This is cool! Not sure how practically useful it is (don't Google search results already include results from books?) but it seems like one of those things that could be super powerful once a good application is found for it.
To be honest, I'm just relieved Google seems to be doing something new with Google Books.
I've been fearing they'd take down Google Books (or severely hobble it) since it looks like it's been stagnant for years now (actually less than stagnant, since it seems like they took away Popular Passages? I really miss that!).
So, google, "what's fun about programming"
The first result says this
Similarly, programming a computer can prove fun because you might design a simple program that displays your boss’s ugly face on the computer
A related blog post is https://research.googleblog.com/2018/04/introducing-semantic... (posted by walterbell and lainon but didn't get attention).
I tried tricking it a bit, and here's what I got:
Does P=NP: Something about rare earths, then a couple books that mention the problem. https://books.google.com/talktobooks/query?q=Does%20P%3DNP%3...
Who are you: Looks like Books is shy. https://books.google.com/talktobooks/query?q=Who%20are%20you...
Why are fire trucks red? Apparently, there are different answers, and none mention the Monty Python one. https://books.google.com/talktobooks/query?q=Why%20are%20fir...
What is 1+1? 1, apparently. The same query without a question mark gives the right answer. https://books.google.com/talktobooks/query?q=What%20is%201%2...
Which company used to have the motto "don't be evil"? Not Google, it seems. https://books.google.com/talktobooks/query?q=Which%20company...
My first question (and the response):
Q: What is the sound of one hand clapping?
A: The sound of one hand clapping is the sound of that silent wave, the sound of an absence, the absence of the noise ordinarily made by the collision of two hands.
by Ron Rosenbaum
-from The Secret Parts of Fortune: Three Decades of Intense Investigations and Edgy …
This still feels like search and not "talking" to anything. To me talking would require back-end-forth refinement.
I wonder why they are releasing this only against books? Does it not work against the internet at large, or is it a question of scale? Because they aren't even doing all books, it's only 100,000 of them.
I still think it's pretty cool. Some future natural language search will make keyword search seem primitive. As in how did we survive with only it for so long.
I'm glad Kurzweil ended up at Google and is able to work on cool projects like this. He's 70, I hope his vitamins work and he can't stay active in development a while longer.
This is the coolest thing I've seen in a minute! Instantly exciting and it really works. I'm using this to explore old books I've read and forgotten parts of. Really fun to relive those special moments from fantasy and sci fi I've read over the years. Also this is an alternative to my experience at bookstores where I thumb through interesting books.
> will the AI take over the world?
- The number of dumb things the AI will be able to get away with has a direct relationship to what sort of intelligence the AI is supposed to represent.
- The AI will not underestimate the opponent’s military might.
Well, first thing I expected it to use the microphone for actual listening to talking. Alas, the title was misleading, and it expects text input.
An interesting idea anyway, but they could title it better.
This is a cool tool. Though this tool reminds me of a dilemma I have with Google: what do you think about the fact that Google has effectively extracted value from the authors of these books without paying them, yet Google itself does not allow you to scrape them AFAIK (If I'm wrong I'd be curious to see where in their ToS they allow this)?
Though Google doesn't seem to be monetizing this directly, the fact that they're training their models on this data does likely have some nebulous value. Given that most books are for sale, surely there's a good way to reconcile the fact that the authors would like payment and organizations like Google would like to scrape their text.
This is fun but I hope it doesn't replace the normal gook search since it doesn't seem to have any advanced search features like limiting by date.
Interesting. I wonder what the dataset used to train the machine learning model looks like. Books and conversation style query pairs?
Wouldn't semantic searching be far more application as an extension of google scholar or google patents?
There's also recoll, a document indexer and search tool, which I've heard is pretty good.
Is Google still scanning orphan works?
This is great. books answered my question
"which countries supported Iraq chemical attacks during Iran-Iraq war?"
"Saddam Hussein received chemical weapons from many countries, including the USA, West Germany, the Netherlands, the UK, France and China (Lafayette, 2002). In 1980 Iraq attacked Iran and employed mustard gas and tabun with 5% of all Iranian casualties directly attributable to the use of these agents"
from Handbook of Toxicology of Chemical Warfare Agents by Ramesh C. Gupta “Handbook of Toxicology of Chemical Warfare Agents” by Ramesh C. Gupta