Ready to learn more?
Request a demo to learn how your team can accelerate R&D
Blog
Exploring Human Validation in the Age of AI at SmartLab Exchange
No items found.
Albert hosted a think tank at SmartLab Exchange on the role of AI in innovation. Read a recap of key themes such as human validation and change management.
The power of partnership
Earlier this year, a team from Albert attended SmartLab Exchange USA in Fort Lauderdale, Florida. We were honored to attend this unique, invitation-only event that brought together R&D leaders and solutions providers to share knowledge around digital transformation in the lab. From presentations and panels to think tanks and one-to-one meetings, this event was thought provoking and informative. A key takeaway was that there isn’t always a perfect vendor or single straightforward way to implement digital change – but if you choose a trusted solutions partner, you can build something transformative together.
The role of AI in innovation
Lenore Kubie, PhD, Principal Solution Engineer at Albert Invent, hosted a think tank on the topic of digitalizing materials science for the age of AI. The discussion began with where different organizations currently find themselves on the path to digital transformation. While experiences ranged from being brand new to further along in the process, we observed several key themes:
The low-hanging fruit
Most organizations started their digitalization journeys using tools that can handle unstructured data, such as leveraging large language models (LLMs) for document parsing and search. In other words, many organizations had created an in-house version of ChatGPT that acts as an internal subject matter expert.
While this is a logical place to start because it does not require clean, rigid formats for data, document parsing tools are just the tip of the iceberg. Structured chemical and experimental data – in other words, AI-ready data – are necessary to unlock the deeper value of AI. Some participants had begun to structure data to tackle specific problems, yet none were doing so at a large scale. Data platforms built for chemical R&D – such as Albert – make it easier for organizations to generate AI-ready data, paving the way for them to move beyond chatbots and into the world of AI-assisted design of experiments, formulation, and synthesis.
The value of human validation
As AI becomes more common in R&D workflows, scientists often grapple with the question of whether AI will replace them. At our think tank, a shared sentiment was apparent: while AI will play an increasingly important role in innovation, human validation is key. An analogy was made to the relationship between a principal investigator (PI) and a new graduate student, where the PI should provide feedback or validate results if data from the graduate student does not match their intuition. This is how a scientist should interact with AI.
Importantly, this kind of relationship still requires the knowledge and expertise of the chemist, and an early career chemist may struggle more with interpreting the output of AI. It also suggests a strong need for building trust in AI models – something that Albert Breakthrough, our suite of AI/ML tools, was designed for. Breakthrough requires constraint inputs from the scientist and provides both model performance (how accurate the model is) and feature performance (what features the model found to be important) in its output.
While AI will play an increasingly important role in innovation, human validation is key.
Change management
Despite increased interest around digitalizing R&D, it was not uncommon for organizations to run into challenges during the rollout process. Change for change’s sake is generally not well-received, meaning any efforts around digital transformation require ties to a broader enterprise strategy and a specific plan. One successful strategy shared in our think tank was to pick a specific digitalization project, use its success story to sell it to the rest of the company, and ensure that the win is closely tied to clear company goals.
Henkel ran into similar challenges bringing digital transformation to their organization of 2,800+ scientists. Here’s how they implemented a successful tiered rollout.
Building the future together
Our team came away from SmartLab Exchange USA with new perspectives and connections. We’d like to thank the organizers for hosting such a valuable event and we look forward to staying in touch with those we engaged with.
If we didn’t get the chance to connect at SmartLab, Request a Demo to learn how we can be your trusted partner and build the future together.
The power of partnership
Earlier this year, a team from Albert attended SmartLab Exchange USA in Fort Lauderdale, Florida. We were honored to attend this unique, invitation-only event that brought together R&D leaders and solutions providers to share knowledge around digital transformation in the lab. From presentations and panels to think tanks and one-to-one meetings, this event was thought provoking and informative. A key takeaway was that there isn’t always a perfect vendor or single straightforward way to implement digital change – but if you choose a trusted solutions partner, you can build something transformative together.
The role of AI in innovation
Lenore Kubie, PhD, Principal Solution Engineer at Albert Invent, hosted a think tank on the topic of digitalizing materials science for the age of AI. The discussion began with where different organizations currently find themselves on the path to digital transformation. While experiences ranged from being brand new to further along in the process, we observed several key themes:
The low-hanging fruit
Most organizations started their digitalization journeys using tools that can handle unstructured data, such as leveraging large language models (LLMs) for document parsing and search. In other words, many organizations had created an in-house version of ChatGPT that acts as an internal subject matter expert.
While this is a logical place to start because it does not require clean, rigid formats for data, document parsing tools are just the tip of the iceberg. Structured chemical and experimental data – in other words, AI-ready data – are necessary to unlock the deeper value of AI. Some participants had begun to structure data to tackle specific problems, yet none were doing so at a large scale. Data platforms built for chemical R&D – such as Albert – make it easier for organizations to generate AI-ready data, paving the way for them to move beyond chatbots and into the world of AI-assisted design of experiments, formulation, and synthesis.
The value of human validation
As AI becomes more common in R&D workflows, scientists often grapple with the question of whether AI will replace them. At our think tank, a shared sentiment was apparent: while AI will play an increasingly important role in innovation, human validation is key. An analogy was made to the relationship between a principal investigator (PI) and a new graduate student, where the PI should provide feedback or validate results if data from the graduate student does not match their intuition. This is how a scientist should interact with AI.
Importantly, this kind of relationship still requires the knowledge and expertise of the chemist, and an early career chemist may struggle more with interpreting the output of AI. It also suggests a strong need for building trust in AI models – something that Albert Breakthrough, our suite of AI/ML tools, was designed for. Breakthrough requires constraint inputs from the scientist and provides both model performance (how accurate the model is) and feature performance (what features the model found to be important) in its output.
While AI will play an increasingly important role in innovation, human validation is key.
Change management
Despite increased interest around digitalizing R&D, it was not uncommon for organizations to run into challenges during the rollout process. Change for change’s sake is generally not well-received, meaning any efforts around digital transformation require ties to a broader enterprise strategy and a specific plan. One successful strategy shared in our think tank was to pick a specific digitalization project, use its success story to sell it to the rest of the company, and ensure that the win is closely tied to clear company goals.
Henkel ran into similar challenges bringing digital transformation to their organization of 2,800+ scientists. Here’s how they implemented a successful tiered rollout.
Building the future together
Our team came away from SmartLab Exchange USA with new perspectives and connections. We’d like to thank the organizers for hosting such a valuable event and we look forward to staying in touch with those we engaged with.
If we didn’t get the chance to connect at SmartLab, Request a Demo to learn how we can be your trusted partner and build the future together.