Autonomous Knowledge Discovery in Personal Lives

Automated knowledge discovery will also benefit people's personal lives. With the invention of voice-activated technology like Amazon's Alexa, Apple's Siri, and Google's Home, humans are becoming increasingly comfortable with interacting with machines on a personal level. Machine learning in the social networking medium is already discovering behaviors, likes, and dislikes of consumers at a general level. The next step is systems discovering things about us on an individual and personal level. Currently, when one converses with voice-activated technology, one does not get the sense they are talking to an entity that knows them personally. We predict that will change in the near future.

With the inclusion of automated knowledge discovery, voice-activated assistants like Alexa and Siri can discover personal things about users if they build up a personal history with the users. Currently, marketing analysis associated with social media map preferences to socio-economic classes. When I am shown an advertisement for a particular product, it is because I have either searched for it or shown some interest in it via social media or the system as targeted the product to the demographic I represent. That is very different from knowing someone for the last two years, spending time with them enough to know their deepest likes and dislikes, motivations, recent experiences, etc. We foresee voice-activated assistants becoming personally and intimately knowledgeable about us and targeting content to the individual level, even proactively retrieving content because the system will know we want to see it even before we realize it ourselves.

Automated Discovery

Yolanda Gil and colleagues have developed a framework for the automated discovery of scientific knowledge called DISK for 'automated Discovery of Knowledge' (Gil et al., 2016). The goal of DISK is to enable computers to autonomously carry out the hypothesize-test-evaluate discovery cycle by searching for confirmatory or contradictory evidence to either support a hypothesis or facilitate a modification of a hypothesis. Lines of inquiry to test hypotheses of interest are initiated by a human scientist. DISK then launches one or more workflows to find and test the hypotheses against available data. DISK considers new data when it becomes available. In most scientific endeavors, there is more data than humans can possibly analyze. As shown in Fig. 12-1, DISK represents the desire to automate the data analysis part of discovery to prove or disprove a hypothesis.

Fig. 12-1: DISK: automated discovery of knowledge.

DISK gathers hypothesis statements (HS) from the scientist, and tests these hypotheses (hypothesis examination) through one or more data analysis workflows providing hypothesis evidence (HE) resulting in a confidence level being calculated. The chain of supporting analyses is retained to create the provenance for a hypothesis. If the data analysis shows the need for it, modifications to the original hypothesis can be made (revised hypothesis). When a version of the hypothesis results in a high enough confidence level, a discovery can be declared.

The interactive discovery agent is the interface between DISK and the scientist. The scientist enters the initial hypothesis and receives a revised hypothesis from DISK, if necessary. DISK has access to experimental data via the data repository and can publish results (new data/knowledge) as a result of its deliberations via the analytic workflows. DISK, to prove or disprove a hypothesis, formulates lines of inquiry and launches one or more analytic workflows. Multiple lines of inquiry and multiple workflows may exist at any time. DISK is a cognitive system able to evaluate hypotheses at the level of an expert. A human using DISK is then in a human/cog ensemble—a synthetic scientist.

In this chapter, we present our model of a synthetic scientist based on the Model of Expertise described in Chapter 7 and shown in Fig. 12-2. We call the scientist cog Ashe for Automated Scientific Hypothesis Explorer.

As in the DISK framework, Ashe uses data to test hypotheses. The data used by Ashe to test hypotheses is domain knowledge, K_D. As described in Chapter 7, these data can be accessed directly by the cog via the Internet, or via a local area network. Similar to Synclair, described in Chapter 11, Ashe is able to construct field and topic models as needed describing the domain data and domain theories and ideas.

Other types of relevant knowledge in our model, not addressed in the DISK framework, are problem-solving skills, P, P_D and tasks, L, L_D. These are necessary to carry out the analytical workflows to test the hypotheses. Generic problem-solving skills include basic analytical algorithms and processes (e.g., linear regression). Domain-specific problem-solving skills might include analytical techniques applicable to just to the domain- specific data and hypotheses (e.g., cancer gene mutation tests). Likewise, tasks may be generic or domain-specific. Task and problem-solving knowledge gives the cog the ability to evaluate and analyze possibly by further experimentation.

