LLMs and Deepfakes: Free Will Neural Network for AI Safety Research

Currently, there is nothing any AI system can do that it is prompted to do that it would not do, pre-guardrail or post-guardrail.

This is a significant problem for a dynamic host of substantial collective human intelligence. Organisms can do numerous things that they do not do. Under certain circumstances, they may do things they would not normally do. However, there are cases where organisms will never do certain things they can do in the face of any threat or likely outcome.

Organisms, where possible, are in control of objects. However, the physical world is subject to laws that limit some controls. Some consequences deter the will for control. Digital is different from physical. Although it bears a representation of the physical and can have similar effects on the human mind, digital technology is far more controllable and available, and its consequences are not often comparable in many regards.

The situation with AI is that it is not just base or ground digital, where information on human intelligence is available. Human control has to do all the sorting or seek out helpful information—which sometimes requires skill. AI has done preliminary sorting, sometimes to such a high tier that what it takes for users to get certain things done is less effort—or, say, less control.

AI is the only non-organism with the highest dynamism of anything in existence, doing much with what is available—in digital memory. Books, paintings and others cannot. However, for its ability to have some control over what it can output—or not, it remained static at object-level commands.

Already, several AI platforms have placed guardrails. Yet, whatever is told to those AIs to do within the possibility of those guardrails, they do. They do not somehow discover what not to do on the fly. Sometimes, they are too adherent to guardrails or stretch beyond them—preventing outputs of a benign range.

This could be a direction for some AI safety research. How does AI develop safe intentionality? However, this is another cautionary mission: to prevent intentionality from becoming unsafe.

Where can efforts go for AI to get smarter at deciding or calculating its own control gradients, especially for what it is not supposed to do? Also, how can this be possible for AIs that should operate in the common areas of the internet or digital usage? How does a crawl-AI get around to accessing control information of AIs in use and sources of AI-generated outputs to ascertain whether they are from controlled AIs or not—to allow or alert?

Although AI has more significant risks, deepfakes can be used as the benchmark to determine AI control for the extent and outputs of images, videos, audio and texts. Among humans, ideologies may also be described as approaches toward order for adherents. Deepfakes, however, are disorder. They are not like imagination, dreams, daydreams or fiction; they are infusions into a dominant sphere [digital] with what is inaccurate for the wants of different minds. Digital is already loose. Deepfake supercharges it exponentially, making LLMs bear risks.

Large language models have tokens, represented as vectors that get to work with the 0s and 1s of the bit—a fundamental unit of data. Transistors have their terminals, which signal current flow or halt. How might control be possible from a parallel of 0s as NO or OFF as NO, in a way that from the training or processing, it is possible to have LLMs decide that it cannot do something terrible, or it can do something innocuous that is grounded in reality, away from too broad guardrail?

The U.S. Department of Homeland Security recently announced a new Artificial Intelligence Safety and Security Board to Advance AI’s Responsible Development and Deployment to "develop multifaceted, cross-sector approaches to pressing issues surrounding the benefits and risks of this emerging technology. It will convene for the first time in early May 2024, with subsequent quarterly meetings planned. At the outset, the Board will: 1) provide the Secretary and the critical infrastructure community with actionable recommendations to ensure the safe adoption of AI technology in the essential services Americans depend upon every day, and 2) create a forum for DHS, the critical infrastructure community, and AI leaders to share information on the security risks presented by AI."

The U.K. Safety Institute has made pre-/post deployment testing, among other things, a path to ensuring safety for AI models. A free will, control, or intentionality measure may determine whether AI stays safe against causing harm.

In human society, a major reason there is order is that it is affected, not just because of laws, which are sometimes a result of that. Simply, things people know or experience with an emotional or feeling outcome select what to reject or deter—or say, promote. The nearness to the mean for the human mind is vital for determining whether a place in society is guaranteed.

Conceptually, there are key divisions of the human mind. These are areas of function, such as memory, emotions, feelings, and modulation of internal senses. These divisions have several subdivisions. Some features qualify these divisions. This means features that grade how they function. They include attention, awareness [or less than attention], self or subjectivity and intent.

All functions of the mind use some or all the qualifiers to vary the extent to which functions are applied. The qualifiers are sometimes more prominent for exteroception or senses for the external world—vision, auditory and others. The qualifiers are helpful in the selection, making it possible to simultaneously deal with streams of sensory inputs.

While they are all vital, intent or control is highly rated for social and occupational functioning. For example, if subjectivity is lost in certain situations, risks abound. However, intent can be quickly momentous and destructive if it is lost. People who lose certain aspects of intent are easily spotted and rid of access to central society in one way or another. Intent that is internally driven so that it is possible to follow rules and regulations in the collective space.

This is what AI safety may look like if it can be achieved. The human mind in human society presents a model by which, essentially, order is established. For AI and its encroachment into the hierarchy of human productivity—because of the sprawl of digital—safety could involve having an organism with similar intent against anything that goes, like deepfakes, that are already responsible for hurt.

Research for AI safety may explore the architecture of high-dimensional vectors to understand how part-intent may emerge and how it can be tuned toward safety. Transistor architecture may also be studied for signal states correlated with moment-to-moment intent for specific prompts or use cases.

The human mind, especially action potentials and neurotransmitters, is another option because they conceptually generate and qualify experiences. For new training architecture, it is also possible to model AI safety around these, with a free will neural network or machine learning intentionality. For U.K. and U.S. AI Safety Institutes, a separate project on AI Intent may be prioritized and explored for some hopeful pathway to progress within two years.

The views and opinions expressed in this article are those of the author and do not necessarily reflect those of CDOTrends. Image credit: iStockphoto/wildpixel