Artificial intelligence can’t quite start an anti-human uprising yet, but it’s getting pretty good at understanding what people mean when they ask stupid questions.
One of the latest products leveraging that newfound ability comes from Ask-AI, which last week completed a $9 million funding round for its platform: a program that reads and digests large business documents, manuals and databases and allows users to simply enter a search query in order to yield a response to their exact question.
If you’ve ever needed the answer to a very specific question about a general subject within a multi-thousand page document, you can see why this is a big deal.
“Companies today find themselves awash in information and data from many internal and external sources and a multitude of text-heavy communication channels. Employees spend extensive amounts of time searching for knowledge and summarizing and tagging customer communications. Our platform aims to mimic Google’s success in making answers from the web directly available but in the corporate world,” said Ask-AI founder and CEO Alon Talmor.
The platform can process a multitude of corporate data sources, from manuals to guides to customer feedback sources, and learn them all well enough to formulate answers to most – if not all – questions that one might ask.
The company was only founded at the end of 2021, but has already picked up several clients who have liked Ask-AI’s platform so much, they’ve been spreading it around to others before it even officially came out.
Yanai Oron, managing partner at Vertex Ventures (one of the recent funding round’s leaders), said, “One of our larger portfolio companies is a design partner and paying customer of Ask and has expanded the use of the product to many dozens of users even before the product was officially launched.”
Natural Language Processing based capabilities
Ask-AI’s ability to conjure answers based on naturally phrased human queries is possible thanks to natural language processing. NLP has undergone significant progress in the last few years as developers find ways to train AIs with larger and more complex language models, teaching them how to dissect the way people communicate.
Yoav Shoham, a professor of computer science at Stanford, founder of the Artificial Intelligence Index and co-founder and co-CEO of AI21 Labs, a company that uses artificial intelligence language models to change how people read and write, said, “NLP seems inherently tied to progress in AI, because it captures complex human thinking.”
In an interview with The Jerusalem Post, Shoham explained that in the last half-decade, the NLP rubber has hit the road, leading to swift advancements in the way that computers understand people.
“You didn’t see a lot of progress in natural language processing until about five years ago. There was progress, but all the benchmarks were kind of puttering along and it didn’t come close to human-level performance,” he said.
“You didn't see a lot of progress in natural language processing until about five years ago. There was progress, but all the benchmarks were kind of puttering along and it didn't come close to human-level performance,”
Yoav Shoham
Now, thanks to more advanced neural net architecture, the sky might be the only limit on a future computer’s ability to understand us. As Shoham put it, “Natural language is the lens into the human mind, because there’s nothing as subtle and complex. To understand text is really to understand thinking.”