How prescient is Ray Kurzweil and how useful are his predictions?

Even a blind squirrel finds a few nuts. But Ray Kurzweil missed the Cloud, Facebook, Twitter, SaaS, phone apps, and Angry Birds.

In case you have been in a cave for a while, Ray Kurzweil has joined Google. According to Ray “In 1999, I said that in about a decade we would see technologies such as self-driving cars and mobile phones that could answer your questions, and people criticized these predictions as unrealistic. "

There already were self-driving cars in 1999 i.e, Kurzweil can't have been predicting the concept. He must have been predicting they would be mainstream. Which they aren't. Self-driving cars are still research projects, decades after their invention. As for the phones, correct that they exist. Ray is certainly towing the line of his employer now: "Fast forward a decade — Google has demonstrated self-driving cars, and people are indeed asking questions of their Android phones." Android phones???? I'm no Apple fanboy but I think we need to acknowledge that Siri exists and came first. But have they progressed past a novelty? I have never ever once seen anyone ask their phone a question in my life. If those are Ray's best two predictions, they're lame ones.

I thought I'd look at a few more of Ray's predictions, but as is so often the case in our rapidly accelerating world, it has already been done for me. From an excellent Wikipedia section examining his predictions, these are predictions from The Age of Intelligent Machines (written in 1990) predicted for the early 2000s (i.e. before 2010):

Translating telephones allow people to speak to each other in different languages.
Machines designed to transcribe speech into computer text allow deaf people to understand spoken words.
Exoskeletal, robotic leg prostheses allow the paraplegic to walk.
Telephone calls are routinely screened by intelligent answering machines that ask questions to determine the call's nature and priority.
"Cybernetic chauffeurs" can drive cars for humans and can be retrofitted into existing cars. They work by communicating with other vehicles and with sensors embedded along the roads.
The classroom is dominated by computers. Intelligent courseware that can tailor itself to each student by recognizing their strengths and weaknesses. Media technology allows students to manipulate and interact with virtual depictions of the systems and personalities they are studying.
A small number of highly skilled people dominates the entire production sector. Tailoring of products for individuals is common.
Drugs are designed and tested in simulations that mimic the human body.
Blind people navigate and read text using machines that can visually recognize features of their environment.

So shooting 20 years out i think Ray's aim is pretty wide.

How about 10 years out? From The Age of Spiritual Machines (1999) predicting for 2009

Most books will be read on screens rather than paper.
Most text will be created using speech recognition technology.
Intelligent roads and driverless cars will be in use, mostly on highways.
People use personal computers the size of rings, pins, credit cards and books.
Personal worn computers provide monitoring of body functions, automated identity and directions for navigation.
Cables are disappearing. Computer peripheries use wireless communication.
People can talk to their computer to give commands.
Computer displays built into eyeglasses for augmented reality are used.
Computers can recognize their owner's face from a picture or video.
Three-dimensional chips are commonly used.
Sound producing speakers are being replaced with very small chip-based devices that can place high resolution sound anywhere in three-dimensional space.
A 1000 dollar pc can perform about a trillion calculations per second.
There is increasing interest in massively parallel neural nets, genetic algorithms and other forms of "chaotic" or complexity theory computing.
Research has been initiated on reverse engineering the brain through both destructive and non-invasive scans.
Autonomous nanoengineered machines have been demonstrated and include their own computational controls.

I'd give him about 50/50 on those. This more rigorous analysis of 10 randomly selected predictions rated Ray at 54% accuracy, which the analysis says is "excellent". Perhaps. I think that is over-kind, especially when that analysis rated the prediction "while schools are still not on the cutting edge, the profound importance of the computer as a knowledge tool is widely recognised. Computers play a central role in all facets of education, as they do in other spheres of life." as strongly true. I suggest if you took every computer out of every primary school and most secondary schools, teaching would continue with little disruption. Administrative paperwork would suffer but not education.

More to the point, where's the Cloud? Facebook? Twitter? SaaS? Phone apps? Angry Birds?

Ray Kurzweil rates himself 95% accurate, which is simply preposterous. I think Ray's powers of prediction are at least matched by his powers of self-promotion. Tweet this. This industry loves to hype ideas and people - Saint Steve Jobs, for example, or Kurzweil for another.

And that's for 10 years out. Looking further into the future, many of Ray's predictions are based on a fallacy: that any mathematical trend will continue into the future. That fallacy is dumb enough when one uses a linear trend: for example, in 1894 The Times estimated that by 1950 London would be nine feet deep in horse manure. Most of Ray's extrapolations are not linear, they are exponential, which is a stupid basis for extrapolation (more horse manure, as it were). Models are just that: models. They only reflect reality within the bounds in which the model was constructed. Beyond those bounds, all physical systems have counter-acting factors which limit a curve. Even if climate prediction models of rising sea-levels are true that doesn't mean that the sea-levels in a billion years will reach to the moon. Some limiting factors that predictions have come up against or will come up against: cultural resistance (e.g. self-driving cars), cost of implementation (e.g hydrogen fueled cars).
[update 2013-04-09 See more on extrapolation and Kurzweil's "second half of the chessboard"]

Other Kurzweil predictions are based on simplistic descriptions of reality which aren't true, for example comparing the human brain to a digital processor. It isn't: it is a custom-made analogue processor with multiple parallel processing mechanisms, both electrical and chemical, specialised "hardware" areas, and active biological self-modification.

[Update 30/12/12: The curve of increasing digital processing power (Moore's Law) is waved around as somehow portentious because in 12 years or so we might hypothetically have enough processing power - if the extraordinary and historically unprecedented curve of this graph is sustainable into new orders of magnitude for another decade and a half and if no other factor intervenes such as economic collapse - to run a simulation with the same CPU instruction rate as the overly simplistic model of the human brain.

This does not say we will know how to model the human brain. Only that we can run a model with the same number of operations per second as we think might maybe be happening in the human brain. What that model will actually DO is completely unknown and this curve tells us nothing about that.

Even if we come up with some sort of brain model to use up all these FLOPS, there is zero evidence that it will behave as an intelligence of any sort, only that it will have the same horsepower as our simplistic neuron models of the brain suggest we might have. I can build a fire that releases the same energy as a car, but it won't tell me much about how cars work.

So please don't derive any mystic inference from this curve that computers are gonna go sentient in 2025.
]

I love science fiction such as Kurzweil's: it is a fascinating mental exercise. To treat it as science or use it as a basis for business modelling is silly. Tweet this


See also exponential systems.

Syndicate content