What’s Next for Generative AI? Everything We Overheard at ‘VentureBeat Transform’

Technologists, entrepreneurs and investors descended on the VentureBeat Transform conference in San Francisco last week for what was billed as “the first independent conference on generative AI.” A few of us from Mission North’s Enterprise Practice Group attended to hear how the tech industry is navigating the hype, opportunities and risks of this fast-growing technology. The overall tone of the conference was one of excitement and optimism, tempered with a healthy awareness of the need to deploy this technology intentionally and responsibly. 

‍

Here’s what we learned:

‍

The Biggest Enterprise Opportunities

‍

Gerrit Kazmaier, Vice President of Data and Analytics at Google, and Matt Wood, Vice President of Product at AWS, spoke at length about how to leverage generative AI for enterprise success. They identified two major areas where customers are already seeing benefits:

‍

• Software development: This is one of the most constrained enterprise resources. Generative AI is making coders more efficient by generating source code quickly and accurately, and can even enable people to write code who normally don’t.
• Customer experience: Generative AI is enabling new types of personalization and search, where an end-user may not even realize they’re using a large language model (LLM) but it’s working in the background.

‍

Crucially, generative AI has the opportunity to even the playing field for organizations of all sizes. “Big companies have been able to use machine learning to better serve customers for decades,” said Wood. “Gen AI makes this technology much more available. I think we’ll see hundreds of organizations emerge that are going to be pretty transformative and more disruptive.”

‍

‍

How to Navigate the Risks

‍

The ethical and legal implications of this technology were another focus throughout the conference, with much discussion devoted to the potential for disinformation, privacy and copyright violations, bias and security vulnerabilities. “Generative AI should not be deployed thoughtlessly,” said Ravi Jain, Chair of the Association for Computing Machinery (ACM) working group on generative AI. “We have to make sure as we’re developing this technology that the benefits are commensurate with the risks.”

‍

Jain spoke to the need to look holistically at AI models to mitigate unintended outcomes. “Look at the implications of the technology. Have we introduced bias into the data? Have we made it inaccessible to certain populations?”  

‍

He recommended that organizations develop guidelines around transparency and limits of use, as well as a risk and benefits framework in order to develop and deploy AI responsibly. For example, an end-user should always be informed when they’re talking to an AI and not a real person.

<split-lines>"We have to make sure as we’re developing this technology that the benefits are commensurate with the risks." - Ravi Jain, Chair, Association for Computing Machinery - working group on GenAI<split-lines>

The Right Balance Between Humans and AI

‍

There was wide agreement among the panelists around the opportunity for AI to accelerate and optimize the work that humans do, rather than replacing it. “When deployed correctly, AI can give employees super powers,” said Uljan Sharka, CEO of financial education company iGenius. She spoke about the importance of human experts supervising not only AI tools but the data that feeds into them. 

‍

Gail Muldoon, Head of Customer Data and Analytics at Stellantis, echoed the importance of human oversight to mitigate security and ethical concerns. “AI is not the be-all end-all. You still need a human to interpret its outputs and make business decisions. There needs to be a human strategist behind the scenes.”

‍

“Humans are essential,” said Usama Fayyad, Executive Director of the Institute for Experiential AI at Northeastern University. “The organizations with the biggest advantage will be those who have well-trained people using AI who are experts in their domain. That’s the hidden secret. The technology won’t work well without that continuous human intervention.”

<split-lines>"The organizations with the biggest advantage will be those who have well-trained people using AI who are experts in their domain..." - Usama Fayyad, Executive Director, Institute for Experiential AI, Northeastern University<split-lines>

Tips for Successful AI Deployment

‍

AI implementation presents a number of roadblocks. Several panelists spoke to the need for intentional, patient deployment of AI to ensure organizations can reap the full benefits of this technology.

‍

“Avoid shiny objects syndrome,” advised Mark Tack, CMO of TreasureData. “Everyone is excited about AI. Many organizations are moving quickly. You need to slow down in order to speed up.”

‍

AI is an accelerator, Tack continued, but organizations must pause and answer the question of what they’re accelerating and why. Make sure the right strategy, processes and tools are in place. Without those foundational elements, companies risk veering in the wrong direction.

<split-lines>"AI is an accelerator...but organizations must pause and answer the question of what they’re accelerating and why." - Mark Tack, CMO, TreasureData<split-lines>

Smaller LLMs Are the Future

‍

The most successful deployment of generative AI will come from smaller LLMs that are trained on a company’s internal data, said Kjell Carlsson, Head of Data Science Strategy at Domino Data Lab. General-purpose LLMs trained on wide-ranging internet data lack the specialized knowledge that companies need. Smaller LLMs, Carlsson noted, will be less expensive to train and minimize the risks of exposing sensitive company data to the outside world.

‍

Cohere Co-Founder Nick Frosst agreed. He expects generative AI to become just another tool we don’t think about or ever realize we’re using, embedded in every website and app and as ubiquitous as a touch screen. Every company will have multiple specialized LLMs to address internal employee questions, he said.