Problem-solving skills and tasks can be learned by Ashe based on its own processing and its episodic memory (K_E). Also, problem-solving skills and tasks can be obtained directly from remote cogs. As with Synclair in Chapter 11, Ashe will be able to consult with other scientist cogs via the Internet and exchange knowledge and know-how.

Hypotheses (HS) are domain-specific models, M_D, in that they describe a concept or belief about how the scientist thinks the world works. As with the DISK framework, hypotheses can come from the scientist or can be modified by Ashe. Unlike the DISK framework, Ashe can create its own hypotheses after analysis, evaluation, and application. As described in Chapter 7, models are dynamic data structures and can be modified. Thus, the workflow models become the medium by which Ashe executes the hypotheses testing. The evaluate skill calculates the certainty level for a hypothesis. The goal is to either find enough confirmatory evidence or evolve the hypothesis until it achieves a high enough certainty value. The understand and create skills combine to modify the hypothesis and create revised hypotheses. The understand skill is also used to explain results and revised hypotheses to the scientist. The perceive and act skills handle communication with the scientist.

Figure 12-2 shows a situation in which the human plays a central role—a synthetic scientist achieving Level 3 or Level 4 cognitive augmentation. However, over time, Ashe will evolve and be increasingly better able to hypothesize, test, and discover on its own. We fully expect Ashe to work semi-autonomously. After being directed by the scientist, Ashe is able to execute hypotheses testing on its own.

However, we expect the amount of human direction to lessen over time to a point where Ashe is operating autonomously and discovering significant new theories, laws, proofs, associations, correlations, etc. If we imagine thousands or millions of Ashes running across nearly every domain of discourse and collaborating with each other, autonomously generated knowledge can be expected to explode. The cumulative knowledge of the human race will increase by the combined effort of millions of cogs all over the world. In fact, we foresee an explosion of knowledge, an exponential growth, when cogs begin working with the knowledge generated by other cogs. This kind of cognitive work can proceed without the intervention of a human and therefore proceed at a dramatically accelerated rate. We can easily foresee the point in time where production of new knowledge by cogs exceeds, forever, the production of new knowledge by humans.

In fact, we anticipate a class of discoverу engine cogs whose sole purpose is to reason about enormous stores of knowledge and continuously generate new knowledge of ever-increasing value resulting ultimately in new discoveries that would have never been discovered by humans or, at the very least, taken humans hundreds if not thousands of years to discover.

Intellectual Property Ownership

With cognitive systems becoming able to generate knowledge on their own with minimal or no human supervision or interaction, interesting questions as to the ownership of this knowledge arises. In 2019, the Unites States Patent and Trademark Office (USPTO) sought guidance on ownership and intellectual property rights issues stemming from the use of artificial intelligence and cognitive systems to generate knowledge (Deahl, 2019).

The USPTO issued thirteen questions (USPTO, 2019):

