Products

What we do

SoftPerson is currently developing a range of desktop software products using AI targeted from consumers to developer. Compared to the competition, our main differentiation is the heavy reliance on the meaning (through symbolic mathematics) of the content being modified or generated with its products.

These products have not been released yet.

Static Analysis for Bug-Finding and Code Conversion

NStatic uses denotational semantics to convert a program into its lambda calculus representation. This is a purely mathematical representation, that when simplified and normalized, allows any two equivalent programs to reduce to the same underlying representation. The "meaning" of a program can be identified with its normalized mathematical representation. For example, we can say 1 + 1 and 2 are equivalent, because the standard evaluation rules for addition reduces 1 + 1 to 2.

The benefits ofthis approach enables high-fidelity tools that understand code:

  • Bug-finding and Specifications.
  • Code conversion. Code conversion between different programming languages or different platforms/libraries within the same language
  • Refactoring.

More information can be found in http://netundocumented.com/nstatic/page/1/.

NStatic Static Code Analyzer for Bug-Finding from Wesner Moise

Descriptive Graphics

What if you can think of a computer as an personal artist, who you can dictate your needs to in order to render what scene exists in your imagination? You wouldn't need any artistic skill or advanced Photoshop knowledge to quickly produced the mental image.

Descriptive graphics is a way of contructing graphics from a top-down, high-level approach. The user can construct a scene by describing it in natural language. Thousands of preliminary scenes can then be automatically be generated by the software and then later refined through finer-level controls. Parts in a scene can be moved around and modified freely. User images and graphics can be incorporated right in. Scenes can be rendered in either 2D or 3D, and non-photorealistic techniques can be selectively applied to create images that appear hand-drawn or man-made.

The following image was constructed by a research program called "Word's Eye," which is unaffiliated with SoftPerson, but illustrates some of the ideas mentioned earlier.

More information can be found in Conversational Interfaces Redux.


Natural Language Wordprocessing

Current generation wordprocessors focus on formatting and layout of documents, but offer little help in the production of content. There are some aids offered in the form of spelling, grammar checking, document comparison, and auto-summarization. However, these aids do not rely on any deep semantic analysis.

SoftPerson actually takes written text and parses it into its syntactical structure. It then goes one step further and produces a denotational semantics representation. Various representations include flavors of predicate calculus, description logics, and natural semantics, augmented with modal and tense logic.

Some of the reasons why semantic wordprocessors have not arrived in the 1980s and 1990s, when wordprocessors underwent rapid progress:

  • Skepticism. The public is generally skeptical of the idea that computers are capable of human understanding.
  • Limited storage and power. Machine-readable dictionaries and databases require hundreds of megabytes of storage--a prohibitive cost when RAM was a few megabytes and disk drives were not yet capable of gigabyte storage.
  • Limited natural language data. Today, there are massive freely available databases for a wide variety of natural language data.

Famous AI Quotes

Some AI quotes by famous people in the computer industry.

If you invent a breakthrough in artificial intelligence, so machines can learn, that is worth 10 Microsofts.

Bill Gates in the New York Times, March 3, 2004

People always make the assumption that we're done with search. That's very far from the case. We're probably only 5% of the way there. We want to create the ultimate search engine that can understand anything. Some people could call that artificial intelligence.

Larry Page in Memepunks, May 23, 2006