Generative AI in the enterprise space
ChatGPT, Bard... Key players in generative AI are moving swiftly to secure a position in the enterprise AI space.
Recently, ChatGPT unveiled its enterprise version1, laying the groundwork for the race to begin.
However, what are the real challenges faced by generative AI in the enterprise space?
Gartner2 defines these risks as follows:
* Domain adaptation
* Copyright issues
* Concentration of power
* Hallucination
* Potential for misuse
* Opacity
These are significant challenges that require a collaborative effort between AI companies and those implementing the AI. This collaboration should address both the architectural and usage challenges from various perspectives.
It's logical to seek a collaboration when we consider the impact generative AI could potentially have. McKinsey3 estimates that generative AI could further reduce the volume of human-serviced contacts by up to 50 percent and increase sales productivity by 3 to 5 percent of global sales expenditures.
AI companies should carefully analyse and address copyright issues and opacity, as they hinge on the formulas and base data used to train the AI. It is quite possible that specific legislation might be needed to clarify what can and can't be used in the training of these large foundational models.
Two areas that must be considered by platforms implementing such technologies in the enterprise space are 'Hallucination' and 'Potential Misuse'. These concerns extend to the control of message content and brand safety.
How are implementers or platforms mitigating these risks? José Luis Cantero, CEO of WorkbAI, explained that they've found a solution by creating a layer on top of the Generative AI. This layer is able to understand the context and fine-tune the prompts in real time. He said, “The capabilities of generative solutions are massive, but it's like driving a car without brakes. Companies like WorkbAI are focusing on creating this 'conscience' layer on top of the Generative AI brain. This allows the AI to understand the objective and keep the conversation within safe boundaries."
He added, "Another big challenge is delivering the proper legal disclosures during a conversation with an AI-powered bot. This isn't something that can be guaranteed with generative AI, and guaranteeing is the key for enterprises. That’s why our 'conscience' layer takes on greater significance. It has the ability to deliver a specific message in predefined situations to meet legal requirements.”
In conclusion, the solution lies in collaboration. Tech giants like Google, Microsoft, AWS, and OpenAI need to enhance their core technology, while the growing ecosystem of solutions based on these technologies must develop a containment network to guide the AI within safe boundaries.
1. https://openai.com/enterprise
2. Understanding ChatGPT, LLMs, Foundational Models and Generative AI (Gartner)
3. June 2023 The economic potential of generative AI (McKinsey&Company)