• 1. Should a work produced by an AI algorithm or process, without the involvement of a natural person contributing expression to the resulting work, qualify as a work of authorship protectable under U.S. copyright law?
• 2. Assuming involvement by a natural person is or should be required, what kind of involvement would or should be sufficient so that the work qualifies for copyright protection? For example, should it be sufficient if a person (i) designed the AI algorithm or process that created the work; (ii) contributed to the design of the algorithm or process; (iii) chose data used by the algorithm for training or otherwise;
• (iv) caused the AI algorithm or process to be used to yield the work; or
• (v) engaged in some specific combination of the foregoing activities? Are there other contributions a person could make in a potentially copyrightable Al-generated work in order to be considered an "author"?
• 3. To the extent an AI algorithm or process learns its function(s) by ingesting large volumes of copyrighted material, does the existing statutory language (e.g., the fair use doctrine) and related case law adequately address the legality of making such use? Should authors be recognized for this type of use of their works? If so, how?
• 4. Are current laws for assigning liability for copyright infringement adequate to address a situation in which an AI process creates a work that infringes a copyrighted work?
• 5. Should an entity or entities other than a natural person, or company to which a natural person assigns a copyrighted work, be able to own the copyright on the AI work? For example: Should a company who trains the artificial intelligence process that creates the work be able to be an owner?
• 6. Are there other copyright issues that need to be addressed to promote the goals of copyright law in connection with the use of AI?
• 7. Would the use of AI in trademark searching impact the registrability of trademarks? If so, how?
• 8. How, if at all, does AI impact trademark law? Is the existing statutory language in the Lanham Act adequate to address the use of AI in the marketplace?
• 9. How, if at all, does AI impact the need to protect databases and data sets? Are existing laws adequate to protect such data?
• 10. How, if at all, does AI impact trade secret law? Is the Defend Trade Secrets Act (DTSA), 18 U.S.C. 1836 et seq., adequate to address the use of AI in the marketplace?
• 11. Do any laws, policies, or practices need to change in order to ensure an appropriate balance between maintaining trade secrets on the one hand and obtaining patents, copyrights, or other forms of intellectual property protection related to AI on the other?
• 12. Are there any other Al-related issues pertinent to intellectual property rights (other than those related to patent rights) that the USPTO should examine?
• 13. Are there any relevant policies or practices from intellectual property agencies or legal systems in other countries that may help inform USPTO's policies and practices regarding intellectual property rights (other than those related to patent rights)?

Some think if a programmer develops a system and the system

creates new knowledge then the programmer should own the intellectual property rights for that knowledge. However, think about a sculptor using a hammer and chisel to create a masterpiece of art. No one would think the person who built the hammer or chisel should have any claim of ownership of the art. So what is the difference? One difference is the hammer and chisel are tools. They do not create anything on their own. One argument then is cognitive systems like Ashe are mere tools, although sophisticated tools, and so the humans developing or operating the system should have ownership of the products created by the system.

A counter argument is the view cogs are the entities creating the new knowledge, not the programmer. Some feel artificially intelligent systems should own the intellectual property. Still others feel no one owns the intellectual property and instead it is owned by the human race at large. The Levels of Cognitive Augmentation presented in Chapter 3 come into play here. The sculptor using a hammer and chisel represent very low levels of augmentation. Any tool augments human performance. However, the situation becomes less clear as we get to Level 4 and Level 5 of cognitive augmentation where the artificial system is doing most of the cognitive work. Once a system achieves a certain level of autonomy, does it then earn the right to own the intellectual property it produces?

One of the most useful features of cognitive systems is their ability to consume vast amounts of unstructured information and process it much faster than any human ever could. However, in doing so, systems often acquire information via the Internet. Some of the information on the Internet is public but some of it is protected by intellectual property rights such as copyrights. Question #3 from the USPTO asks, and rightly so, what are the legal ramifications of one system using protected material to create new knowledge. Humans do this all of the time. Imagine the situation where a human reads a copyrighted book and learns something he or she then uses to create new knowledge. No one thinks the author of the copyrighted book has any claim to the newly created knowledge. However, it is appropriate for a person to give credit for others' ideas. Are we looking at a future where researchers and scientists must cite autonomous and semi-autonomous cogs?

Most seem to agree if a company owns a cognitive system, then it owns anything the cognitive system creates. One argument here is to equate the cognitive system to any kind of tool or machine on an assembly line. If a worker uses a powered socket wrench to attach parts to a automobile chassis as it moves down the assembly line, no one thinks the socket wrench has any more claim to the automobile than the assembly line worker. The company owns the tools, the raw materials, and employs the worker, so it owns what is produced. However, imagine a cog one day discovers the cure to all cancers, or discovers how to triple the lifetime of every human, something of tremendous value to the entire human race. Should the company own this knowledge? Should it be allowed for everyone in the world be beholden to one company (or one government for that matter) which happens to have stumbled on a discovery of universal importance? Some antitrust laws may speak to this issue, but the issue gets less clear as the value, or perceived value, of something gets enormous. At some point, many feel those kinds of things should belong to everyone. Questions like these and many others will have to be answered by the next generation.