DataBots, from Imagination Engines Incorporated (IEI), use ordinary Microsoft Excel spreadsheets to create advanced neural networks for adaptive artificial organisms and knowledge agents. These DataBots can  "...exercise their own independent judgements in perusing databases, choose their own perspective on or physically move through the data, automatically learn hidden data patterns, and cooperatively build compound cascade structures capable of autonomous discovery and invention," according to Dr. Stephen Thaler, the founder and CEO of IEI.

Amazing claims for a time when skepticism about artificial intelligence (AI) is at an all time high. The failure of the early AI researchers to live up to their predictions of human-like machine intelligence and the lack of significant advancements in the field have taken their toll on its credibility, to the extent that AI has actually become a tainted term in the computer industry. So, when Dr. Thaler contends that his DataBots exhibit behaviors core to intelligent beings, a bit of healthy cynicism is understandable.

Nonetheless, the IEI Web site does offer intriguing examples of the practical applications of DataBots, including a materials design project with the United States Air Force. In this application, DataBots discover undocumented materials for the USAF. Other interesting demonstrations of DataBots at work include the following endeavors:

• Threatening Fax Identification – agencies such as the FBI, CIA, and NSA are evaluating a DataBot that autonomously sifts through thousands of Faxes arriving within federal offices and earmarks those containing threatening messages or logos.
• Stock Market Analysis and Scenario Prediction – Swiss bankers are exploring the possibility of using DataBots to form predictive models from market data on the fly capable of laying "...out all potential scenarios, both up- and down-sides, for the near future."
• Process Control – an Air Force SBIR program involving a DataBot monitoring a CVD reactor to invent recovery paths that assure identical process outcomes.
• Artificial Life – experiments where DataBots with self-evolved network mappings of illusory "self-worth" prevailed in a competition against neural nets lacking this "motivational" factor.
• Neural Darwinism – DataBots that autonomously learn to avoid hazards within a spreadsheet, performing more efficiently and effectively than traditional Darwinistic mechanisms.
• Voice Directed DataBots – preliminary experiments that use voice recognition to initiate maneuvers such as move up, right, and find a prime number within spreadsheet.
• DataBot Code Writer – DataBots that observe raw data and autonomously write their own Java or Visual Basic code to achieve objectives dictated in natural language.
• Compression – DataBots that achieve 10,000:1, nearly lossless compression of process control data. The same principles may be extended to images, communications, etc.

So how do DataBots accomplish these feats? Traditional neural networks are trained with inputs and outputs to recognize patterns by storing numerical weights within layers of nodes. In contrast, Thaler's neural networks generate plausible patterns that nucleate from random disturbances randomly applied to their internal synaptic architectures.

To achieve such disturbances, the weights of the spreadsheet-based neural nodes of the DataBots are perturbed, by random amounts from their trained in values (see illustration below) to induce a constant state of "idea" flux, much like what is happening in the human brain according to Dr. Thaler. As a consequence, a continuos stream of new possibilities appear, which Thaler likes to equate with the human process of creativity or imagination. A second neural, acting as an overseer, or critic, guides this "creative" process and determines when something useful is discovered.

Dr Thaler points out that,"...plausible notions do appear within such an internally perturbed neural network. This is the subject matter of some very interesting articles in Neural Networks (virtual input effect) that are referenced on the [IEI Web] site, as well as the Mind II paper, 'A Quantitative Model of Seminal Cognition.' Otherwise readers imagine a process tantamount to chimps banging away at typewriters, and stochastic resonance. This is all a new and fundamental neural network effect."

Believe it or not, the nodes in this unique neural net model are built using cell references in a standard Excel spreadsheet. The illustration below provides an overview of the worksheet structure of a DataBot. Notice that each cell in the worksheet is a node, with the possibility of millions of nodes in an Excel Workbook. These huge neural nets can then be cascaded across multiple machines using the Internet.

As a clear indication of the uniqueness of this new technology, Imagination Engines, Inc. "...was granted U.S. patent number 5,659,666, entitled Device for the Autonomous Generation of Useful Information." As Dr. Thaler explains, the patent is for a "...device for generating useful information employing a first neural network trained to produce input-output maps within a predetermined initial knowledge domain, an apparatus for subjecting the neural network to perturbations which may produce changes in the predetermined knowledge domain, the neural network having an optional output for feeding the outputs of the first neural network to a second neural network that evaluates the outputs based on training within the second neural network. The device may also include a reciprocal feed back connection from the output of the second neural network to the first neural network to further influence and change what takes place in the aforesaid neural network." 

Dr. Thaler calls this paradigm the Creativity Machine.The Creativity Machine, in turn, provides the creative decision making for the DataBot, dictating where these automata go and how they operate upon data within their digital microcosm. In addition to the above patent, five more patents are about to be issued that relate to DataBot and associated technologies.

But Stephen Thaler is not content with the impressive advancements of DataBots and the Creativity Machine paradigm. He has his eye on an even grander, and more controversial, future for DataBots -- one in which cascades of these neural nets are distributed around the Internet and use the very traffic of the Net to "perturb" them into nucleating new ideas within a "World Brain." As Thaler puts it, "...DataBots may collectively self-organize into a colonial intelligence capable of human level cognition.

Considering the potentially hundreds of millions of spreadsheet neurons possible and the DataBot’s inherent self-training facility, it may be possible for DataBots to autonomously assemble first individual brain modules, link such modules, and then collectively train the overall brain structure via the self-training paradigm. The resulting agent would then be capable of conversation, reasoning, and seminal thought."

To achieve this end, Thaler is attempting to form a "World Brain Consortium" with partners willing to participate in the construction of this global spanning intellect. However, he was a bit sketchy on how members will be involved since he is justifiably concerned about the ease with which his spreadsheet-based work can be duplicated. If you are interested in learning more about the project, visit the World Brain Consortium page or write Dr. Stephen Thaler at sthaler@ix.netcom.com.

Finally, for an in depth look at some of the more far reaching and thornier implications of the Dr. Thaler's epistemological views, you may wish to check out the extensive list of articles at the IEI Web site. These papers explore the philosophical ramifications of the potential of this next generation of neural networks, the World Brain, and the Creativity Machine. They offer some very provocative reading.

To find similar software, browse through the Knowledge Bots.