Generative AI Knowledge: The More you Know

demo

Many of us grew up watching those ubiquitous (and often parodied) public service announcements called “The More You Know.” With famous figures speaking earnestly to the viewer on topics like public health, safety, the environment, etc., these bite-sized PSAs work on the straightforward premise that the more you know about a tricky and important subject, the better prepared you will be to handle it when it touches your life. Today, industrial CEOs, business owners, government leaders, and everyday citizens are experiencing that old familiar “the more you know” feeling regarding generative AI knowledge. They are on a fact-finding mission to understand what this technology means for their business. What are the risks? What are the rewards? And what exactly is it again?

Broad public awareness of generative AI’s capabilities skyrocketed less than a year ago when OpenAI launched ChatGPT, DALL·E 2, and Whisper. Since then, volumes of explainers and editorials have attempted to find the middle ground between hype and reality, fact and fiction. Some hot takes offer fear that generative AI’s resemblance to artificial general intelligence—AI virtually indistinguishable from human intelligence—signals the rise of machines someday taking over the world. For the record, we don’t see it quite that way. 

Even if generative AI is the stairway to AGI (and time will only tell), we’ve argued (and proven) over the last ten years that AI is just a tool, albeit a unique and exponentially powerful one. Used appropriately and precisely, AI technology in general and perhaps generative AI, in particular, can help us discover critical insights to improve human quality of life and solve entrenched problems like climate change, school safety, early disease diagnosis, etc. For this reason, generative AI knowledge gathering should be explored to its utmost. 

The more you know, the more you’ll be able to leverage generative AI successfully

We recommend two recent resources to add context to your personal generative AI knowledge. The annual McKinsey Global Survey on AI brings timely perspective and analysis on how organizations have reacted during the early days of generative AI for business. The State of AI in 2023: Generative AI’s Breakout Year provides fascinating findings from its survey respondents, showing that generative AI is already attracting plenty of early adopters, for example:

  • 33% of respondents reported using generative AI tools in at least one business function.
  • 60% of organizations that reported some form of AI adoption use generative AI, with 40% expecting further investment.
  • 75% of respondents predict generative AI will disrupt their industry within three years.
  • AI high performers—respondents who attribute 20% of EBIT to AI—likewise use generative AI the most, including in product and service development, risk modeling, and HR. They are more focused on increasing the overall value of their current offerings through AI-enabled features versus reducing business costs through AI strategies.

 

On this last bullet point, McKinsey’s Bryce Hall adds adept commentary:

“Over the past six years as we’ve conducted our annual global AI research, one consistent finding is that high performers take a broad view of what’s needed to be successful. They are particularly strong in staying focused on value, and then rewiring their organization to capture that value. This pattern is clear when looking at how high performers are working with generative AI as well.

For example, on strategy, leaders from our analysis are mapping out where the high-value opportunities are from AI across their business domains. Tellingly, they’re not doing this for only generative AI. As excited as we all are about the dazzling new gen AI applications, significantly more than half of the potential value for companies comes from AI applications that don’t use gen AI. They are maintaining discipline in viewing the full range of AI opportunities based on potential value.

The second generative AI knowledge resource you should bookmark is our recently published webinar, Unleashing the Power of Generative AI: Revolutionize your Industrial Operations. This hour-long webinar featuring SparkCognition’s Chief Science Officer, Dr. Bruce Porter, and moderated by Stephen Gold, Chief Marketing Officer, covers the fundamentals of generative AI, including using tailored large language models (LLM) and industry-tuned solutions. You will learn key considerations for applying and deploying generative AI technology to augment human skills, capitalize on data insights, and accelerate business processes. 

Dr. Porter gives essential background to learn how generative AI works, how it differs from other AI categories, how large language models are trained and fine-tuned, how users can master prompt engineering using context and constraints, and more. A perfect example of the insights you’ll find in this webinar is Dr. Porter’s advice for how companies should approach getting started with generative AI:

“We’re at an early stage on assimilating this technology. So what would be a good way to put your toe in the pond? Here’s my suggestion. First, focus on use cases. Don’t focus on the technology itself. Focus on use cases within your organization. Explore them to see whether they have legs. [Determine that] they are useful to you before you drill down on their technical feasibility. 

So, how do you do that? How do you explore use cases? I would start with publicly available LLMs. So instead of standing up something local in your organization, start with what’s publicly available. Start with the GPT family or offerings from Microsoft, Amazon, or Google. Use these to try out a use case in your organization. Commonly, the use cases involve giving supplemental documents from your organization to the LLM. Have it process these supplemental documents. Maybe they’re training manuals or technical manuals? Whatever you operate. 

Now since you are working here with publicly available LLMs, you may have some concern about data leakage to the cloud. So, mitigate that problem by using proxies for the actual documents that are proprietary—[use] last year’s technical manual or one that comes from open source. It doesn’t matter the content so much as the type of information that you’re putting into the LLM. And drive the use case! Experiment with prompt engineering. We’ll talk a little bit more about that. And if the ticket or use case is promising, then continue on to the next step of this workflow. 

In that next step, look at your concern about data leakage. If that’s a big issue to you, then move to a private LLM. That involves some technical work on your side. Because of that technical hurdle, I suggest you not start with a private one—start with a publicly available one. 

Step three. Fine-tune the LLM to use your data and to fine-tune it around your use cases. And finally—I think this is where the real payoff is—integrate that LLM into your workflows and your software applications. So it’s not just a chat system. It’s rather a proponent of an otherwise comprehensive software application. 

SparkCognition’s Unleashing the Power of Generative AI: Revolutionize your Industrial Operations webinar and McKinsey’s State of AI in 2023: Generative AI’s Breakout Year report are excellent companion pieces to elevate your generative AI knowledge, providing a much-needed level-set and multilayered perspective on how generative AI is being leveraged today across industries. 

As they say: the more you know.

Latest blogs

SparkCognition is committed to compliance with applicable privacy laws, including GDPR, and we provide related assurances in our contractual commitments. Click here to review our Cookie & Privacy Policy.