Lab Safety in the Age of AI

Alright, looks like we've got a few more folks joining, but we'll just go ahead and get
started. So welcome everyone again to our OneLab Network event today. My name is Chelsea
Parsons and I'm a consultant with Guidehouse supporting CDC's OneLab initiative.
Just a few notes to mention before we get started, before we dive in, if you're having
any technical issues throughout the session today, feel free to e-mail our OneLab inbox. It's
OneLab@cdc.gov, OneLab@cdc.gov. You'll see that popping up in the chat, and you'll also see
some more helpful links pop up in the chat throughout the day, so you can keep an eye on that.
If you have any questions throughout the session and that is regarding the session content,
please use that Q&A function and you'll see that in the little ribbon down at the bottom of your
zoom web page. You can use the Q&A to put questions throughout the session. We'll have a
Q&A session at the very end of the presentation, where we'll try to get to as many of those
questions as possible, but please pop those into the Q&A. That way we can prioritize those
questions to get them answered during our Q&A.
Session. Please note also that we posted the link to the live captions in the chat. If
you're going to use those live captions today, we just ask that you keep that live captions
window open as well as this zoom window so you can continue to have the presentation on one
page and the live options on another.
So let's go ahead and take a look at our agenda for today? We're going to begin our
session today by just briefly going through a few OneLab resources and then we'll go in and
introduce our main speaker for the session today. And then finally, we'll end the session like I
mentioned with that Q&A followed up by an announcement about some of our upcoming
network events.
Many of you have already discovered that we're going to keep the chat feature open for
your engagement today. Thank you for introducing yourselves and welcoming each other. Just
want to lay a couple ground rules for using the chat box today. So please keep in mind the
appropriate rules of engagement you see on the screen. So again, use the chat to connect with
others, react to what you're hearing, share your stories and experiences with each other.
Maybe even share best practices or questions you may have to your fellow peers and
laboratory professionals. But if you have a question specifically for our presenter, so if you have
a question that is specifically around this content for our presenter today, that's when we'll ask
for you to use that Q&A function. So engaging with one another, reacting to what you're
hearing, sharing stories, use the chat. Engaging with our presenter to ask her question. And
please use the Q&A for that and if you have any questions, you can let our team know. Please
engage with respect and professionalism in that chat. Any inappropriate language, improper
conduct or any form of discrimination may result in removal from the webinar. And Please
ensure your comments are relevant to the topic. So if the moderator gives any direction
regarding chat behavior, just please go ahead and comply with that accordingly.
And again, if you have questions, you're free to shoot me a message on zoom or you can
shoot an e-mail to that inbox, OneLab@cdc.gov. So I'm now going to turn it over to our OneLab
Network lead, Miss Alicia Branch, to share some of our new and relevant resources. Alicia.
Thanks Chelsea. We're excited to announce the official launch of the fundamentals of
Public Health Laboratory e-learning course. Developed from the September 9th, 2022, OneLab
Network event, public Health Laboratories 101. The course provides an overview of the 11 core
functions of public health laboratories and explains how they are implemented to serve
communities effectively.
It also compares the similarities and differences between clinical and public health
laboratories, focusing on their respective roles within the public health system. Next slide,
please. As you can see from our background today, October is biosafety and biosecurity month.
And the division of laboratory systems is excited to celebrate in ways to raise awareness of the
importance of biosafety and biosecurity as clinical, and public health laboratories adapt to
evolving public health challenges.
Strengthening biosafety and biosecurity practices through education, collaboration, and
continuous improvement is critical. That's why the division has featured some initiatives and
they're aimed to raise awareness and foster a culture of safety and security. Also promote the
highest standards of biosafety and biosecurity practices, enhance the safety of our laboratory
professionals, communities and the environment, and support the critical work of laboratory
and biosafe professionals.
We encourage you to engage in the DLS biosafety initiative and a link to the initiatives
has been provided in the chat. Next slide please. I will now read 2 disclaimers. The first
disclaimer I'll explain is that it's actually external presenter disclaimer and then the second
disclaimer will be the professional acknowledgement for continued education, which is our
page disclaimer. Then I'll introduce today's speaker.
Slide decks may contain presentation material from panelists who are not affiliated with
CDC. Presentation content from external panelists may not necessarily reflect CDC's official
position on the topics covered.
