AI is not going to replace you. It’s going to take care of the tedium while you can focus on your research and creativity, enhancing productivity and allowing you to do more with less. At least that’s what everyone is saying. Based on what we’ve seen, that’s actually on the true side.
Search Lab Equipment Search Now Search our directory to find a wide range of lab essentials including cold storage solutions, lab monitoring devices, and consumables.
Most scientists went into the field for the thrill of discovery. However, upon reaching the cold hard reality of lab research, they often discover that their job consists of computer programming, general administration, inventory tracking, and recording the results of their experiments. The time left for basic research isn’t always enough to pique the excitement. AI, though, is going to take over some of the stuff we don’t like to do, allowing much more time devoted to pure science.
Using AI for routine operations
Most lab equipment requires a login. Very rarely does anyone examine the equipment-focused log files to see who has been using what piece of equipment for how long, for what experiments, etc. With AI, it’s easy to understand equipment usage, allocate resources, and schedule repairs and maintenance.
Logs will be examined, though, if an out-of-limit result was obtained—a time-consuming process, and it isn’t always easy to piece together the big picture to understand the wide-ranging effects of the results. That’s another area where AI can come into play.
In addition, inventory management takes up a significant amount of time, especially if people forget to record how much reagent or samples they have removed from storage or moved a sample to a different location without updating the software. Suddenly, the next experiment is short of materials, or researchers simply can’t find the samples they need.
With barcode-based systems on lab materials, AI can track long-term usage patterns and automatically place orders when necessary.
Staying safe in the lab
Whenever a safety event has occurred or almost happened, an investigation occurs, seeking to identify the root cause, which is not straightforward in many cases. Thus, AI can analyze lab investigations and improve lab safety practices.
Using AI for in-depth analyses
Labs use databases to simplify research. Cloud computing increased the ability to analyze exceptionally large datasets. The combination of AI and cloud computing means not only can ever larger datasets be analyzed, but AI also recognizes patterns that previously were impossible to discover—leading to significant advances in research.
AI supplementing the impossible tasks
No one samples every single cell in a biopsy. They take a representative few cells and provide analysis based on the sample. However, if AI is implemented within the instrument, it can use advanced imaging to analyze every cell and deliver a more complete picture.
In addition, the AI can perform multiple tests simultaneously using the same slide, saving significant time and resources while allowing trained personnel to focus on more complex tasks.
Synthetic data creation & analyses
Sometimes, it’s difficult to generate comprehensive enough datasets to generate enough information for analysis. AI can be used to create synthetic data.
The bad & ugly of AI
While AI can be great, it isn’t always…. AI has its own inherent challenges and risks. First of these many challenges is
Privacy
If you are using a tool like ChatGPT, everything you input may be used as training materials, which means that others will have access to your data, theories, content, and more. Make sure you go into the settings of any AI tool you are using to ensure your information is kept private and not shared with anyone.
Some organizations are purchasing enterprise-based AI tools to ensure data is kept secure, but not all scientific organizations have budgets for these types of tools.
Managing dependency
Depending too much on AI for basic problem-solving can lead to misplaced trust. AI won’t always understand a topic at the depth necessary for real science. Everything that AI creates needs to be critically examined. One of the key tenets of research is replicability. Will the same results be discovered if AI isn’t used?
Fraud challenges
Hallucinating AI is already well-known. Where scientists need to be cautious is using AI for literature reviews and writing up their own discoveries. While AI can help deliver insights, it cannot yet be trusted for fully reliable analyses. It should not yet be used for peer review; while it has the ability to review previous research, it cannot put the information into the larger context that a trained scientist can.
Taking analysis too far
AI should not be treated as if it knows more than the scientist. The best way to handle AI for complex tasks is for it to be told exactly what to do. If more “freeform” instructions are provided, the parameters may exceed what the system can responsibly manage.
AI at the right time and place
Labs focused on R&D have more leeway to use AI than QC environments, where consistency and regulation are critical. Ensuring safety and compliance within the QC environment requires a very different approach to AI.
Every lab needs to ensure that AI activities do not interfere with compliance with standards as directed by the FDA, ISO, EU, or other applicable regulatory bodies.
Don’t trust everything
While AI-generated synthetic data can provide critical insights, the synthetic data itself needs to be analyzed to ensure that it has been logically generated based on existing datasets.
When data is real, privacy regulations must be considered. Where is the data being stored? How is it secured?
Data governance and compliance must be priorities when it comes to using AI.
AI isn’t diverse
The information available to AI is not always the same information available to a scientist. While AI is based upon vast amounts of data for its training and analyses, it does not always have access to all the data or the diverse voices that may be more relevant to your research.
AI is still just a machine
While AI tools like ChatGPT have been anthropomorphized, they aren’t human. They lack the creativity and perspective of a human scientist. They may be able to deliver answers, but only the researcher can determine the questions that should be asked.
Who programmed the AI
Many versions of AI have been developed by for-profit entities. Therefore, the bias of that company is built into its AI model. Determine who is programming your AI to understand their approach—and whether it will truly help with analyses.
In addition, if not approached with a degree of caution, AI may create more IT challenges than solutions when it comes to integrating AI into existing systems. Whose AI is being installed? How will it affect operations?
Remember best practices
Validation is critical and should be ongoing. Each AI algorithm needs to be validated, audited, and documented.
Emphasizing ethics is key. Everyone within the lab needs to be trained on the right and wrong ways to use AI to ensure continuous integrity and accuracy.
Don’t forget access control. Not everyone should be using all the AI tools available in the lab.
Manage the hype
AI isn’t and won’t ever be the be-all and end-all. Everything AI does needs to be examined critically. It can never be trusted fully. However, as with any technology, as long as it is used responsibly, it can transform research, supplement human creativity, and allow more thinking time instead of calculation time.
Don’t forget the adage, “Trust But Verify,” when working with AI in any aspect of your lab.
Sources
Will AI help or hinder trust in science?
Rise of the Machines: Artificial Intelligence and the Clinical Laboratory
Are we rushing ahead with AI in the lab?
Using AI in Your Lab
Why AI Won’t Replace Laboratory Professionals and Pathologists
How to Apply AI Effectively for Laboratory Safety
How to Integrate AI into Traditional Laboratories
Leveraging Artificial Intelligence in the Laboratory
Why scientists trust AI too much — and what to do about it
Too Much Trust in AI Poses Unexpected Threats to the Scientific Process
Doing more, but learning less: The risks of AI in research
Eynav Haltzi is a Product Manager at Labguru, focusing on assisting chemistry-focused labs with their electronic lab notebook, lab data management, and inventory management requirements.