[QuietGuy]

Hello, could anyone give any comments on the following approach?

1. Take the massive of the whole language - download the whole internet in some language.
2. Treat like a collection of separate words, sentences, paragraphs, texts and dialogs.
3. Search for rules, similar to today's grammar rules that are common between those bits
4. Generate other bits like that, including question-answer pairs!

I'm not saying that this is computationally possible, but has anyone even tried to estimate how hard would it be?
This could be like a numerical upper bound on how hard it is to make AI.

[GT40]

Hi QuietGuy, welcome aboard! If you wrote this post, I guess you need a reply. Nobody answered you and I'm not supposed to have any idea, but I don't want to leave you alone.

I think it's a VERY hard task, a never-ending story, a full-time job, that's all I can say. But if you have nothing else to do... Good luck!!

[Data]

Download the whole internet? I have no idea how you could find every site on the Internet, it's not like all sites are on Google, only the sites they allow in.

But if you could, then you are going to need a huge server farm to house all that data.

Maybe Apple could contribute with some of their $60 billion profit mountain  .

Having said that,
Welcome to AiDreams QuietGuy 

[Art]

Welcome QuietGuy!!

You actually posed 3 questions....

Hello, could anyone give any comments on the following approach?

OK...Yes.
Was this approach to AI already tried?
I don't think so, given the various methods you've lumped together in your posting.
...but has anyone even tried to estimate how hard would it be?
I think the majority of regular users here did a quick roll of the eyes due the the immense nature of your suggestion. The Internet IS a very, very, very...etc., large entity and to attempt to download it would likely not be possible. It is also alive in the sense that it never sleeps, devours information of all kinds, is constantly changing and evolving and can never be killed (unless electricity of all kinds becomes outlawed).
The mere collection of data / info does NOT make a good chatbot. How and where the bot obtains the information, collects, sorts, filters out good from worthless [rejecting Spam, porn, virus / malware attacks, porn, propanganda, etc.] analyzes, and finally runs the pertinent data through its algorithms and final internal assesments before either displaying or speaking its appropriate response, is what matters.
Most students of BOTology have yet to master the method of getting the computer to have what we call an "Understanding" of our words, no matter the language. A computer simply does not know what words actually are like blue, wispy, love, swirling, inquisitive looks, wonder, amazement, emotions, etc., not to mention proper grammar, usage, synonyms, antonyms, homonyms, etc. The methods of many have been tried over the years and perhaps the bot doesn't really NEED to know how all the above sentence items mean but rather how they work or should work in conjuction with the other given parameters. It is no trivial task to build an intelligent bot. One day soon we'll see one, perhaps like a well connected bot such as IBM's Watson, and maybe even a scaled down version that can run on your own computer or smart phone (some are pretty close now like Suri , Iris, Android's Assistant, vehicle navigation devices, etc. Others will soon follow but for the single developer / user? All I can say is search, ask questions, pick some brains (talk to) of some of the creators of various bots and try to put together your own combination of approaches. There is no one "right" method...there are as many as you wish to explore.

Sorry but I know I haven't been much help with regard to the premise behind your initial postings but hopefully a bit of feedback is better than no feedback at all.

Having said that, Good luck in your quest!!

[Freddy]

Hi Quiet Guy, I have to admit that this approach broke down at point 1 for me. 

[victorshulist]

yep, already tried -- it is called WATSON.

great little search engine, but narrow ai

[QuietGuy]

Thanks for answers and encouragement!