Next slide. CDC, our planners, and our presenters wish to disclose that we have no
financial interests or other relationships with the manufacturers of commercial products,
suppliers of commercial services or commercial supporters.
Next slide please. And supporting the division's biosafety and biosecurity initiatives, I'm
excited to introduce Doctor Imke Schroeder. Doctor Schroeder is the Research Project Manager
of the University of California Center for Laboratory Safety, where she is engaged in laboratory
safety, accident analysis, and the develop of research safety training. She frequently conducts
safety climate surveys at academic institutions and is proficient in risk assessments, chemical
and biological agents, and prostheses, including BSL 3 best practices. She has served on various
organizing committees, including the Laboratory Safety Workshop events since 2016, and the
2022 American Chemical Society Safety Summit. And since 2023, she has been the associate
editor for the American Chemical Society's Chemical Health and Safety Journal. Our speaker for
today is Doctor Imke Schroeder. Over to you, Doctor Schroeder.
Thank you so much and let me share my screen. I think you should be able to see my
screen now. Yes, we can see it. Perfect. Let me first give you an overview of what I will be
talking about. I will begin by discussing artificial intelligence applications and their limitations.
The majority of my time I will spend on giving you demonstrations of various AI applications,
compare different visual and voiceover tools, and then finally summarize AI features of
different presentation tools.
So at the end of this presentation, you should be able to differentiate between machine
learning, large language models, and generative AI. Recognize hopefully the usefulness and also
the limitations of AI applications. Evaluate different AI tools for different work needs you might
have and then apply AI tools to create images and presentations.
So I want to start out with a poll question. Are you using AI tools for your work? Are you
using it currently? Do you use it frequently? So it looks like about a third of you, 29% are using
AI tools for their work. But the majority of you is not yet. Or may not ever. Who knows. All
right.
So first let me give you a little bit of background. The first digital computers were built in
the 1940s and some of you may remember the early days of the Internet in the 1990s. And you
probably all remember the appearance of the first smartphone in 2007.
And open AI launched ChatGPT almost two years to the date ago and started a new era
of data, process, data access and manipulation. And here we are. Nobel Prizes were awarded in
physics this year for the discovery in machine learning that paved the way for the AI boom and
also in chemistry for the application of AI for protein structure prediction and design of new
proteins.
Some call this time the artificial intelligence era as it transforms our lives similarly to
how the industrial revolution changed how we live. But potentially it has also negative impacts,
especially if it's not regulated.
So when we talk about AI, we generally use three terminologies or models. All three
models have in common that they are being trained on data. Machine learning can be trained
on very specific data sets, including pixels that are used in science, for example, to analyze
images. Machine learning is used to predict 3D structures of proteins, toxicology outputs or for
example also the generation of synthetic DNA.
Machine learning is really the foundation of large language models and generative AI.
Large language models are trained on massive word data sets and use deep learning to
interpret, manipulate and generate human type language. An example for a large language
model is the GPT model, which stands for generative pretrained transformer. While machine
learning makes predictions and large language models, we use existing content to perform
specific tasks. Generative AI creates new original content such as text, images, music, audio and
video. And if I have time at the end I will play you a Halloween themed biosafety song which I
created using an AI application. So how much do you trust AI technology? Would you ever ride
in a self-driving car?
This is not a poll question, but just a question to yourself.
And so I've been in one of those and it was definitely weird. In the following I will
demonstrate applications for chatbots. There are more chatbots than listed here, but I will
mainly talk about the top five here. All of these chatbots operate slightly differently. ChatGPT,
and Claude, and Copilot are really great for language content creation. I use Perplexity a lot as a
research tool.
HARPA is an interesting Chrome extension that gives me a deeper Google search
experience but note that you can now download pretty much all of these chatbots as Chrome
extension if you use Chrome.
All chatbots are free. However, you can also subscribe to them to gain access to their
enhanced features and faster processing times.
In contrast to other AI's, the free version of ChatGPT and Claude are not connected to
the internet.
And they have been pre-trained pretty much on data or information that's available on
the internet up to 2021 for ChatGPT and 2022 for Claude. So that could pose some limitations
on their usability. I want to mention here that ChatGPT now offers a Bing subscription called
Browse with Bing, in which case it's connected to the internet.
So before I go into and just disclaimer here, I am not promoting or advertising for any of
these particular chatbots and I'm not being paid by them either.
So before I go into the demonstrations, I want to point to some of the limitations.
