Google’s new AI search tool is a stunner and book lovers now have a new way to explore titles – both fiction and non-fiction. It brings finding text embedded in books to a whole new level. The tool, called Talk to Books, allows you to type in a question or a statement and instantly receive responses “that would most likely come next in a conversation.” The speed in which responses are returned is truly impressive. If you type in a question the responses, gleaned from over 100,000 books, provide answers in the form of sentences that are shown in bold along with surrounding text to help with context. You can click on the linked title to find additional information about the book such as publication date and publisher. There also is an option to view the passage on the actual book page.
According to Kurzweil Accelerating Intelligence website, the tool, based on semantic search, was developed by a Google team headed by noted futurist Ray Kurzweil (inventor of the first text-to-speech machine for the blind in 1976 and many other tech breakthroughs). “Semantic search is based on searching meaning, rather than on keywords or phrases. Developed with machine learning, it uses ‘natural language understanding’ of words and phrases.”
Probably the most important thing to remember when using Talk to Books is that it starts at the sentence level, rather than the author or topic level. This makes it an entirely different search experience. It helps to visually keep this in mind when forming queries or statements. Google clearly states that some experimentation is needed to use the tool effectively. Also, it is evident that you will receive responses that are simply off the mark.
My first thought when experimenting with the tool was how are the responses ranked and why are they shown in the order that they appear? An article appearing in Steemit does a nice job explaining how to think about responses: The model is trained on one billion chat sentences and learns to identify which ones might be good responses. Once you ask a question (or make a statement), the tool searches for all the sentences in the 100,000 book, finds the content corresponding to the input sentence according to the semantics of the sentence level; there are no preset rules that limit the relationship between the input and output results.
Since the responses are pulled from mostly books, I would not recommend it for current events, company information, medical research or any other topic that requires continuously updated information. But that is just fine because the intent of the tool is for book discovery and literature review. The beauty of the tool is that any topic you can ever imagine is probably covered!
For example, these statements and queries worked really well:
What is the scariest movie?
What is a kpi in procurement?
Tours for foodies
Why do horses sleep standing up?
Most important cubism artists
What are Peter Drucker’s views on management?
I did not however, have as much luck with this question: What can Procurement do to get a seat at the table?