januari 05, 2024
David Giard, Cloud Solution Architect at Microsoft and Cameron Turner sat down to discuss predictive and generative artificial intelligence (GenAI). From the fundamentals of these technologies to the ethical implications and the critical role of human influence, the conversation shed light on the nuances and potential of AI for organizations. Here are some of the highlights of this thought-provoking discussion.
Generating value from data
We are digital transformation consultants. Turner explains that he helps clients understand what solutions are out there, mapping opportunities to those solutions, and — crucially — building them. Data and AI are at the center of his role and he is excited about the opportunities organizations have to generate value from existing data.
Organizations have already collected and stored data. And they're wondering what to do with it. That's where the fun begins. We model against that data asking: Is it predictive of what we want to look at? Is it giving us answers that we can trust? Now, with GenAI we have a whole new set of tooling that we can apply to those kinds of questions.
Cameron Turner, who became VP of Data Science at Kin + Carta in 2020.
Valtech acquired Kin + Carta in April 2024.
The blend of predictive AI and generative AI
Turner describes how we’ve “graduated over the years from systems that were purely analytics” to machine learning and AI. AI is an umbrella concept that includes natural language processing (NLP), supervised learning, unsupervised learning, and large language models (LLMs) that work with vast datasets — including semistructured data.
While GenAI focuses on creating new content, predictive AI forecasts future outcomes by analyzing historical data. Though the math and source data may be different for each, these two AI approaches are interconnected; we can now build GenAI systems that create recommendations, which in turn become new content in the generative cycle.
Generative and predictive AI are really starting to spark the imagination. We’ve gone beyond the threshold of the Turing test. People are starting to see the value of systems that can generate compelling content to solve real business needs.
Cameron Turner
Embracing a new type of creativity (with caution)
Turner says that GenAI has the potential to surprise us with its creativity, offering exciting possibilities across sectors like pharmaceuticals and music. He believes these advances — based on examining data in new ways — aren’t too different from what humans have done in the past, “If we think about the Renaissance or great artists, aren’t we always building on what was there prior?”
However, he cautions that human curation is essential to success. Humans must guide and tune outputs, taking an active role as “guardians of these systems”. AI tools like DALL-E, OpenAI, and Chat GPT can only create content based on the data they have been trained on and can and do make mistakes. A hallucination is when a large language model generates incorrect or illogical outputs. This is because these models produce information based on probabilities, not a meaningful understanding of the questions they are asked.
We're on this pendulum. We swung forward in the initial euphoria of volume content. And then there was this backswing or backlash when we started to see hallucinations. I think, in that swing, we recognize that AI can be great at improving productivity and throughput but there is still work to be done.
Cameron Turner
Building impactful AI solutions
We build the right kinds of AI solutions through really listening to clients. “We’ve learned that even the most non-technical executive has fantastic intuition about the potential of what their data and AI can do for their business.” Turner explains that it starts with perceived opportunity and exploring the technical and non-technical challenges an organization faces — including cultural integration, process change, and training.
He adds, “Once we have a clear understanding of what the opportunity space is, we can then go about identifying the right tooling. A lot of times that discussion starts with data foundations. So developing a data corpus, a data platform that you can start to build not one, but many different solutions against.”
Examining data challenges
AI has broad human foundations, for example, the labeling of data that goes into LLMs requires substantial human intervention. Turner explains this surfaces some important ethical questions: “Are these people getting paid well? Are they exposed to too much harmful content? Are they getting a fair share of the economic value that's coming out?”
He also believes that organizations themselves can be a blocker to development. They often think that their data is not ready for AI. And while it’s true that there can be lots of heavy lifting to do in terms of data slinging, normalization, and cleansing, there is never a perfect time to start. Turner advises that it’s okay to not be all the way there as you’re beginning your journey.
The thing to do is to get ahead of it and abstract a level and develop a governance process that can then manage that chaos. That can be a hard thing to sit with when you're trying to get to the most accurate response based on the highest quality data.
Cameron Turner
AI adoption may not be for everyone
Turner is clear that AI adoption is “absolutely not” a solution for every problem in every business. But he believes that it offers exciting opportunities for a huge range of organizations, from “mom and pop shops” to enterprise. The key to current AI accessibility lies in cloud technology. Easy access to large language models means companies can now combine existing technology with their own data to answer questions that they couldn’t before.
However, Turner cautions that “the biggest place to get tripped up is when human factors aren’t addressed”. This means companies must lead with data, encourage a “core curiosity” and increase data literacy.
Azure AI tools of the trade
We are a longstanding Microsoft Solutions Partner with strong experience in app migration and app modernization in Azure. Turner thinks about applications as a key source of data that can then be harnessed for AI processes inside the new AI workshop. He says that the security and flexibility of Azure make it an essential tool, “In addition to the advantages of convenience, optimization opportunities, and cost, the security of OpenAI models, inside a secure Azure environment, is the headline.”
If you put every Word document and every PowerPoint your organization has ever made into an LLM and start asking questions, you might learn some things you didn't know. And so the goal is how do you harness that and give your employees superpowers versus generating new risks?
Cameron Turner
The possibilities of AI are endless
Learn how to leverage Microsoft AI technologies to support smarter decision-making, personalize customer experiences, and optimize operations by booking a discovery call.