Chat bots have significant security risks when dealing with sensitive information. Any
data documents you provide to a chatbot will be used for their training.
There are also ethical considerations when data are mishandled or shared without
consent, or purposefully used to infringe on user privacy.
There currently is no transparency on what data are collected by the companies that
actually run the chatbots.
If AI chatbots are trained on bias data or designed with biased algorithms, they can
inadvertently propagate and reinforce these biases, including gender, racial, and cultural biases.
Also, chatbots can generate false information and this is especially problematic if where
they receive their information from is not revealed. So it's very, very important if you use any of
this technology check and edit the content that's created by them. So here's my second poll
question. Does your institution currently have a policy for AI use? It seems that again about 1/3
of you 35% are at institutions that have a policy, but 65% of you are at institutions that does not
have a policy or you may also not be aware that your institution may have a policy. So there is
lots of work that still needs to be done by institutions.
So what should you do if your institution does not have a policy for AI use? So here's
some recommendations that I would provide. Chatbots should not be used for applications
involving sensitive or confidential information. This includes, for example, student records,
medical records, research data, or any intellectual property that your institution, or you may
want not to be shared. They should not be used for anything of questionable ethical nature,
such as generating academic content.
They should not be used in admissions or hiring processes because of potential biases.
Also, they should definitely not be used for counseling where human empathy and judgement
are required because I just last week went to a talk by someone who was in the developing of
early AI applications and they just described it as an empty, emotionless box.
Finally, do not use them for formulating a confidential communication. However, there
is a way to use AI tools for confidential data.
If your university or institution has a subscription service to it and a firewall, and I will
describe this, how we are doing this at UCLA later in my talk.
So here are some of the potential chatbot applications, but and I'm going to go give you
examples for most of them, but make sure if you use them to fact-check and edit the content.
So all right, now I'm going to ask ChatGPT the question: how much energy is consumed
by AI computing? This is ChatGPT. It states here at the very bottom in fine print check, GPT can
make mistakes. Consider checking important information.
And the nice part of ChatGPT is that if you're on the site, it has created a library of what
I've worked on with it in the past and you can go back at any time and revisit what you've
queried?
And you can also ask it follow up questions, but I now want to do a new search, so I'll go
to new chat. Paste in my question. And hit enter and you can now observe how quickly ChatGPT
generates the answer to my question. So this was ChatGPT.
And I queried Claude with the same question and it gave me a very succinct summary.
The energy consumption of AI servers could range from 85 to 134 terawatt hours annually by
2027, which is right around the corner which compares to the electricity usage of an entire
country like Argentina, Sweden, or the Netherlands.
And so that does not even take into consideration the substantial amount of water used
to cool the supercomputers, and in fact the energy uses by AI data centers rivals that of Bitcoin
mining. So AI tool usage has to become sustainable. Let's say you want to understand Einstein's
theory of relativity. In this case, we will start out with a normal Google search and then add
HARPA to it, and HARPA serves now as an assistant or copilot to your search. So here we go.
I'm now going to do a Google search using Google Chrome. So I'm entering my prompt
and up come a bunch of websites, some videos,
and more. So now going to call up my Chrome extension HARPA right here on the top.
And HARPA immediately picks up the prompt and provides me with a brief summary
and even related queries that I can enter. But this is still too complex for me to understand, so
I'm going to ask it to explain this to me at the level of a 5th grader. And here you have a much
simpler language that you can use to impress your office mates.
All right, so you can also now use HARPA to translate the text into any language, and in
fact any chatbot will do this for you. I cannot verify this text, but when I translated it into
German, the translation was spot on. So note that in the translation it retains its specialty
characters. This is especially important if you want to translate text into any of the the Asian
languages.
So check the chatbots. Pretty much all the chatbots can be used to improve language. So
here's an example of an excerpt we received by a subject matter expert for electrical safety
training.
And so it reads: if your institution allows it and you have been trained, you may
deenergize and isolate anything that is on fire and use a fire extinguisher of the C type, which is
non-conductive. And when I saw this I said, "What?" So I entered it into Claude.
And Claude transformed it into something that's very succinct and informative without
being given more background about the subject matter. So it really recognized that this had to
do with an electric fire and was just amazing in transforming the text for me.
All right. So let me show you another example for using AI to create. We use it a lot for
safety training. In this case, I'm using it to create a bloodborne pathogen training. So first I'm
prompting the chatbot with tell me about bloodborne pathogens as they relate to research
safety.
