ISM World 2024: AI and LLMs Advancing at a Record Pace

A new skill set is emerging: prompt engineering. It’s the practice of crafting and making prompts to ensure that the large language model (LLM) you are using — like Google’s Gemini artificial intelligence (AI) — gives you the answer you want, said Patrick Marlow, staff engineer, Vertex Applied AI Incubator team at Google.

Marlow spoke to members of Institute for Supply Management^® (ISM^®)’s Committees and Councils on Sunday at the ISM World 2024 Annual Conference at the MGM Grand in Las Vegas. More than 1,500 supply management professionals are expected to attend the Conference, which runs through Wednesday and includes keynote presentations, workshops, educational sessions and the ISM Awards Gala.

Marlow said that LLMs and prompt engineering will make job roles like software engineering and data analytics more accessible to everyone. That’s because coding knowledge is not a prerequisite, he said. The new coding language is English — or Spanish, Italian or the user’s own language, he said.

Prompt engineering is still nascent, but it and generative AI approaches and tools are developing quickly. Having real-time information, rather than relying on only historical information, is critical for many organizations, and an approach called retrieval augmented generation (RAG) can be effective, Marlow said.

According to Google, it is “a design (or architecture) approach for generative AI that combines information retrieval techniques with generative language models (LLMs) to enhance the quality and relevance of generated text.”

“It’s like a Google search for your information,” Marlow said.

He cautions that large language models can be hacked, so companies need to establish a security posture. Also, understand that there are limitations: Models may not have an answer for every prompt, and instead may give the next probable answer, which could be incorrect. These are called hallucinations.

Key Takeaways

• LLMs aren’t a one-size-fits-all application: Prompt creation is different for each.
• Eventually, all LLMs will advance to the same levels or complexity, speed and the like. Which LLM to choose will eventually become like choosing between an Android phone or an Apple iPhone.
• No experience is necessary to use LLMs.

Quote

“In the last 18 months, generative AI has happened. And it’s advanced. It’s unheard of.” — Marlow

(Image credit: Getty Images/Da-Kuk)