Hacker News Reader

Comment by thangalin | Hacker News Reader

While playing Go at a local pub, one of the servers, after being told about AlphaGo, said it was scary. We talked of Siri and other forms of artificial intelligence and her fears were assuaged with a well-known question that humans handle effortlessly.

    A glass ball falls on an iron table, and it shatters.
    An iron ball falls on a glass table, and it shatters.
    To what does "it" refer to in the previous sentences?

She didn't know that such a simple question could stump computers.

It's anecdotal, but leads me to think that an average person doesn't understand AI, much less the difference between AI and AGI. From their perspective, machines that can answer simple questions ("How do I get to the nearest movie theatre?") are as knowledgeable as anyone.

View on HN

While playing Go at a local pub, one of the servers, after being told about AlphaGo...

I was expecting this to be the set-up for a joke :)

A server, a client, and a user walk into a bar ...

It would surprise me if there wasn't already an AI that could handle that class of question. Or, at least, I'd imagine it being relatively straightforward (non-trivial, but straightforward) to develop an AI that could handle it.

This question requires either understanding or experiencing physics: gravity (a force that can cause objects to fall) and properties of glass (brittle) and iron (not brittle). The question also contains certain assumptions. For example, that the (hollow) glass ball is smaller than the (well-constructed, non-rusty) iron table. Lastly, answering this question depends on a deep comprehension of English combined with complex memory abilities.

If someone has solved this particular type of problem (Winograd Schema), I'd be keen to know.

https://www.cs.nyu.edu/davise/papers/WS.html

https://en.wikipedia.org/wiki/Winograd_Schema_Challenge

http://commonsensereasoning.org/problem_page.html

http://spectrum.ieee.org/automaton/robotics/artificial-intel...

couldnt you solve it with just a large enough dataset of language and grammar.

it refers to a breakable object made of either glass or iron. the ai then scans usages of the word shatter to see if it more commonly is used to describe glass or iron.

i searched google. "shatter iron" returns 500,000 results. "shatter glass" returns 2 million. 4/5 chance it refers to the item made of glass. in quotes the phrases return 4,640/186,000.

You can solve it for any specific instance of this structure. To solve the general case, you would need to teach an AI of the physical properties of and relationships between every word in the chosen language. That's much more than just language and grammar.

Wikipedia has another example (the original Winograd Schema) that doesn't allow you to "solve" the problem by asking Google:

  The city councilmen refused the demonstrators a permit because they feared violence.
  The city councilmen refused the demonstrators a permit because they advocated violence.
  To what does "they" refer in each sentence?

Plot twist: The iron ball had just been submerged in liquid nitrogen and it was actually the thing that shattered.

Reply on news.ycombinator.com