And then I will ask it to create a 20-minute course outline based on two PDFs that I'm
attaching.
So I'll go to the first prompt.
Let's start there.
Here I am using Perplexity. I have entered my prompt and within seconds perplexity
provides me with the answer which I can now review.
What I like about Perplexity is that it provides me with the sources that it used to
retrieve the information.
So now you can hear this is a different voice because when I was recording this, I had
lost my voice. And so this is what you heard now is an AI voiceover. And I'll talk about AI
voiceover towards the end of my talk again.
So what I'm going to do now, I'm going to attach two PDFs.
One of them is an OSHA bloodborne pathogen training and the other one is of course,
the Biosafety in Microbiology Laboratories book, which is also available as a PDF, and this is a
pretty big book or PDF.
Now I am going to use Perplexity to create an outline for a 20-minute interactive elearning course on bloodborne pathogens for scientists and research labs.
The more specific you make your query, the better response you can expect.
I am attaching 2 documents. The documents should best be in PDF format. It is
uploading now which can take a few minutes since especially the BMBL file is large.
And I hit enter, it now takes a moment to analyze the two documents.
Perplexity now asks me follow up questions to see where I am going, since I want to
make my training not too heavy on regulations, but want to emphasize understanding of the
subject, I will leave out the OSHA standard. I can always include that in a later query. It now
analyzes and comes up with an outline, which I can now rephrase, refine, and flesh out.
So this is Perplexity and PDF's can be added not to all chatbots but Claude and
Perplexity will accept PDF analysis.
So I can now take the content.
And so just the beginning of the course outline and you can see it looks pretty
reasonable.
And so I can take now the content and ask Perplexity to flesh it out.
And so you see here more detail that I can now use to build my training.
I can also use Claude to write multiple-choice questions, and that's a huge time saver.
This is Claude.
I am copying into the prompt my query and the fleshed-out introduction from Perplexity
and then I hit enter and Claude generates 3 questions, provides the answers and the feedback
for the correct answer.
So also very, very reasonable test questions.
As I mentioned before, it's a huge time-saver for these kind of applications.
So you can not only use PDFs, you can also use videos as a source of information and I
just want to mention here, if you use PDFs validated material.
to create content that really helps to avoid misinformation that's potentially created by
the chatbots. So I'm going to now use Perplexity to create a video outline for an electrical safety
training.
But you can also use the videos or PDFs to summarize and to create potentially fact
sheets or other materials that you need.
I'm now going to use Perplexity to create an outline for an electrical safety course from
the two-hour webinar by Jim Gilson.
So I've entered my prompt and asked it to create an outline and what I also like about
Perplexity is that besides the original YouTube video it provides me with additional resources
that I might find interesting in that backs up also the material.
And what you see here is the answer.
So here's the outline introduction, understanding electricity.
Regulatory requirements.
Risk assessment.
Establishing an electrical safe work condition, PPE. Safe work practices. And emergency
response.
So this is pretty impressive what it generates and it gives me a very structured start to
present to my subject matter experts for them to flesh out.
All right.
I now want to give you an example for using AI for sensitive data.
UCLA now has a subscription service to Copilot by Microsoft, which we can access by
logging into the university site, so it's behind a firewall.
And so this slide was taken from UCLA's website to train users on the application, and
it's important to note that Copilot encrypts data that are inputted and also deletes all data after
the session is complete. So the data are not being used for training the chatbot.
So I want to now show you how you can analyze data from an Excel file using a chatbot.
This could be a survey or could be sensitive incident data.
And for this demonstration purpose I'm using publicly available data that I downloaded
as a CSV file from the Internet. It was really hard looking for Excel files on the internet with data
in it. So I found data on refugee camps that were pretty depressing. So I settled on data from
the World Happiness Report 2023.
So. I copied the data directly from the Excel sheet and pasted them into the dialog query
following my prompt to analyze the data and it looks like this. And so then I asked Copilot to
analyze the data. And so this is basically the output. It not only summarizes the first column, but
also independently looks at the other columns in the data set. I formed, so you see here GDP
per capita, social support, healthy life expectancy, freedom of making choice, generosity, and
perception of corruption and how that all correlates with happiness. And I performed the same
analysis side by side with Perplexity and Perplexity, really comes up with a very similar analysis,
very similar output, but has a slightly different take maybe. So for example here it lists also
again a strong correlation between GDP per capita and happiness.
