This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.
2023 in Review: Enterprises Focusing on How Data Was Used in Gen AI
As 2023 rolled along, and the generative AI announcements began to accumulate, the wind picked up and drove vendors toward the necessary discussion about the data on which the models were trained, how the models were trained, what biases may have been introduced, and how those inputs (among others) might affect the outputs of those LLMs – toxicity, bias, hallucinations, factually wrong responses, etc.
According to Omdia’s AI and intelligent automation research director Natalia Modjeska, one major challenge companies face is data. “Specifically, high-quality, scalable, reliable, and trusted enterprise data to fine-tune and ground foundational models to make them usable. We know that many organizations still struggle with that because of the chronic underinvestment in data quality, management, and governance.
“I like to say that one can build a beautiful generative AI house and maybe even host a house-warming party, but you won’t be able to live there unless the house has a proper foundation and unless utilities such as water and gas mains, telecommunication, etc., have been laid in the ground. Data is that foundational infrastructure for AI.”
Related to that, enterprises have grown increasingly wary of how their data might be exposed on the Internet and/or how their data might be used, without their consent or perhaps knowledge, to train the LLMs. According to a recent article by Metrigy analyst and frequent NJ contributor Irwin Lazar, “IT and business leaders continue to express concern over the training of models and how company data is protected. We expect vendors to continue to message (and differentiate) around their security, governance, and compliance capabilities and partnerships.”
Focus on AI and Data in No Jitter Roll
Over the course of 2023, responsible AI, usage policies, protecting personally identifiable information (PII), managing data, etc., all became part and parcel of the discussion around generative AI and how it uses data. A few examples include:
- Salesforce’s launch of the Einstein 1 Platform, a suite of AI-powered tools offering analytics, automation, and customer data capture at scale. Per an announcement at AWS re:Invent 2023, Amazon Bedrock will be available through the Einstein Trust Layer.
- In mid-December 2023,Salefsorce updated its Einstein 1 Platform, as well.
- Amazon’s launch of its Bedrock managed service and then updated that service in late November 2023, with new models and security approaches.
- Emplifi announced a unified analytics component to provide a single view of business and customer data.
- Mindbreeze’s InSpire solution which emphasized the security and integrity of LLMs for enterprise use.
- Verint launched a bot that automatically redacts personally identifiable information (PII).
- Kyndryl and Microsoft partnered to drive generative AI adoption.
- Snowflake's new Cortex managed service which provides customers with access to LLMs and AI models.
- Miro added more than 50 collaboration features along with tighter data security and privacy,
- ai brought enterprise-level privacy and security to generative AI applications.
- Thales and Google Cloud partnered to improve data security via machine learning and AI.
- VMware and NVIDIA extended their partnership to a generative AI software/hardware platform for enterprises,
- The Partnership on AI invited public comment on its framework for safe AI model deployment.
For Further Reading
Much of NJ’s Conversations series dealt with asking many of the companies directly involved in the incorporation and implementation of AI and generative AI-based solutions about how the capabilities of those models could be used and controlled. A few key points included:
- The need for data security, guardrails, etc.
- Asking questions, at least, about the data used to train the models.
- Keeping “humans in the loop” to evaluate what the generative AI “engine” produces as a backstop against toxicity, hallucination, incorrect information, etc.
Also look at these articles regarding the use of data by AI:
- CX software that leverages an open data platform at its core may better to orchestrate customer experiences across the enterprise.
- When it comes to data integration for CX many companies are in fact getting quite a bit of the data they need.
- Generative AI is only as good as the data it draws on.
- As organizations evaluate AI they should consider some time-tested principles of technology adoption.
- AI and the role of knowledge management in contact centers is becoming more and more prominent.
- Zoom’s AI Approach: Leading with Openness and Accountability.
- Generative AI is already embedded in contact centers, but trust concerns of the technology are widespread.