Right, there are things I can do in a second with a computer that would have took lifetimes in the 1800s. There are also things I still cannot do.

Having enough energy to power these computers (or whatever problems people in the future are trying to solve) could take 10^100 people 10^100 years

Survivorship bias is a terrible basis for an argument here.

This criticism of Kurzweil is a very poor and unusual usage of the term.

Survivorship bias normally matters in one of three cases:

1) You have a series of binomial trials and there are a sufficient number of predictions that there is likely a match. (ex: finance scams)

2) You require one success to accept a claim, but the failures are hidden. (medicine)

3) You measure performance for a group based on how the surviving members of the group performs. (both finance and medicine)

Kurzweil's predictions are distinctly different from these cases.

He didn't choose a characterized random process and guess on the outcome.
In short, the sample space of possible outcomes is so large than even a staggering pool of futurist-predictors will fail to guess correctly.

Good Survivorship bias example:
2^n financial analysts making unique sets of predictions regarding whether a stock will either go up or down by the end of a day, for n days, is guaranteed to have 1-super prophet.

Bad Survivorship Bias:
Many political pundits predict what President Obama will say.
One predicts President Obama will use the phrase "eliminate all capital gains taxes on small businesses" in the first 1/4th of his state of the union.

No matter how many pundits you have, the odds of a pundit guessing a sequence of words by pure chance is exceptionally low (1/250,000)^9. There must have been a skill element to the capital-gains claim.

Here's an example Kurzweil prediction:

(made in 1990, predicting 2000s): Telephone calls are routinely screened by intelligent answering machines that ask questions to determine the call's nature and priority.

This unfortunately happened on schedule. Do you really expect us to think we should weight the significance of this prediction with the significance of "Microsoft will go up tomorrow?" Of course not! Kurzweil didn't have a 5-50% of generating a correct prediction when he sat down and started writing.

Do we need a minimum of two futurists to have had one futurist get it correct? This is what your usage of the term survivorship bias implies.

Comparing Kurzweil to a donkament winner / Peter Schiff is a similar to the frequent probabilistic error of thinking that any two outcomes are equally likely.

Kurzweil didn't have a (50%)^n shot of generating his predictions. There is clearly a massive skill element in his forecasting.

Wrong. It's exactly what we'd expect from someone who is good at predicting the future. The more you publish the harder it is to be lucky (it would be harder to maintain a high percent accuracy through luck) - the fact that he's published a lot works against the survivorship bias.

We know that Kurzweil is good at predicting because he's shown us his reasoning behind his predictions and it makes sense. If you looked at a player play poker, it would be very hard to tell if he was actually good unless you looked at a large number of hands. You would also be able to tell if he was good at poker by asking about why he played the hand in a specific way. If his reasoning was weak and inconsistent you'd know he was a bad player. Kurzweil has made a lot of correct predictions AND he's shown us the strong logical thinking behind them.

Perhaps you should dig deeper into the sorts of decisions in financial advisors' success rates. They have to make hundreds of decisions a year about buying/selling stocks that constitute mutual funds. Say a fund supervisor has a run of 10 20% profit years. You'd immediately think that he's an all star and extremely talented. He's made thousands of great decisions and produced results. But, alas, it's survivorship bias.

Same thing is likely funcitoning here. Your analysis of the source of survivorship bias is too simplistic and doesn't capture how it can happen.

Not "wrong": the outcome is what we'd expect given both processes. So, the point is that it's prima facie underdetermined whether the "evidence" (track record) indicates skill or survivorship bias. The fact that you don't even consider the latter is an error.

This is a HALLMARK of survivorship bias. Things "look" like it's skill when it isn't. That's why the bias is so difficult to detect.


Bad example. People can "rationalize" sub-optimal decisions and make them appear optimal quite easily. Just look at top pros disagreeing over reasoning processes. This indicates an ignorance of disconfirming cases (confirmation bias now!). Look at the reasoning behind his bad predictions, not just his successful ones.

Then, look at the reasoning behind less successful futurists: do they look equally impeccable? What explains Kurzweil's success but their failure if their reasoning looks equally good?

Gee, maybe it's luck viz. survivorship bias?

future predicitions aren't coinflips.

Don't need to be. Understand my analysis better please.

Kurzweil may have some skill in his predictions but I'm conjecturing that a large source of his success is the result of survivorship bias. It has all the hallmarks.

Asymptote? Not really. He simply thinks that the exponential trend will contine.

E=mc^2 is pretty simple (but the proof for it isn't): I can write it in 5 characters. The fact that it's simple is not evidence against it, in fact I would say the opposite. Simplicity is better. If you keep having to patch up a hypothesis by adding bits to make it correct (and more complicated) then it's more likely to be wrong.

Actually, no he doesn't. He addresses this issue.

There exist fund managers (Jim Simons) that are pretty much guaranteed to be successful for reasons other than surviorship bias and this can pretty much be proven mathematically.

Did you read my post?
The odds of someone running over expectation in a game of coinflips is completely different than the proverbial monkey's chances of typing Shakespeare's works.
There is a fundamental difference between "run good in a game with chance" and "generate a proposition."