But it also notes that however, some countries like Costa Rica, is anybody from Costa
Rica here, or likes to travel there, achieves high happiness scores relative to low GDP per capita
data, suggesting other factors that may play a role there. So it's amazing in this analysis that
these chatbots do statistical correlations. So that's just an example. I now want to switch gears
and talk about image creation.
This is an image of UCLA's mascot that was created with Adobe Firefly. So we have a
brown bear. And the brown bear is now in a lab coat with gloves and safety glasses. And so we
like Adobe Firefly best as compared to Canva, DALL-E or Midjourney. And I want to give a big
shout-out to Alicia, who created the images, these images here, and she's the instructional
designer for training development for our organization. So this is a comparison of Adobe Firefly
images. Adobe is a standalone web app. Or it also works as part of Photoshop and other Adobe
products. One could almost use the first image that you see here and from 2023, except if you
look closely the hands are a little weird here and maybe face are also a little weird. A little bit
improvement over all the faces.
But the lab coat is open. The hands are still little problematic here. The lab coat is open,
although we asked for the lab coat to be closed. So it's still not quite there yet. This is an
example by Canva. Canva in April 2023, pretty Halloweenish, I'd say.
Didn't get the safety glasses correctly at all. Scary looking faces, Almost like vampire
maybe. Hands, well, improvement in July 2024, but the hands, I don't know what's happening
here and I have no idea what happened with the hands here. All right. OK, here's another very
creepy example by Canva with a different prompt.
I don't know how they came up with the hair net here and the hands very problematic.
DALL-E's much better. But not photorealistic in 2023. If you now have Copilot for Windows, you
can enter a description of the image you want to create into Copilot, and Copilot now uses
DALL-E 3 to create images. So this is still DALL-E 2. And so DALL-E is improving as compared to
the earlier versions here. But you see here the two white males. We're looking for a little bit
more diversity.
And not entirely happy with the hands.
Okay, Midjourney. Example Midjourney 2023. Vey cartoonish. Very sincere looking
scientists but big improvement in July. Also, to be much more photorealistic.
Midjourney is available for free. But it works very very slowly. So in order to generate
this image we have to subscribe to it to get a faster processing speed. All right, image
limitations. There's still problems with rendering hands. Incorrect PPE, lab coat not closed,
safety glasses problematic. Definitely look out for biases and also read the user rights carefully.
Do you have to cite the image? Is the image that you create actually yours to use?
Look closely into that. OK, AI voice-over. So I've talked about, used it earlier and we are
using ElevenLabs. And you heard in some of the voice overs earlier that there is some
problematic language like he, the, OSHA. That can be corrected. I'll give you a demonstration of
that. Other developers are using MURF AI or WellSaid, and they're all better but we found
ElevenLabs works best for us and was the cheapest solution.
So. Hi, my name's Matilda and I'm one of the voices available from ElevenLabs. In this
video, I'll show you the process of generating voiceover using ElevenLabs. Before you begin,
adjust the settings to select the voice that best matches your project's needs. You can change
the voice settings to make the voice sound more variable or stable. You can also make it sound
more clear and similar to the target voice.
In this example, we'll leave the default settings. There are also different models of
ElevenLabs, depending on which languages you need narration in. For my voice, ElevenLabs
recommended using version one for the best quality. After adjusting the settings, simply paste
in the text you need narrated. Note that Elevenlabs only allows you to paste 2500 characters at
a time, so you may need to generate separate audio files for longer scripts.
Select the generate button to listen to the voiceover. While I'm generally accurate with
most words, I sometimes struggle to pronounce technical terms correctly and am sometimes
inconsistent in how I pronounce the same word. For example, by default, I pronounce the word
reagent as regent.
The developer had to add a hyphen between re and agent to ensure proper
pronunciation. When the narration is acceptable, select the download icon to download an
MP3 file. So you can use these voiceover tools also for different languages. And I now want to
finish up my presentation with talking about building presentations and want to give you an
example for Slides AI, TOME, and at the end for PowerPoint.
So this is an example for Slides AI. Which is an extension of Google Slides of the Google
Slides app. The pros are that it creates decent placeholders for talking points, although the
designs are quite simple so not too much better off as compared to your general. PowerPoint.
