The News: On September 5, Gleen announced it has raised $4.9 million to accelerate its work in solving a major issue with large language models (LLMs)—hallucination. Gleen AI is focused on improving LLM-based chatbots that are focused on customer support/customer service, and it is now publicly available.
LLM-based chatbots tend to hallucinate, responding to queries with completely fabricated information. To address this problem, Gleen created a proprietary AI layer, independent of the LLM, that ingests enterprise knowledge across multiple sources, manages it, selectively feeds knowledge to the LLM, and cross-checks the quality of the LLM’s response, eliminating hallucination. Gleen AI is LLM-agnostic. It currently works with GPT 3.5 and 3.4, Anthropic, and Llama, and it is integrated into Slack, Discord, and other leading help desk solutions. Gleen provides software development kits (SDKs) and REST APIs for customers to integrate directly. Read the full blog post on the public availability of Gleen AI on the Gleen website.
Gleen: Solving LLM Hallucinations
Analyst Take: It is interesting that given the potential impact of LLMs there is so much work involved in making them behave properly. Hallucination is a big challenge. Is Gleen the type of solution to solve it? What about other LLM challenges? Here are some of the key takeaways from Gleen’s debut.
Gleen Is on To Something
Hallucination is a massive issue for LLMs, and if Gleen can solve this problem, it could translate into real productivity gains for generative AI applications. On paper, the Gleen AI solution is interesting and makes sense as a solution that might be able to mitigate LLM hallucination. The fact that Gleen AI is an abstraction layer, independent of the LLM, makes this solution compelling and enables it to be LLM-agnostic. It is unclear whether navigating that layer with data will mean additional costs for data processing.
Hallucinations Are Not the Only Accuracy-Focused LLM Challenge
Hallucination is only one of several challenges enterprises face in deploying LLMs. To be fair, Gleen is also addressing false confidence and some accuracy issues with Gleen AI. Other issues that Gleen AI might also be able to solve explanability issues; however, Gleen AI probably does not provide the sources to back up its conclusions of the corrected answers. Another issue it might not solve is mitigating bias—LLMs tend to require a heavy dose of pre-production monitoring to weed out bias language and answers.
Cost Issues for Running LLMs
A cottage industry has sprung up to address cost issues for leveraging LLMs. Current compute costs for LLMs can be expensive. As a result, there are massive efforts by a number of chip manufacturers to build more efficient, purpose-built AI chips; a range of development tools have been designed to help AI models for LLMs to run more efficiently; and there are LLMs that are trained on smaller data sets.
Conclusion
Hallucination is not the only challenge enterprises using LLMs face, but it is a significant one worth solving. If Gleen’s concept works, players will scramble to build similar solutions, particularly larger AI development platforms/tools vendors, including the LLM players themselves. Gleen’s focus is on hallucinations or customer service chatbots, but LLMs do not discern in their hallucinations, which means it is likely that savvy players will develop hallucination fighters for all LLM applications.
Disclosure: The Futurum Group is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of The Futurum Group as a whole.
Other insights from The Futurum Group:
Qualcomm-Meta Llama 2 Could Unleash LLM Apps at the Edge
Top Trends in AI This Week: August 25, 2023
OpenAI ChatGPT Enterprise: A Tall Order
Author Information
Mark comes to The Futurum Group from Omdia’s Artificial Intelligence practice, where his focus was on natural language and AI use cases.
Previously, Mark worked as a consultant and analyst providing custom and syndicated qualitative market analysis with an emphasis on mobile technology and identifying trends and opportunities for companies like Syniverse and ABI Research. He has been cited by international media outlets including CNBC, The Wall Street Journal, Bloomberg Businessweek, and CNET. Based in Tampa, Florida, Mark is a veteran market research analyst with 25 years of experience interpreting technology business and holds a Bachelor of Science from the University of Florida.
