While most of you will be familiar with ChatGPTwhich is a creative artificial intelligence (AI) is built on a large language model (LLM) that provides relatively intelligent answers to questions, few of you will use it at work. ChatGPT is not generally considered safe for serious business effort and mostly used for tinkering at this point.
Now, efforts are underway to package language models into an enterprise environment, with a focus on resident enterprise data. But at the same time, the AI practitioners and Experts are calling for caution with the development of AI and LLM.
These are the findings from a survey among 300 AI experts and partitions released by Expert.ai. “Enterprise-specific language models are the future,” the report’s authors state. “Business and engineering executives are being asked by their boards and increasingly by shareholders how they plan to capitalize on this new dawn of AI and the promise it offers. to unlock the language to solve the problem.”
Research shows that more than a third (37%) of businesses are considering building business-specific language models.
At the same time, AI practitioners realize that building and maintaining a language model is no small task. The majority of enterprises (79%) found that effort required to train a business-specific language model that is usable and exactly “one main business”.
However, efforts are underway — the teams have budgeted for LLM training and adoption projects, with 17% having a budget this year, another 18% planning to budget and 40 % discuss budget for next year.
“This makes sense, as most of the public domain data used to train LLMs like ChatGPT is not enterprise-grade or domain-specific data,” said the expert.ai author. “Even if a language model has been trained on different domains, it is unlikely to represent what is used in most complex enterprise use cases, whether vertical domains such as financial services, insurance, life sciences and healthcare, or very specific use cases such as contract reviews, medical claims, risk assessment, fraud detection and review network policy. Training effort will be required for consistent quality and performance in specific domain name use cases.”
For enterprise AI advocates in the survey, The top concern with general AI is security, cited by 73%. Dishonesty was another issue, cited by 70%. More than half (59%) expressed concerns about intellectual property rights and copyright protection — especially with LLMs like GPT, “trained in multiple streams of information, some of which are protected copyright and because this information comes from publicly available internet data”. the authors of the report maintain. “It has a fundamental problem of garbage in, garbage out.”
AI can reduce the need for human resources in specific tasks, but ironically, it will require even more people to build and maintain it. More than four-tenths (41%) of AI advocates expressed concern about the shortage of skilled professionals with the expertise to develop and deploy innovative AI for businesses.
More than a third (38%) of survey respondents expressed concern about the amount of computing resources required to run an LLM. The report’s authors say infrastructure, such as powerful servers or cloud services, is needed to support the large-scale deployment of language models.
Enterprises adopting language models require careful planning and consideration of a range of factors, including privacy and data securityinfrastructure and resource requirements, integration with existing systems, ethical and legal considerationsand knowledge and skills gaps.
As with any emerging technology, successful adoption depends on use cases that represent a significant leap over previous approaches. There are a few sure use cases for general AIas discovered in the survey:
- Human-computer interaction: Enterprise language models will serve to provide end-users and customers “the ability to quickly and easily access information and support, such as product details, troubleshooting instructions, etc. troubleshooting and frequently asked questions.” The most common use cases at this stage are chatbots (54%), Q&A (53%) and customer care (23%).
- language generation: “Creative AI can write new content, create immersive visuals, create marketing content, compose music, and even generate programming code.” The two most popular examples at the moment are content summary (51%) and content creation (45%).
- Information extraction: The top use cases here are knowledge mining (49%), content classification and metadata generation (38%). Content classification for routing (27%) and entity extraction (20%) are also mentioned.
- Search: General search (39%), semantic search (31%) and recommendation (29%) are considered “important tools to help people find the information they need quickly and accurately without need to go through a lot of unrelated results.”
While many businesses may be looking to adopt an enterprise LLM, most AI advocates in the survey advise caution when going ahead with AI. Nearly three-quarters (71%) agree that government regulations are needed immediately to address legitimate commercial and malicious use of AI. The report’s authors warn that AI and LLM “can have significant ethical and legal implications, particularly around issues of bias, fairness and honesty”.