Perspectives on LLMs
Boosting Business or a Risky Gamble: Enterprise Perspectives on Large Language Models
Recent breakthroughs in large language models (LLMs) have made it possible to create almost superhuman agents for a variety of business-related tasks. These models have proven valuable especially for generative use cases: creating code, summarizing documents and synthesizing knowledge – while engaging in conversation as increasingly adaptive assistants.
OpenAI has been at the vanguard of this progress, publicly releasing disruptive products faster than public discourse – let alone business practices – can keep up with. Yet amid all the resulting attention and, often, confusion, it is important to highlight that some challenges will inevitably require more than a single transformer architecture to solve. General AI is still far from reality, and LLMs will continue to operate within limitations.
For the above reasons, integrating such technologies into a business scenario involves risks. These notably involve matters of privacy and security, the sharing of sensitive data and ensuring factual correctness for business-critical queries. In our talk, we provide a data scientist’s perspective on why these concerns arise and how they could be mitigated to effectively leverage LLMs in an enterprise environment.
Szilvia Hodvogner holds a degree in computer engineering with a specialization in artificial intelligence and computer vision. She has extensive experience working for research-oriented companies, where she developed a wide range of predictive models and natural language processing model applications. As a data scientist at Starschema, Szilvia currently delivers solutions for companies via advanced analytics and machine or deep learning with a focus on NLP and GIS-related projects.