And the cons are that many images, Slides AI suggested were really not relevant, so I'm not
impressed by this very strange looking image that it suggested, and also the copyright of the
image, was unclear. This is an example slide created by TOME.
TOME creates the text and the image after your prompt. But it takes more of a
storytelling approach. Pros is it provides you with many options to trim down the text and edit
the text and also change the tone. And you can tell it you want to have a little bit of text,
different text, or shortened text.
The cons, the free version is only accessible via the web, and a paid account is required
to download a PDF of your presentation, and so it cannot be downloaded as a PowerPoint file
or a Google Slides. Does not have that type of export option. Also, many of the AI-generated
images by TOME uses DALL-E and we not generally like those images. So this is the
demonstration by Alicia on PowerPoint AI, which is now available through the Microsoft
subscription, and you may or may not have played with it. So note that Alicia made this video
that I'm going to show you next.
Accessible by adding caption at the bottom of the screen and this is a feature also
available by PowerPoint. In this video, we'll explore how to use PowerPoint's AI tools to both
build and improve your presentations. So in this example, I have a Microsoft 365 subscription
and I am logged into microsoft365.com.
I've selected the PowerPoint app and I've entered some simple text and I will now select
the designer option from the home tab over here. The designer option to get some suggestions
about slide design layouts. So as you can see, it's come up with some options here. You can
scroll down and if you want you can see more design ideas by clicking on that. But I actually like
this option here.
So I'm going to wait for that to load. All right. And then if I wanted to, I could make this a
little bigger to adjust to the slide dimension size there. Alright, now that I'm happy with how
this slide looks, I can move on to the second slide and then Designer will give me other options
for this slide. So I can scroll down, I can see more design ideas. There.
And we'll just go with this one for now. But I do have this empty area and I would like to
fill it with a photo and so for that I can go up here to the image options. Select that. And you
want to be careful about copyright. So right here it's selected Creative Commons only and I can
scroll through. I can see it's suggesting some images relating to tools, other things relating to
software. I can scroll through this until I find something that I like. We'll just add this image as
an example here.
So once I'm happy with the overall design of the PowerPoint I can move on to practicing
the presentation by going up to the slideshow tab and then selecting the rehearse with coach
option. And if you select this arrow here, you can also choose to get feedback on your body
language as well as certain settings which we'll go over in a little bit. Once I select the rehearse
with coach option you can see that there is this pop-up here and I will select the start
rehearsing button. I can also get real-time feedback.
So for this example we'll start with the presentation and we'll pretend that this is a
longer presentation. It does mention that it will give you better feedback if you do practice for a
longer session, but for now this will be good enough to give you a snapshot of the type of
results you can get. We'll pause this session and then we will close out the slideshow.
And you can see this rehearsal report of the time spent, the slides rehearsed, which was
just the one here, and then whether we use any filler words such as, umm or ah, whether we
repeat the same language that is on the slide or if we add some original words with our
presentation. Inclusiveness, whether you use inclusive language. And then you can also get
feedback on the speed of your presentation. So how fast you talk and then how fast you talk
over time.
And if you want to rehearse again, you can select that button to get more
recommendations. So those are just a few of the examples of tools within PowerPoint AI that
can be useful for you as you get ready for your next presentation. Of course, Alicia is the
absolute presentation nerd. If you want to become a presentation nerd you can try this
presentation tool by PowerPoint. So let me summarize my talk. You can use chatbots to
increase your knowledge, create content, summarize existing content, translate content into
different languages.
You can play around with different chatbots. There are more than just one, more than
just ChatGPT, but be aware of their limitations and definitely verify the content.
Lab-related AI images still need improvements. AI voiceovers are relatively user-friendly,
although some people still prefer the human voice over the AI voice. And presentation AIs can
save you time and can make the presentation more interesting. And so thank you to all of you. I
now will give you a little a thank you AI biosafety song. Are you ready? OK, let's go.
Ebola in a test tube glowing neon green. Salmonella creeping like it's Halloween. Rats
with tiny love comes running wild at night. CRISPER/Cas is casting spells with genetic light. In
the lab of horrors, we mix terror into science. Biosafety rules are fighting off defiance, masks
and gloves our ghostly armor gleam in this haunted lappy dance in the nightmare dream. Petri
dishes whisper secrets of diseases. Microscopes reveal the monsters in the yeast. Pathogens
are partying under black light globe.
