Just a few months ago, a lawyer and his New York-based law firm barely dodged sanctions from a judge, after submitting an artificial intelligence (AI) -assisted legal brief that contained six completely fictitious cases. The lawyer had used ChatGPT to help research the brief, but things went awry when the AI’s algorithm hallucinated some of the case law.
In theory, AI can add value to virtually any profession, including the practice of law. However, as illustrated by the fake case brief, current AI models can also create potentially serious problems for attorneys and their firms, from an ethical standpoint.
All licensed attorneys, those acting under their supervision and those supervising them must abide by the rules of professional responsibility for their specific jurisdiction and practice. While these rules vary from state to state, the American Bar Association’s Model Rules of Professional Conduct provide a solid baseline. Among other things, they require a duty of “Candor Toward the Tribunal.” This means that attorneys must refrain from knowingly making a false statement of fact or law or offer false evidence to the court and that they proactively disclose case law that is adverse to their stated position. This rule, or the New York equivalent, became an issue in the fake case brief. The judge chose not to ultimately penalize the attorney, because alleged he did not know that AI could hallucinate cases.
By now, most of us know better: current AI systems are prone to hallucinating, or providing confident responses not justified by its training. They can give inaccurate information, come to conclusions consistent with their missions but possibly adverse to humans or can simply make things up by mashing existing pieces of data together to create a new thing (as in the case of AI-generated art…or fake case law). Any or all of these outcomes can happen because current AI systems, such as ChatGPT, rely on a generative model, which incorporates neural networks and machine learning based on algorithmic statistics and probabilities. Essentially, these AI systems are nothing more than word predictors that fail to display true intelligence. They don’t “think,” as humans do, or might believe they do. In short, they lack a basic level of consciousness.
The next iteration of AI, artificial general intelligence (AGI), provides a higher level of functionality to support human endeavors. Valmiz, Inc. has been working on AI to do just this. They call it Valmiz, a Valmiz Aurora (VA) software. This alternative to contemporary AI, has been in the making for the past 20 years. Its goal is to enhance informational reach to inform better decision-making, while keeping the human, and human morals, ethics, and values, at the center of any given process.
Unlike current AI that uses a single algorithm, the engineers that created Valmiz Aurora designed a system using multi-agent algorithms. Valmiz Aurora ingests a client’s own raw, internal and organizational data to create an enterprise-level super knowledge base. Because the primary inputs into Valmiz Aurora are an organization’s own, they come pre-validated.
Here’s how it works: Valmiz Aurora combines several individual agents – Veda, Vera, Vela, Vega, and Vix – as part of a larger whole. They can work separately or together. Vela searches the different data sources in a client’s network (or the internet, if desired) to gather and compile it. Veda, the core Al system that binds everything, fuses knowledge graphs and bases together. Vega is a resilient data storage system that ensures continuity of operations, even in the event of a power Interruption. Vera tracks key-value-metadata changes across different data sets. Vix is the human-machine interface. It receives commands and communicates back to users similar to current AI assistants but smarter because it gathers information from the other agents of Valmiz Aurora.
Valmiz Aurora is like having a thousand highly qualified experts to work on a task to provide input to the human operator at the center of the operation. This augments his or her capabilities. It enhances decision-making with powerful, precise and accurate information.
For ease of use, Valmiz Aurora can connect directly to a user’s system as a desktop app that can be deployed on a machine or mobile device or by web API or it can be inserted as part of an organization’s pipeline. For attorneys, Valmiz Aurora can integrate with other systems, such as a law firm’s internal workflow system. It can also presumably link up with respected proprietary systems, such as the LexisNexis search engine.
Valmiz Aurora’s compounding capability can streamline and validate legal research and writing tasks. Using Valmiz Aurora, an attorney can type a few simple keywords (as opposed to long string queries) into Vix. Vix will then search the firm’s internal database and others, if asked. Because the original data is internal to the firm, the compiled information can be trusted. With regard to any external data, Vera verifies and validates that all information sources actually exist. Using the Valmiz Intelligencealmiz Aurora to assist in research for brief writing prevents the fake case brief scenario.
It can also preclude a legal office from “reinventing the wheel” or providing inconsistent advice. Too often, In both civilian law firms and military Judge Advocate’s General Corps legal practice, individual attorneys create work products that remain largely inaccessible by others in the same practice or firm. Often this work remains siloed on one’s own hard drive or in the minds of long-standing employees with corporate knowledge. Even when written opinions or briefs are included into some kind of knowledge management systems (KMS), these systems are rarely searchable. This often results in attorneys repeating work previously done by others on the exact same issues. At worst, these inefficient practices could result in conflicting opinions emanating from the exact same office.
Valmiz Aurora, on the other hand, dynamically updates information by continuing to run and search out information based on keywords that the humans give it, to provide up-to-the-minute precise information. In this way, Valmiz Aurora can change the game by allowing for easy access to past opinions, as a starting point for current research and writing projects.
Valmiz, Inc. can also provide great value on either the receiving or giving end of the discovery process. Some cases literally involve pallets full of information. Valmiz, Inc. designed its Valmiz Intelligence specifically to address great volumes of data.
For example, imagine a Request for Production of Documents that seeks medical records addressing a client’s cervical spine injury. A simple query (e.g., “C-4, C-5, cervical”) would scour these records and produce an accurate list of these records in mere hours. Contrast this with the days that would normally be required in a hand search of these same records.
Conversely, with the same rapidity and precision, an attorney could use Valmiz Aurora to review similar records that the firm received from an opposing counsel to find all the documents that mentioned the cervical spine injury.
Attorneys could also use Valmiz Aurora to review actual radiographic imagery provided in discovery, and use its backend analytics to provide valuable insights. Valmiz Aurora could compare the imagery, over time, to identify physiological changes in the C-spine, for example. Valmiz Aurora captures all the previous incarnations of the area being evaluated. Vera does the comparison, using data in Vela. Vix allows the user to communicate to the system via keywords and bring them together and provide the required result. Thus, instead of just looking at one image at a time, Valmiz Aurora could connect the dots over a series of images over the years. This could provide an advantage to a firm’s medical expert, while reducing review time and costs.
These are but a few examples of how legal practitioners can employ Valmiz Aurora in support of their legal endeavors. The Valmiz Aurora novel multi-agent approach can streamline efforts while providing peace of mind that the information received, and ultimately provided to others, is not hallucinated but rather real, authoritative, accurate and valid. In a legal setting, where attorneys and law firms cannot gamble on AI that makes wrong guesses or conjures up case law that does not exist, Valmiz Aurora will take AI to a higher level for the legal community.
