The News: On July 25, large language model (LLM) startup Cohere announced it has introduced Coral, described as a “knowledge assistant” in private preview with select customers. Coral, like Google’s Bard and Anthropic’s Claude are alternatives to OpenAI’s ChatGPT.
In the announcement, Cohere is intentional about how Coral will stand out from “existing consumer chatbots” (assume that means ChatGPT):
- Coral is powered by Cohere’s Command model, which is trained with chat, reasoning, and writing capabilities.
- Customized: According to the post, “Customers can augment Coral’s knowledge base through data connections. Coral has 100+ integrations ready to connect to data sources important to your business across CRMs, collaboration tools, databases, search engines, support systems, and more.”
- Addresses hallucinations with explainability: According to the post, “To help verify generations, Coral can produce responses with citations from relevant data sources. Behind the scenes, our models are trained to seek relevant data based on a user’s need, even from multiple sources. This grounding mechanism is essential in a world where workers need to understand where information is coming from in a consumable way.”
- Secured data in private environment: Data used for prompting and the chatbot’s outputs will not leave the company’s data perimeter.
Read the full announcement on the Cohere website.
Cohere Launches Coral, a New AI-Powered Knowledge Assistant
Analyst Take: Open AI introduced ChatGPT a mere 9 months ago. In that short amount of time, the concept of an AI assistant has evolved and mutated rapidly to a point where the challenges for LLM-based AI assistants are being addressed and enterprise requirements are part of their design. What does Coral’s debut signal about what happens next with AI assistants? Are we on the cusp of next-generation computing? Here are the key takeaways:
Evolution of Consumer Assistant to Enterprise-Grade Assistant
AI assistants are getting better. As ChatGPT has been adopted, its potential has been quickly tempered by market challenges that would prevent enterprises from adopting it – hallucinations and other inaccuracies, bias, security, explainability, assistants’ lack of sufficient memory, and a need for specific methodologies to obtain the best results, like prompt engineering. Evolved LLM-based AI assistants and organizations that use them are layering in ways to address these issues. For example, Coral’s approach to hallucinations is to provide citations, some enterprises limit other informational inaccuracies, bias, and security by pointing the models not at public domain data, but rather strictly at private data (as exemplified by Salesforce and Adobe). Coral’s approach to security is to access public domain data for the assistant, but to limit where and when Coral uses private domain data. Anthropic and Meta’s Llama 2 have both increased the length of inputs and outputs users can use in each prompt, which in theory increases the assistants’ memory to improve their performance.
The Next-Generation Interface
Coral and the evolving enterprise AI assistants are bringing the world closer and closer to the next-generation interface, one where most software can be operated by telling it what to do in a user’s own words and even link operations across applications.
Since 2015, AI visionaries have dreamed of a day when natural language processing (NLP)-based AI would usher in a new era in computing – when a conversational interface would replace today’s text/mouse/point-and-click interfaces. In this vision, users tell (or type) what they want of software in their own words, and the interface understands what they want and executes operations, literally any function computer software performs today. The vision included this interface being able to perform complex tasks humans typically do in workflows – such as working across disparate applications to complete an operation.
Challenges Remain: How Good Will AI Assistants Be?
While AI assistants are making massive improvements in their capabilities, they are still more about potential and promise than effective, reliable tools. The barriers and challenges mentioned earlier loom large and solving those challenges is not a sure thing at this point. Further, there is the ongoing challenge of natural language understanding (NLU). AI systems have struggled to understand humans effectively. Human communication is one of the most complex, if not the most complex task that exists – so much of how humans communicate – tone, conversational history, reference linking, sarcasm, spatial elements, volume, inflection – are impossible for machines to translate. Generative AI does not necessarily solve the NLU problem, and many experts believe AI never will be able to duplicate human ability in NLU. The effectiveness of AI assistants will depend greatly on their NLU capabilities.
Competition
Enterprise-grade AI assistants will proliferate over the next 2 to 3 years. Competition will be fierce, because winners in this space will have opportunities to sell other services and applications into the enterprise market. Some will be designed as general-purpose assistants, some will be very narrowly focused, some will come from open source, while others will be based on closed systems. Vendors in this space will come from AI compute players, data management and governance players, enterprise software players, independent LLM players, and possibly other sectors we have not yet imagined.
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
Generative AI Investment Accelerating: $1.3 Billion for LLM Inflection
Oracle Launches Enterprise-Grade Oracle Generative AI Services with Cohere
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.