A fund supervisor running above expectation for 20 years is a completely different calculation than someone generating a hypothesis like "gee whiz, I think blind people will be able 'read' books with a machine in the 2000s." How do you luckbox your way into a profound statement?

Please show us a way of randomly generating these predictions, and give us an estimate on how many futurists randomly generating predictions it'll take to get Kurzweil's record.

OK, looked up his wikipedia page out of curiosity. Definitely does not improve my opinion of his opinions.

http://en.wikipedia.org/wiki/Raymond_Kurzweil

You don't understand that the decisions are effectively the same. The fund managaer is making hundreds of predictions a year; taking in information and generating hypotheses regarding geopolitical futures, where various sectors are heading, and so on.

Don't let your ignorance lead you astray.

Also, you didn't read my post carefully enough. I didn't say that Kurzweil is totally random. You get survivorship bias even where there's some skill. I said that he clearly isn't just randomly guessing, but that a significant portion of his success is likely explained by survivorship bias.

That is completely absurd. You cannot argue (without losing all credibility) that because something looks like skill it is more likely to be the result of luck. And the survivorship bias is concerned about survival - it is not concerned about how the decisions that ensured survival took place so it isn't a hallmark.

Show me just one futurist who has reasoning that is anywhere near as logical, smart and clear as Kurzweil's.

Good thing that I'm not arguing anything remotely close to this. ducy?

I don't think that you know that to which survivorship bias refers.


There are all sorts...they're a dime a dozen. The thing is that you won't have heard of many of them since they don't have the track record or noteriety of Kurzweil. But, their reasoning is no less "logical, smart, or clear"...they just didn't flourish by being so fortunate as Kurzweil.

I don't think you understand what survivorship bias is at all... and many of us have tried to show this to you but you refuse to admit that you don't really know what it's about.

Funny how there are so many and yet you are still unable to name just one.

The decisions are not effectively the same.

De novo generation of propositions is completely different than investment decisions.


You are accusing several poker-playing electrical engineers / computer scientists / etc of not understanding elementary probability - you seem to think that it somehow matters whether we are examining the trial of one random variable, or a trial consisting of a summation of random variables.
Not only does our analysis not change, but the very accusation is incorrect.

Your argument doesn't even knock over a strawman. To quote Karganeth, your argument is absurd.

There's nothing 'de novo' about his predictions in the sense that financial predictions of the sort that I'm discussing aren't.

1) Kurzweil prediction - (made in 1990, predicting 2000s): Telephone calls are routinely screened by intelligent answering machines that ask questions to determine the call's nature and priority.

2) Financial Analyst who runs over EV for several years.

Are you suggesting that when evaluating the above examples, the impact of survivorship bias is of a similar order of magnitude?

That's some bold **** there.

Hannah-Barbera predicted in 1962 the emergence of Skype-like services.

Plancer is obviously right, so I'm not going to reiterate what he's said.

I just want to add, that Kurzweil's success / resume, preceded his predictions, so you can't say that's the output of survivorship bias.

1965, age 17 - he appears on I've Got a Secret where he plays music that a computer he built composed.

Age 20 - sells company for 500k (inflation adjusted). Company used a computer program to match high school students with colleges.

1976 - His company successfully produces the first computer program capable of recognizing text written in any normal font aka first print-to-speech reading machine.

17 Honorary degrees.

Has successfully cured himself of Diabetes (to the extent where he has absolutely no symptoms, and no doctor could "spot" it).

He's currently 62 with the mind/body of a 45-50 year old.

The man is clearly a genius even if you think he is crazy.

I didn't say my intuition should have merit. I was explaining to you how easy it would be to prove me wrong. Set you up for a layup (assuming you are right) and you didn't take it.

This is complete bull****.

The vast majority of his arguments are discussing technologies that are in production/testing, or are already demonstrable today.

For every prediction he makes (other than the ones that need no explanation), he goes into great detail to prove the basis for the technology exists. He discusses companies current manpower/budgets. Refers to who some of the lead designers / developers / funders are. He discusses what they are capable of today, why the fall short of currently being on the market, and what is needed to overcome that.

If you had even skimmed the book, you'd know the ridiculous depth he goes into. It's obvious that you're just flat out lying and haven't read TSIN.

Can you please get off his nuts long enough to post or link to his 109 predictions for 2009 that I asked for like 100 posts ago? That's slightly more relevant evidence for his ability than the above.

Maybe he never ****ing had diabetes to begin with? Oh, wait, he didn't. What a shock. Another "misgraded" assertion by a delusional liar (assuming he actually claimed he'd done this).

Here you go.


I was responding to the argument of survivorship bias. Posting his predictions that have proven correct wouldn't exactly argue against it.

Kurzweil is irrelevent. The idea is good or it is not idependently of him. Makes no sense to believe it or not based on Kurzweil's track record. the only relevence of his track record is whether we bother to read about his ideas in the first place. Clearly he has passed that hurdle.