of the predictions, this person is clearly making shit up. Whoever is making this prediction is simply not getting even the most basic facts correct.
In some ways I think the approach of this article is backwards. You don't apply retrospective rigor to old predictions, you define a rigorous process for future predictions, because there's no way someone is going to luck into being rigorous, and I think most of the futurists are even aiming to be rigorous. A lot of futurists are essentially sci-fi writers. What they're doing is fundamentally entertainment rather than rigorous scientific method. So I don't see why you would set out to judge them by a rigorous scientific method.
In some ways I see a parallel between this and effective alturism - EAs often think other forms of charity are misguided, but there's no point judging non-EA giving by EA standards because people just fundamentally don't agree with the core ideas of EA.
The best futurism was probably at the 1939 World's Fair.[1] The General Motors vision for 1960 came to pass. Wide freeways! Endless suburbia! Big airports! Mission accomplished. By the 1964 World's Fair, futurism was kind of stuck. The GM exhibit had moon bases and undersea cities. Nope. The 1964 World's Fair mostly ended the era of futuristic world's fairs, although there were a few later attempts. Usually involving monorails.
The huge misses of futurism:
* Robots. Manipulation in unstructured environments still sucks.
* Atomic power. Works, but every two decades or so you lose a city-size area. Costs too much.
* Space. Costs too much to get there, not that useful to be there.
* Immortality. Not even close.
What we have now is dystopian futurism, about which too much has been said already.
Unless you live in that large part of the world where living standards having been rising quickly.
I will predict we'll have more compute throughput and endless social media in 20 years. Not sure on what the next big public transit innovation will be.
It's worth noting that Dixon's prediction was at the very tail end of the time when CPU portable was a major goal. In the late 90s, anyone compiling for Unixes might reasonably have to support a half-dozen major architectures. Nowadays, even including 32- and 64-bit as separate architectures, there's basically 4 architectures that matter (i386, x86-64, arm, aarch64) for most software. And since 1998, RISC-V and Itanium would have been the only notable new architectures.
This doesn't contradict what Dan says about Dixon being wrong, though.
Dan Luu: "Kurzweil pushes the date this is expected to happen back by more than one year per year (the last citation I saw on this was a 2016 prediction that we would have unbounded life expectancy by 2029), which is characteristic of many of Kurzweil's predictions."
(2) Prediction on long timelines is too hard. It's more useful to generate ideas about possibilities. With the small samples here, people are almost just as likely to be lucky as right.
We saw a microcosm of this early in the pandemic, where some groups got really excited about forecasting the number of cases. (1) I didn't see anybody quantitatively evaluate their own predictions and own up to how good or bad they were. (2) Everyone pretty quickly gave up even trying to make predictions over a month away. It was too hard. So it devolved into fitting short-term curves and then I think mostly they gave up forecasting altogether.
If, say, I were to predict that 40% of intercity travel would be via high speed rail in 2100 in California, I should be able to either argue why high speed rail would spring out of nowhere in 2099 or have a plausible story as to how it might grow.
Kurzweil just says 2045, because that's where his chart or accelerating returns tells him it will happen.
> Such trend-curves... extrapolated 30 or 40 years they are so high and steep that even the most naive futurist discounts their accuracy.
(That's the closest therein to a specific prediction date, from 1983: https://imgur.com/a/HrUxGts)
> let me more specific: I'll be surprised if this event occurs before 2005 or after 2030.
From 1993: https://edoras.sdsu.edu/~vinge/misc/singularity.html
It's interesting that 29 years after he said that, and 8 years remaining on the timeline, it's not completely out of the realm of possibility.
For example: all people will eventually die, both today and 50 years from; Earth’s climate will increasingly get warmer; etc.
By collecting enough of these types of predictions, future becomes much more clear than attempting to predict what will change.
I wouldn't be so sure--would you also be confident about 100 years from now? 500?
https://en.wikipedia.org/wiki/Superforecasting%3A_The_Art_an...
So much of it is "local exponential increase will continue forever" despite all the laws of nature screaming it won't.
What if I told you that's not a coincidence?
https://www.christianevidence.net/2019/01/gods-7000-year-pla...
The only things these "science communicators" have done is parade their buffoonery around and turn science into a daytime TV marathon. They've created a generation or two of people who think they're informed because they passively watch TV shows with pretty renders in them, "nerds" who don't design or build anything, and an idea that you can be right all the time if you venerate the right people. They're grifters, they're novelty guests at parties, they don't actually understand anything or teach anyone anything.
Actual smart people who know what they're talking about don't take these people seriously. The day I listen to these guys is the day Neil DeGrasse Tyson draws me a picture depicting exactly how you write bug free code.
Also any recommendations for material to read about future trends in technology outside of people mentioned in the article?
Ouch!