ASAI 2024 Blog Featured Image

Reflecting on the American Society for AI June 2024 Summit

I recently had the chance to meet with the American Society for AI, which held its annual meeting in June in New York City. It was a pleasure to have the opportunity to interact with people using AI in various ways, ranging from business applications, government, defense, all the way to creative artists who leverage AI tools to generate viral content.

This group brings AI experts together for off-the-record, honest discussions about how AI can be leveraged to make the world a better place across uses in policy and government, business, education, and more. By our charter, the meetings operate by Chatham House Rules: I can’t disclose the exact conversations we had, but I can highlight some of the key topics and learnings I took away.

AI is a major disruptor

The 2024 version of McKinsey’s State of AI report was recently released and contained a handful of themes that came up repeatedly in discussions. First and foremost — AI is a major disruptor, and as organizations, governments, and individuals begin to adopt the use of tools like Large Language Models (LLMs) to automate aspects of their lives, they will begin to grow increasingly dependent on these technologies.

McKinsey found that in 2023, only 33% of organizations were leveraging generative AI for at least one business function — this has nearly doubled in the span of a year to 65%. Predictive analytics, too, has seen a jump from a multi-year plateau. In 2024, 72% of organizations are using some type of predictive model. It has been said that 2023 was the year of AI Consideration, and 2024 is the year of AI Adoption — this appears to be true.

AI policy and regulation will be a major consideration

Governance around AI is also a critical consideration, affecting people and organizations at every level. Much has been made about the risks associated with using AI tools, particularly LLMs, which require prompts to be used. We talked about these topics at length, particularly focusing on impacts that AI will have on how businesses operate, how governments will need to legislate, and how AI will drive the balance of power in the future.

Many companies are concerned with building frameworks that preserve data security while still hoping to use these tools to achieve productivity gains. Development of clear, cohesive policies around how employees are expected to use these tools will be paramount to success.

It is key to remember that data governance in support of predictive analytics is somewhat different from governance needed to support generative AI. In many cases, predictive models will be trained using internal company data, and processes need to be optimized to maximize the quality and consistency of that data. For generative AI use cases, most organizations will leverage pre-existing models, so the goals need to be oriented around building safeguards and guardrails that drive consistent, low-risk usage of these tools.

At a national level, AI policy may be one of the largest — if not the largest — factors that set the stage for the structure of power globally. Creating government regulation and regulatory frameworks that spur innovation in this space while also encouraging responsible, ethical development of AI tools is critical, but much easier said than done. We are going to need structured collaboration between legislators in Washington and AI leaders in private industry to maintain regulations that can keep up with the breakneck speed of innovation in this sector.

Finally, hardware considerations are becoming an important factor. As computational needs rise, the need for chips that can handle both the training and the inference (i.e., the actual answering of prompts/generation of predictions) is becoming important on a global scale, evidenced by the $285 million dollar investment the Biden administration is making into domestic chip/semiconductor manufacturing here in the USA.

Much of the innovation in the LLM research world is also focused on making models more efficient, so, ideally, as the models get better and the chips get more powerful, we’ll be able to do even more with these tools in the future.

Every company is an AI company — and every employee needs to be AI-enabled

It’s no secret that the development of an AI-driven workforce will represent one of the most impactful paradigm shifts since the development of the Internet (and indeed, some experts believe this to be an even bigger deal than that!). Its power to democratize information, provide access to education, drive automation, and lead to more efficient outcomes is going to be critical.

The impact on employees is important to consider, both in the short-term and long-term. Over the next few years, we will see AI tools become integrated into nearly every aspect of the technology we use, from iPhones to CRM to copilots that help generate code. Many of these tools are already being used in the enterprise, but adoption will continue to increase, and the breadth of usage will explode.

Companies and individuals who are not AI literate will find themselves facing a massive competitive disadvantage. Focusing on developing business processes that align with the usage of AI for automation will help people to be more efficient in their roles.

I like to put it this way: Let humans do what humans do best, like creating trust and building relationships, and design processes to let AI automate many of the manual or computational tasks that are part of their workday.

AI maturity: Going from predictive to prescriptive

Much of the usage of LLMs and AI/ML models today fits into the category of “predictive” modeling in the sense that many models drive insights, not actions. This is going to change.

LLMs will someday be displaced by Large Action Models (LAMs) that do both the generation/summarization of content AND the automation of an optimal next step. There’s some complexity behind this concept, but basically, LAMs represent the intersection of pattern recognition (which existing models can already do quite well) with decision optimization (which is an active area of research).

My concluding thoughts

It was a pleasure to get the chance to interface with other industry experts at ASFAI. The breadth of expertise that this group has underscores the multifaceted nature of AI and how it will affect different aspects of our lives in the future.

There is so much happening so quickly in the world of AI development that it’s difficult to really understand where we stand. Certainly everyone can agree that we’re in the midst of a massive hype cycle. Some marketing may greatly outpace real functionality, but in some cases, both out-of-the-box tools and custom frameworks offer an opportunity for organizations to significantly drive up their productivity.

Interested in learning more about Atrium’s Data Science and AI practice? Visit our data science consulting page.