Evolutionary Marker? --Robot to Take the University of Tokyo Math Entrance Exam
Follow the Daily Galaxy
Add Daily Galaxy to igoogle page AddThis Feed Button Join The Daily Galaxy Group on Facebook Follow The Daily Galaxy Group on twitter
 

« The Rings of Jupiter -- Were They Disrupted by Comet Shoemaker-Levy in 1994? | Main | NASA's Hunt for a Twin 'Blue Dot' "May Yield Some Big Surprises" »

September 13, 2012

Evolutionary Marker? --Robot to Take the University of Tokyo Math Entrance Exam

 

                            Shutterstock_1184448_3

 

Artifical intelliegence will surpass human intelligence after 2020, predicts Vernor Vinge, 62, a pioneer in AI, who in a recent interview warned about the risks and opportunities that an electronic super-intelligence would offer to mankind.

Now, Fujitsu Laboratories has announced it will participate in Japan’s National Institute of Informatics (NII) AI project, ”Can a Robot Pass the University of Tokyo (Todai) Entrance Exam?” (“Todai Robot”), led by NII professor Noriko Arai. The goal of the project: enable an AI program to score high marks on Todai’s math entrance exam for admission by 2016, and meet all admission requirements for Todai by 2021.

The test uses high-school math problems. For a computer to solve a math problem, it needs three things, according to NII:

Semantic analysis: Understand the problem text, which is expressed as natural language and formulas easily understood by humans.

Formulation: Convert to a form that can be processed by a computer.* Calculation: Find the answer using the mathematical solver.

Procedure for solving math problems:

 

            Fujitsu_math_problem

So far, Todai Robot can solve about 50–60% of Todai’s Level 2 entrance-exam problems, Fujitsu says.*Fujitsu Laboratories has been researching formula manipulation and computer algebra methods for exactly solving problems related to mathematical analysis and optimization technologies.*

Vernor Vinge is a retired professor of mathematics, computer scientist, and science fiction author who is well-known for his 1993 manifesto, "The Coming Technological Singularity, in which he argues that exponential growth in technology means a point will be reached where the consequences are unknown. He is best known for his Hugo Award-winning novels A Fire Upon the Deep, in which he envisions a galaxy that is divided up into 'zones of thought', in which the further one moves from the center of the galaxy, the higher the level of technology one can achieve.

In May 1997, IBM's Deep Blue won the chess tournament against Gary Kasparov. Was that the first glimpse of a new kind of intelligence? " It seems plausible that with technology we can, in the fairly near future," Vinge continued, create (or become) creatures who surpass humans in every intellectual and creative dimension. Events beyond such an event -- such a singularity -- are as unimaginable to us as opera is to a flatworm."

Source: Akiko Aizawa, Takuya Matsuzaki, Hirokazu Anai, Uniting Natural Language Processing and Computer Algebra to Solve Mathematics Problems, Journal of the Japanese Society for Artificial Intelligence, 27(5), 2012, in press.

The Daily Galaxy via http://www.kurzweilai.net and http://www.computerworld.com.au/article/184070/ai_will_surpass_human_intelligence_after_2020/

Image credit top of page: Shutterstock.com

Comments

I'm pretty sure this is how SkyNet started? Better find john Connor before this gets any further lol

Person: hey computer. given the population and available planetary resources can you solve world hunger?

Computer: processing........ Destroy all Humans!

Maybe get it to sing songs about the benefits of radiation, and they'll let it in with a pass.

AI is already quite impressive in its current forms as just a field of research but with the leaps and advances being made in quantum computing. These 2 fields of research paired together could create an intelligence like nothing we can currently imagine. Something like (Einstein*1000000000000000000000000000000)

Here is something I don't understand. Humans seem to be able to solve logic puzzles better at this point but higher up math problems we always seem to be using graphing calculators for. so why would a computer have issues with math problems?

Intelligence is more than knowledge and the ability to do calculations really fast. Human intelligence includes the ability to see beyond the data, to discern what is not apparent, to imagine what has not yet been imagined and make something entirely new out of existing data or things. It is the ability to recognise the paradoxes of reality, to think symbolically and understand the deeper meanings of symbolic and mythopoeic language and images. Nah! I'm not feeling outdone by computers because while they are really good at some things, they are MUCH slower than bees at figuring out the shortest route between various locations, and they don't have epiphanies while reading Homer, which they wouldn't even understand anyway. The problem is, though, that we humans under-utilize our capacity for intelligence, and so some of us hope that we can create machines to be really intelligent for us.

Computers don't have a problem with the formulas. That what we use them to solve. They have the problem figuring what the actual problem is and developing a formulas to solve that problem.
If you asked it whats 4x4 it will answer. But start asking it... "Billy and Bobby-Joe both leave thier homes at the same time. One is walking 1.2m/s the other 1.65m/s... Bobby stops for 45 secs to let his dog poop....Bobby-joe increases her speed after 10min because she has to poop...." blah blah blah. That is the part computers do not compute.

^ Tosca and Greg. But... Just wait until learning algorithms evolve enough to be applied to linguistic processing. Then things will start to get interesting.

nothing new here, computers are stealing men`s jobs since like the 80`...why not the high school places now ? For example in Japan, meanwhile a few guys earn billions using computer machines, millions of poor people fight for the daily food ...

The potential for packing neuron simulations in denser than neurons exists. The capacity to connect the grid of neuron simulations is where the AI problem lies. Assuming Time Division Multiplexing of the connections high in the net hierarchy, what mux frequency and channel density is required to bring silicon neuron performance?

Oh look. That is why no one is listening, Todai Robot. Nor should they be.... What a stupid hypothesis, trust University of Tokyo to come up with this ill-informed inkling. Has it ever occurred to anyone that it is Math Entrance Exam that is stirring up trouble? You couldn't make it up!


Post a comment

« The Rings of Jupiter -- Were They Disrupted by Comet Shoemaker-Levy in 1994? | Main | NASA's Hunt for a Twin 'Blue Dot' "May Yield Some Big Surprises" »




1


2


3


4


5


6


7


8





9


11


12


13


14


15

Our Partners

technology partners

A


19


B

About Us/Privacy Policy

For more information on The Daily Galaxy and to contact us please visit this page.



E