Syntience Inc.

Syntience Inc.
We understand.
Startup/business
Los Altos, California, United States United States
http://syntience.com
  • Overview
  • Team
  • Documents
Company description

For fifty years, Artificial Intelligence research has been overmuch concerned with Reasoning.  But before you can reason about something you must Understand it. We are a research company developing a new kind of machine learning algorithm called Artificial Intuition that actually understands as it learns, much like children do.

Over a lifetime, humans aggregate a database of experience in the form of discovered patterns. Our everyday Intuition is like an algorithm that uses those patterns in a straightforward and mechanical manner to instantaneously recognize new situations as composites and variations of previously seen ones. Everyday, mundane situations do not require logic or mathematics; they can, in fact MUST be, handled by simple, best-effort methods that produce reasonable answers instantly.

These ideas are well established in fields like Epistemology and in the Life Sciences. However, the AI and machine learning communities have by and large resisted adoption of the programming techniques that can  be derived from these principles. Only a few companies have moved into this new and promising area. We have been developing and adopting what has been called Model Free Methods and Non-parametric models. Google is using these kinds of ideas in their prize winning machine translation systems. At Syntience, we are attacking the larger problem of actually understanding human languages. We want to create software based systems that will actually understand text.

We have experience in web search and document understanding and know what the major quality problems are in these fields. Unsurprisingly, most of these problems would immediately disappear if computers could truly understand human languages. Our first half-dozen products will each address some critical problem in the web search indexing chain. Our customers will initially be large companies in areas like web search, OCR correction, spam filtering, and document classification. We expect to later be able to do excellent document summarization, machine translation, and speech recognition.

Investors would be investing in a new, fundamental technology, much like the Web itself. Acceptable investors should be prepared to invest small amounts for a longer term and to expect a return based mainly on technology licensing/sharing, dividends and buybacks. We strongly favor corporate investors that want to license this technology. And you should want to invest because they understand enough about our technology to realize the impact it will have. Some examples:

Imagine a computer message board where every message is spam-free, porn-free, hate-speech-free, on topic, never duplicated, and sorted by subtopic. Imagine a spelling corrector that knows the difference between "to", "too", and "two", between "luck" and "serendipity", and knows when a passive voice would be appropriate. Imagine parents of teenagers being able to trust chat system software to automatically monitor the language used to ensure everything said conforms to their standards of propriety while quietly blocking everything else. Imagine a word processor that can rewrite your draft for your target audience for you, without further supervision. Imagine human-level reliability in document classification, language translation, or speech recognition. Imagine being able to browse the Web by topic to discover information in your field that you didn't know you needed to search for. Imagine a computer without a keyboard or a mouse, or a cellphone or a car stereo without buttons, that you use by talking to it, and it talks back. These scenarios could be real in a few years if we only had computer software that could truly understand language.

For information about our business strategy, technology, philosophy, history, and news, see:

Corporate site:  http://syntience.com

Theory introduction (2 years old):  http://artificial-intuiton.com

Video site (more recent research):  http://videos.syntience.com

Corporate Facebook:  http://www.facebook.com/pages/Syntience/68992804945

Ms. Anderson's blog: http://monicasmind.com

Ms. Anderson facilitates the Bay Area AI Meetup: http://ai-meetup.org

Ms. Anderson's Facebook:  http://www.facebook.com/pandemonica

Ms. Anderson on Twitter:  http://twitter.com/pandemonica


Business model

Currently we are in the research and development phase, developing a fundamental technology - Software that actually understands the semantics of text. The first goal is to create a convincing demo of this competence.

 

After that we intend to create a multitude of software modules around this capability and license these modules to major companies in any business that requires understanding human language expressed as text, such as web search indexing support, corporate document search, document classification, spam filtering, customer support systems, spelling/grammar/style correction, etc.

Competitive advantage

The Problem:

 Current computer systems do not understand human languages.

This is widely recognized as a major problem. One typical example: Search tools (for the web and for other collections) operate by returning documents that contain words the user enters in their query. But words are ambiguous; the meaning of a word depends on the other words in the sentence containing it, on the order of those words, on the topic being discussed, the language used, etc. This ambiguity causes the good results to be diluted with irrelevant web pages.

As users, we would like our tools to really understand our queries, to understand all web pages, and return as results only the documents that actually discuss what we are asking for. But true understanding of what a sentence of text really means - its “semantics” - is beyond the reach of current computer science. No existing systems can even get close. Some web search systems claim to do “Semantic Search”, but they still really operate at the level of “syntax”, that is, by using words and grammars. Countless computer based applications – from games to automated customer support systems – perform poorly for the single reason that the simplified language the software is capable of handling is incomplete, inflexible, and brittle.

Our Solution:

 Syntience Inc. is developing systems that can learn any human language the way children do – by being exposed to some language, such as English, and learning it by example, as if by osmosis. A breakthrough in this area would not only solve all quality problems in web search; it would enable thousands of new kinds of applications that we currently cannot even imagine.