Lucid is an Artificial Intelligence company that is developing the causal reasoning system Cyc to solve complex business problems. PlanetTech interviews Michael Stewart on the history and future of the project, as well as Lucid’s ideas on what’s coming up in the world of AI.
What is Lucid AI, how did it come about, and what is your historical relationship with Cycorp?
More than 30 years ago, Dr. Doug Lenat, chief scientist and founder of Cycorp and Lucid, predicted the future of problem solving was not in conventional computing, but in an intelligent system that could understand the meaning of the information it was processing- similar to a human.
The platform he created – Cyc – stems from the shorthand of Encyclopedia. Cyc is the 36 year-old brainchild of Dr. Doug Lenat, initiated at the MCC Research Consortium and still continued through Cycorp, Inc. in Austin, TX. Cycorp is a company focused on research and development projects for some of the world’s largest organizations. All profits have been put back into the development of the Cyc knowledge base.
Dr. Doug Lenat and Michael Stewart met while working on separate projects at the MCC building. Michael was working on an Autonomous Software Agent Architecture when we was introduced to Doug and saw firsthand how Cyc’s complex architecture could have a significant impact on the future of AI. About eight years ago, Doug and Michael envisioned creating Lucid together to commercialize the Cyc technology. Lucid is in the process of acquiring Cyc and merging the two companies into one corporation.
What is the main technological idea behind the Cyc project as opposed to other approaches to AI?
Cyc’s platform connects data in a way that pulls in industry expertise to transform information into actionable insights for organizations. Its ability to handle complex dependencies, vast amounts of data and tackle time-consuming tasks is unmatched.
The platform embodies the original architectural belief that strong AI would eventually function systemically by applying knowledge of how the world works and leveraging this through neural networks, mimicking how humans think. Doug and his team stuck with this original symbolic-representation architectural approach, despite its immense challenges. This technology has amassed the world’s greatest symbolic representations of knowledge and multi-faceted modes of humanlike reasoning capabilities, while also enabling math-based solutions like other AI approaches.
Cyc goes beyond pattern matching or making data easier for users to understand. It offers an intelligent system that can understand the meaning of the data it’s interpreting, similar to a human. Cyc intelligence is the only AI platform capable of sharing why it got to the answer, and this additional rationale around the data helps organizations solve issues faster and smarter. By explicitly capturing and representing knowledge about the world, specific domains and user’s tasks, a strong AI can answer questions and provide explanations based on the same concepts and reasoning a human uses. This is a very different approach from more analytically-based AI because while it may provide useful functionality, it doesn’t have any understanding or rationale of the answers provided.
What are the main competing methods to develop strong/general AI?
It all comes down to the insights an AI software can provide. This can make the difference between strong and general AI. Is it an opportunity for simple correlations or is it causation? Intelligence requires more than correlating data, which can link concepts to one another. Cyc’s reasoning engine goes beyond correlation by providing insight into how it derives answers and the potential effects.
The direction the research scene seems to be going is deep learning neural networks – do you have any plans to build deep learning into Cyc?
Deep learning is already present in Cyc’s operating system. It’s this deep learning capability that differentiates Cyc from other AI offerings. Most notably, it can understand structured and unstructured data, bring insights to it and derive valuable conclusions from it. It can do all of this with much less human analysis and interpretation required compared to other AI solutions available.
Now that Cyc is established as the world’s only functioning symbolic-AI platform, it will be used by Lucid in the coming years to extend Cyc’s learning modalities into other functional forms of learning-types, including deep neural, statistical co-occurrence, visual interpretative methods, natural language and other auditory methods.
Expansion into these areas will allow Cyc to compile a cumulative learning machine architecture capable of sensing and understanding the world comprehensively with a combined collection of multi-sensorial modes of information inputs. This process is similar to the human reality assembled in the brain from inputs of the five senses.
Where do you see Cyc in 10 years?
Today, we’re focusing efforts on the optimal business models to drive Lucid’s product development, sales and implementation. We’re currently fully engaged with finance and healthcare markets. During the next 12-18 months, with incremental investment, we’ll be able to more aggressively pursue the demand we see in energy and retail supply chain markets.
Within 10 years, we expect Cyc to expand its distribution to nearly all industries and geographic cultures worldwide. Imagine AI helping economists avert the next financial crisis, enabling engineers to create more sustainable energy options and doctors to cure cancer.
Where do you see the field of AI in general in 10 and 20 years?
From personalized healthcare to banking, energy and education, AI can improve nearly every aspect of our lives. Since AI is beneficial across a diverse set of industries, there’s no limit to who can benefit from this technology. We see AI being adopted on a global scale
We recognize two key trends that will affect the future application and development of AI– the vast amount of data being acquired rapidly today coupled with the exponential increases in computational horsepower needed to reason across this data. These trends have created the need for humans to employ AI technology to help derive value from the mountains of data available to make informed business decisions. As AI becomes integrated into homes and businesses, Lucid is working to create programs for personal and enterprise uses of the Cyc software.
By making connections among data, Lucid can amplify human intelligence one word at a time. This advanced technology has the potential to improve the quality of living worldwide now and for future generations, making it possible to live better and more meaningful lives.
What is it about Cyc that makes it a general AI as opposed to narrow? Do you see it as being truly general in the sense that it could be at least as adaptable as humans; could it control a body (robot), learn a variety of game playing, problem solving, and sensory interpretation?
Just as we weren’t born with more than 30 years of knowledge in place, Cyc wasn’t created overnight. We’ve taught this AI platform truly general information like humans are taught – one concept at a time. More than three decades in the making, Cyc’s knowledge base goes beyond simply recognizing data by actually connecting it with other pieces of data to understand the relationships between them, and draw conclusions. It can learn any of these processes with time.
What in your opinion is the biggest problem in creating the truly human-like AI that we see in science fiction? Is it that that we don’t have the hardware to run such an advanced AI? Or we can’t implement the software correctly? Or is it more basic and we just don’t have any good ideas how to do it?
Similar to Maggie Boden, research professor of cognitive science at the University of Sussex, we recognize the biggest challenge isn’t with specific AI tools. The real barrier to human-like AI is the need for a “central nervous system” that assembles all the pieces of AI, and the three to five decades and hundreds of millions of dollars required to make this vision a reality.
Assembling and interpreting what AI is “seeing, hearing or touching” is where the real challenge lies, and this is where Cyc intelligence thrives. It goes beyond pattern matching to actual understanding.
Has the development of AI progressed faster or slower in the last 30 years than you anticipated? Why do you think this is the case?
AI development has most certainly picked up the pace in recent years, as humans become more overwhelmed with the data we create. There must be a system put in place to help us make sense of it all. This need has led to multiple advances in AI computing capabilities.
What organizations, apart from your own, do you think will make the most impact on the future of AI and what groups are doing the most interesting work?
Google and IBM are two leading organizations with vast resources available to make an impact on the future of AI. Lucid is ready to commercialize Cyc intelligence to disrupt industry applications like never before.
What AI project so far have really impressed you, made you think, yes, that is really a leap above everything else out there?
Cyc and only Cyc, really.
What do you think of the concept of the Singularity as it relates to an intelligence explosion caused by exponentially improving general AI, whether we improve the AIs or they improve themselves (or a combination)? How seriously do you take this idea?
Very seriously, and lucidity is the result of singularity. That is where Lucid got its name, with Cyc as the core enabler.