In the lab of horrors, where the wicked vials flow. In the lab of horrors, we mix terror
into science. Biosafety rules are fighting off defiance. So I'm apologizing for ending the song,
but I wanted to leave some time for questions because we're running already late here and if
you wanted to have the song, I'm super super happy to send it to you. So I'm happy to share it.
So with that, I'll open up the questions. OK, we'll take the question. If we don't answer your
question today, we'll do our best to answer via e-mail. If you have any questions after today,
you can e-mail us at OneLab@cdc.gov. Let's see.
Let's take a couple of these. This is a good one. How do you envision AI being used by
the average clinical hospital laboratory? It seems like the hospital workers will be the one
generating data for the AI to use versus using AI in their everyday work. That yes, and that's a
very good question. And if you have data, data analysis by AI is amazing.
And again, I would recommend a subscription service. I would definitely check out
different AI options. I know that I know one university subscribed to ChatGPT data analyst, so
there are some even specialty AI applications that you might want to explore. But even sending
out routine emails that you might want to use, you can summarize certain policies that you
have or certain emails that should be. You want to have maybe an e-mail that has an
encouraging tone.
You can have that generated by AI. Yeah. And we'll take one last question. Do you know
if there is a course that provides some training? Because I think that, you know there was you
presented a lot of great information. Sure, you know a lot of the participants will need like
some additional training. So do you know of any platforms that offer the training, or a course,
maybe YouTube?
No, that's a very interesting question and I think I would be very open in helping to
develop such a course or giving maybe even live online a hands-on exercises. So one could
envision potentially having a workshop on training people on how to use AI. So I think that
might be a very good possibility moving forward. I'm very much advocating for regulation of AI
use to work together with the institution and really home in on what our employees allowed to
use in their AI application. I mean unfortunately AI is not regulated currently in the United
States. There's more and more coming out.
It is regulated in the EU. That's another entire talk on AI regulation, but I just really want
to emphasize that currently the institutions themselves have to take responsibility. And just like
with any other technology, users like you and I, we have to be accountable and responsible for
the use of any type of technology, including AI.
Great. So that's all the time we have questions for, but what we can do is send you the
questions and then we can have you answer them and then we can e-mail them back to the
participants. Yes, absolutely. I'll like say thank you again for the excellent, insightful
presentation. Thank you all. Thank you for inviting me.
You're welcome and we hope to have you back to maybe do one of those trainings.
We're offering one P.A.C.E. credit for today's webinar, the P.A.C.E. certificates are immediately
available on the OneLab REACH in your my learner hub for easy access.
You receive your P.A.C.E. credit after participating in today's session by logging into your
OneLab REACH account. Visit the link in the chat and use the passcode to complete the
evaluation within two weeks. You must be logged into your OneLab REACH account to access
the evaluation. You will also receive an e-mail containing these instructions if you miss the link
and the passcode after today's presentation.
So before we closeout I want to highlight some upcoming OneLab events. The first event
is OneLab Test scheduled on October the 23rd at 12:00. PM Eastern Standard Time. Patient
Testing is Essential. Get the right results. This two-part series and final presentation will focus
on the test phase of Ready? Set? Test! It aims to provide insight and best practice for patient
testing. Please register using the link in the chat and also I would say consider viewing the video
and slides from the first session in the past events on passing this section on the OneLab TEST
web page.
Next slide please. And our final OneLab Network event for 2024 titled Something is
Strange. Let's Notify the LRN will occur on November the 19th at 12:00 PM Eastern Standard
Time. This webinar is a panel discussion with experts Erin Bowles, who is the Wisconsin Clinical
Laboratory Network Outreach Coordinator. Shoolah Escott, who is a biosafety, biosecurity, and
bioterrorism preparedness trainer, and Chris Manual, who is the director of Public Health
Preparedness and Response at the Association of Public Health Laboratories, or better known
as APHL. They will present a clinical laboratory response example, new tools, and valuable
resources. The link is posted in the chat. Next slide. And as a reminder, the slides and audio
recording of this event will be posted to our website within two weeks of today. And in
conclusion, we encourage you to utilize the OneLab inbox to share your training needs and
feedback on OneLab with us.
We use your input to select event topics and better understand your laboratory training
needs. The OneLab e-mail address is posted in chat for easy access. And again we thank you for
joining us today and have a great rest of your day.