At this week’s Snowflake Summit, CEO Frank Slootman opened up his keynote by commenting on the pervasiveness of the letters ‘A’ and ‘I’ in conversation today.

“It's almost like they’re the only two letters left in the alphabet,” he said.

Discussions of the inseparable vowels didn’t let up today, particularly in a panel discussion on generative AI and large language models (LLMs) in the enterprise. The lively back and forth took place between Andrew Ng, founder and CEO of Landing AI, Ali Dalloul, VP of the Azure AI Platform at Microsoft and Jonathan Cohen, VP of applied research at Nvidia.

Moderator Christian Kleinerman, Snowflake SVP of product kicked things off with the provocative question: “Is this gen AI thing real?”

The resounding consensus: Yes.

[ Related:  Nvidia and Snowflake CEOs on partnership  ]

“I think this is going to be completely transformative in how our society functions,” Cohen said.

Ng agreed, saying that: “I think most people still underestimate the magnitude of the transformation that's going to come. There is a huge wave of transformation ahead of us.”

Dalloul, for his part, emphasized that “it is very, very real. There is hype and we need to be very responsible and very grounded. [But] the use cases that emerge out of this are indeed very durable and it is a paradigm shift in the industry.”

From harnessing data to 'catalyzing the creative'

Cohen pointed to the dramatic evolution of computing: Originally, computers spoke 1s and 0s and “so you had a very small number of expert humans who learned how to speak in ones and zeros.”

Then higher-level programming languages like C and Java came along, but those still required a lot of expertise (if not quite as much).

“Now we're entering this era where computers are coming to where humans are, and computers can interact with us in ways that are natural to us,” he said.

Granting “any person on the planet” access to that computational power without them needing expertise will be “transcendent,” he said.

The most notable implication for AI will be its ability to help organizations (finally) successfully leverage data, he said. All organizations have a tremendous amount of data — but much of that is unstructured. AI can turn that data into automated systems that can make decisions or recommendations based on historical business patterns.

“That’s something that's very high value and very achievable,” he said. “There's so much value locked in this data and especially in unstructured data.”

Dalloul agreed that, “there are a lot of very, very, very high value use cases that we are seeing right now in a lot of enterprises that truly are delivering remarkable results.”

For instance, AI models can optimize fraud detection in financial services while “catalyzing the creative” in content creation and bolstering morale and productivity.

Ng, meanwhile, noted that just before the panel started, he had been texting with former Tinder CEO Renate Nyborg on applying AI to relationship coaching. “This is an interesting case study because as an AI person, I don't know anything about romantic relationships,” he quipped.

But AI is allowing the two professionals to come together on “a very interesting vertical, and I think so too will be with your businesses,” he told the audience.

Enabling $500 projects — not just $1B ones

Ng pointed out that previously, AI was considered most valuable for “billion dollar applications,” such as building web search or better product recommendation systems.

But now, he’s seeming more of what he dubbed “$500 projects”: A pizza maker using computer vision to ensure that cheese was evenly spread, for instance, or a farm leveraging the technology to determine when wheat is at optimal height for harvesting.

The “old recipe” of hiring dozens of engineers doesn't work for those smaller projects, he said.

“This is lowering the barrier to entry,” he said. “We now have people saying, ‘You know what, I’m just gonna build a prototype, I'll do it over a weekend.’”

As they evolve, models will also be both horizontal and vertical; that is, more generalized or broad in scope or highly tailored.

AIs will have “different personas and different values,” said Cohen, and organizations will build custom versions based on their own data. That won’t necessarily be from scratch: They can architect them based on large models such as ChatGPT or PaLM 2.

In a year, he predicted, everyone will be using an AI assistant in one way or another to help do their jobs. Also, models will be much more multimodal. For instance, he said, “you can throw a chart at it and ask it to explain the chart and also ask it to predict,‘What is my sales forecasts going to be next year?’”

“It's going to be transformative and incredibly powerful,” he said.

AI starting at the leadership level

Enterprise leaders should look at AI as an “enabler technology” and a tool, Dalloul said, and they should approach it by first identifying their core business values. CEOs should consider “the five things that keep them awake at night,” then work backwards to see where AI could solve those problems and provide the greatest efficiency.

“Therefore, you can justify the economics because at the same time, these things are not cheap,” he said.

Organizations must also adopt AI with consideration for security and ethics and establish principles around transparency and trust, he said. For instance, a media company using generative AI must inform clients that some content was augmented by models.

“You also want to have a social contract with your employees that you're trying to increase their productivity,” Dalloul said.

Responsible AI standards must be set along with governance processes and best practices, as well. Furthermore, it is critical to be prepared for the unknown.

“You want to have a mechanism to get ahead of the curve so that you're not surprised down the road,” he said.

Ultimately, “it has to start at the leadership level and it has to be done in a very grounded way.”

Invest in education, yourself

Ng advised that organizations invest in education (his team, for instance, has released several free and paid trainings on Coursera). “Certainly get your engineers to go learn about this,” he said.

Also, “if you let your engineers experiment, even if their first few projects are not massive successes and learnings, it will add up to bigger and bigger successes over time,” he said.

Dalloul’s advice: Invest in yourself. “Embrace a growth mindset, embrace the change,” he said, and understand the bigger picture scenarios.

Enterprise leaders should have the courage to experiment with these technologies: “Watch videos, attend talks, read books; you know, you don't have to go on to become a Python programmer.”

Also, don't feel guilty that you're falling behind. “It's like ‘Oh yeah, how do I keep up? You don't have to keep up,” he said.

Instead, understand the bigger picture and the foundational elements and develop a thesis and a strategy for your organization, he emphasized.

Cohen, for his part, urged organizations to focus and start small.

“Just pick something you do everyday and see if you can automate it, see if you can make it more efficient,” he said.