The AI Show - May - Enabling Enterprise LLMs and GenAI.

Amir from PrompTech: Focused on the challenges of data privacy and talent gap in LLM adoption. He introduced PrompTech's "AITeam Space" platform as a solution for secure and easy LLM implementation with features like no-code prompt creation and access to various LLMs including secure options.

Matt Weston from Lexical: Highlighted security concerns as a barrier to AI adoption and showcased Lexical as an enterprise-grade platform for secure LLM access. He demonstrated real-world applications of Lexical in improving communication, salesprocesses, and data analysis. He also provided guidance on choosing the right LLM provider and implementing AI solutions.

Steve Naude from Smartspace: Addressed misconceptions about generative AI, emphasising its limitations and potential riskslike lack of inherent understanding, data governance issues, and vulnerability to attacks. He stressed the importance of responsible AI practices and introduced Smartspace as a platform that facilitates secure and responsible generative AI implementation with features like data spaces, model agnosticism, and comprehensive risk mitigation strategies.

---

Our first talk from Matt Weston, Founder of Lexical focuses on the challenges and solutions for adopting AI within businesses, specifically focusing on secure implementation of large language models (LLMs).

Key points:

• Security Concerns: Many companies avoid AI due to data security and privacy risks, leading to "shadow IT" where employeesuse unauthorized AI tools. Lexical addresses these concerns by providing an enterprise-grade platform with secure access toLLMs.

• Real-world Applications: Matt showcases three examples of how Lexical has been used:

  • Improving Communication: Ensuring consistent and clear email communication, particularly for multilingual teams.

  • Sales Process Enhancement: Automating content creation and customization in documents like proposals and contracts.

  • Data Analysis: Enabling teams to analyze data using natural language and generating insights without needing codingexpertise.

• Model Selection: Matt advises on choosing the right LLM provider, considering factors like data security, pricing models, and feature availability. He recommends sticking with existing cloud vendors like Azure or AWS for simplicity and security.

• Implementation Tips:

  • Start with a small, well-defined project with clear goals.

  • Focus on internal applications initially to mitigate risk and refine the process.

  • Work with experts to identify opportunities, implement solutions, and ensure ongoing optimization.

Overall, Matt’s talk emphasises the importance of secure and well-planned AI adoption in businesses, highlightingLexical as a valuable tool to achieve this goal.

This talk by Amir, founder of PrompTech, focuses on the challenges and opportunities of using large language models (LLMs) within enterprises, specifically in the context of New Zealand.

Key Points:

• Privacy Concerns: The biggest barrier to LLM adoption is data privacy. Research indicates significant concern amongorganizations regarding data security when using these models.

• Talent Gap: There is a shortage of skilled professionals who understand LLMs and can effectively implement them withinorganizations.

• Understanding LLMs: The talk delves into the mechanics of LLMs, explaining how they predict words and generate text based on prompts and tokens.

• Prompt Engineering: The importance of crafting effective prompts to get the desired output from LLMs is emphasised. Theprocess of designing, debugging, and optimizing prompts is discussed.

• Prompt Tech Solutions: Amir introduces Prompt Tech's "AI Team Space" platform, which allows no-code developers to createand share prompts as a service. The platform provides access to various LLMs, including secure and private options like open-source LLaMA and Azure OpenAI, addressing privacy concerns.

• Call to Action: The talk concludes with an invitation to learn more about LLMs through a Udemy course and to explore PromptTech's offerings, especially during the upcoming M2 AI Summit.

Overall, the talk aims to educate businesses about LLMs, address privacy concerns, and showcase PrompTech's solutions for effective and secure LLM implementation within enterprises.

Our final presentation of the evening was from Steve Naude, Chief AI Officer at Smartspace. Steve delves into the critical aspects of implementing generative AI within an enterprise environment. It tackles key misconceptions about generative AI, explores potential risks, and emphasises the importance of responsible AI practices.

Key Points:

• Misconceptions about Gen AI: The talk debunks common misconceptions, emphasising that Gen AI models are not original, creative, conscious, or self-learning. They lack inherent understanding, accuracy, and are prone to biases present in their trainingdata.

• Risks of Gen AI: Several risks are highlighted, including:

  • No concept of truth: Outputs can be misleading due to reliance on training data and context.

  • Data governance: Issues related to copyright, data privacy, and security require careful consideration.

  • Responsible AI challenges: Overly restrictive AI policies can lead to unintended consequences, as seen in the GoogleGemini example.

  • Regional laws: Adherence to specific regulations in different regions is crucial.

  • Prompt injection attacks: Vulnerability to attacks that manipulate model outputs through malicious prompts.

• Data weighting: The importance of providing sufficient context through prompt data to ensure accurate and relevantresponses is emphasized.

• Open book testing analogy: Gen AI models should be treated like students taking an open-book test, requiring ample contextfor optimal performance.

• SmartSpace approach: The talk introduces Smartspace, a platform designed to facilitate secure and responsible Gen AI implementation within organisations. It emphasises features like data spaces, workspace automation, model agnosticism, and a comprehensive approach to addressing the risks associated with Gen AI.

Overall, the talk urges organisations to carefully consider the challenges and risks associated with Gen AI before implementation, and to adopt responsible AI practices and tools to ensure a successful and ethical integration of this technology.