Advancements in artificial intelligence (AI) are reshaping industries by introducing ground-breaking technologies capable of processing vast amounts of data efficiently. The integration of AI models, particularly those utilizing token processing, is leading to new operational efficiencies, as highlighted by recent developments from prominent entities like NVIDIA and Foley & Lardner LLP.
According to recent insights from NVIDIA, AI models heavily rely on tokens—small units of data stemming from larger information chunks. These tokens are fundamental to how AI learns relationships and enhances its capabilities, including prediction and reasoning. For businesses hoping to maximize their AI investments, the efficiency with which tokens are processed can directly impact profitability. NVIDIA's innovations, such as AI factories engineered to handle high-volume token processing, allow enterprises to streamline their AI workloads at significantly lower costs. A notable improvement was reported where integrating software optimizations with the latest generation of NVIDIA GPUs resulted in reducing cost per token by 20 times, yielding 25 times more revenue within just four weeks.
This progressive approach to tokenization exemplifies how financial viability can be enhanced through the mastery of AI technologies. By effectively breaking down data, AI factories can translate complex datasets, moving from raw information to meaningful insights—fueling the new industrial revolution powered by AI.
Another practical application of AI technology is seen at Foley & Lardner LLP, where the firm, alongside Relativity, began exploring GPT-based products as early as 2023. This collaborative effort culminated in the creation of the “aiR” suite of AI tools, significantly streamlining legal processes, especially within multilingual contexts.
One of the most notable experiments involved analyzing Spanish language documents for an internal investigation. Traditionally, bilingual review processes demand excessive time and resources, as they necessitate fluency for accurate analysis and translation. With the aid of aiR, the legal team was able to bypass these hurdles. They initially drafted prompts based solely on English, prompting the AI to identify five core issues relevant to their investigation.
Results from this trial were astonishing. aiR demonstrated strong comprehension of both Spanish and English, delivering analysis and citations quickly. A series of validation tests assessed the accuracy of AI-identified issues at both record and issue levels, resulting in verification rates exceeding 99%. The use of aiR effectively eliminated the previous needs for document translation and time-consuming reviews. Instead, English results became available overnight, allowing the team to move quickly to advise clients—a considerable time saving and efficiency boost.
The key to successful deployment of aiR at Foley lay within the establishment of prompt criteria. The input parameters underpinning aiR’s analysis were refined iteratively, taking initial results and human feedback to modify prompts accurately. For example, early tests involved 50 already identified documents, allowing the legal team to evaluate how aiR interpreted these prompts against known results. Regular feedback led to revisions, ensuring more accurate categorizations during future iterations.
Once the prompt criteria were set, aiR was applied to analyze 2,292 records related to the investigation, generating findings on relevance and whether documents were suitable for review—589 were identified as relevant or borderline. After quality control (QC) checks, only two documents from aiR’s recommendations were incorrectly classified, against thousands of documents, highlighting the technology's reliability. Impressively, less than 1% of suggestions were later overturned by human reviewers.
Beyond efficiency and accuracy, aiR’s capabilities hint at broader opportunities for businesses facing multilingual challenges. By transitioning the majority of the analytical workload onto AI systems, firms can navigate language barriers effectively, allocating resources to hire subject-matter experts rather than multilingual skills. With generative AI's insights aiding the interpretation of foreign language documents, organizations can engage experts based on their knowledge rather than their language fluency.
Foley’s experience with aiR demonstrates the considerable potential of generative AI technologies to impact various industries positively. Integrations like these could define the future of data-driven applications. By eliminating language barriers and expediting document reviews, firms can realize significant time and cost savings, making generative AI not just preferable but necessary for future advancements.
By embracing AI technologies—and particularly the capabilities to process tokens efficiently—agencies can prepare themselves for the challenges and opportunities of tomorrow. Staying at the forefront of AI innovations, as demonstrated by both NVIDIA and Foley & Lardner, provides companies with insights and tools to maximize their competitive edge and capitalize on new revenue opportunities.
