Historical Archive

Actions and Decisions

• Dr. Emery to send link to Dustan regarding program to create academic content.
• Dr. Xiaoqian Jiang and Tony Murry will join Christina Solis' task force to update BAA AI language.
• Dustan will create a page for accessing AI offerings.
• Tony Murry offered to assist council members in setting up AI resources as needed.

Key Discussion Points

• Update on ADS inventory and focus on Gen AI
• Introduction of Bassel Choucair as new member and officer of data management
• Updates to UTHealth's BAA AI language
• Concerns about transparency and deidentification in BAA signing
• Proposed design for internal marketing of AI solution
• Questions about accessing AI offerings.
• Requests for AI for encounters and English as a second language
• Idea for gateway to monitor AI tool usage. Something similar to this project

Proposals and Suggestions

• Grouping inventoried tools by user case (small research groups, admin staff, clinical support, etc)
• "Watermarking" AI-generated content
• Publishing Educational resources on Gen AI through Coursera

Next Steps

• Follow-up on task force updates to BAA AI language
• Finalize design for internal marketing of AI solution.
• Establish a page for accessing AI offerings.
• Explore options for monitoring AI tool usage and "watermarking" AI-generated content.

Actions and Decisions

• Minutes from previous meeting are accepted by the council.
• Council agreed to post minutes to the council page.
• Mr. Dustan Brennan will post before the next meeting.

Key Discussion Points

• Dr. Art Markman spoke on UT Austin's approach to AI in teaching and learning:
  • Aim is to be both AI forward and AI responsible.
  • Encourage faculty to engage with Gen AI and bring it into the classroom.
  • Created guidance on how to avoid feeding the publicly available systems sensitive data.
  • They are building a RAG (retrieval augmented generation) system for teaching and learning in AWS:
    • Available to faculty to create 'virtual tutors.'
  • Make sure "human in the loop" for automated grading of assignments.
  • They revamped their honor code in a positive framing. Faculty need to be clear what assignments are for.
  • Dr. Markman gave some more examples of tools being used around campus with Generative AI and learning.
• Dr. Christine Weaver raised a question about how UT Austin is navigating copyright issues with using Generative AI models to help write content. Dr. Markman stated that submitting work directly created by a Generative AI model isn't acceptable but utilizing it as a helper or editor is a better workflow. Dr. Markman also stated that students are encouraged to disclose if they utilize a model in helping create original works.
• Mr. Amar Yousif asked for a link to the code of ethics from UT Austin around AI and teaching and learning.
• Mr. Amar Yousif mentioned Dr. Markman's podcast, 'Two Guys on Your Head' to the group and shared the links.

Actions and Decisions

• Christine Weaver asked if the minutes are published externally.
• Minutes were amended to reflect a name change and adopted.
• Amitabha Palmer gave a presentation on ethics and AI.

• Reviewed the ambiguity problem.
• Spoke about top-down vs bottom-up approach to values.
• Gave some examples of a good bottom-up participatory process between organizations and patients.
• Include paper link into final minutes.
• Amar Yousif asked a question about "opting out of AI" for patients in healthcare.
• Dr. Palmer stated that healthcare providers should opt for including nuance in decisions.

• Susan Fenton has sent out the emails for teaching and learning and CoPilot.

• There will be a representative from Dr. Kevin Moranos' office that will sit on the workgroup.
• Fri, Sept 13 is the date for the CoPilot training.
• She asked the council to help get the word out.

Updates

• Ms. Diarra Boye raised a concern about advertising the available Generative AI tools publicly and the cost associated.
• Dr. Weaver raised a concern about the use of data and if we should run this by the data use committee and the availability of trainings.
• Dr. Fenton suggested to perhaps include a statement to say that you need to use these versions vs a private account.
• Amar said to review the public communications to state that they must use this version to remain compliant.
• Dr. Martin Citardi mentioned that there are other Gen AI tools (Doximity, etc.) and suggested creating some kind of mandatory training on Gen AI tools and their utilization.
  • Perhaps we could build this into the Service Now process.
  • Dr. Jiang asked if maybe a small quiz should be included.
  • Christina offered to help with the language.

Actions and Decisions

• Dr. Frost asked a question about what the difference is between the licensed version and free version of CoPilot | *move this to next meeting for discussion.
• Dr. Weaver reiterated the need for staff get adequate training for these tools
• Amar suggested we add this to the next meeting
• Dr. Fenton motioned to accept the meeting minutes for Aug and Christina Solis seconded
• Dr. Fenton gave an update on the CoPilot training, we have 297 attendees with good feedback about the overall training but wanted more hands on training.
• Dustan Brennan gave an update about ChatGPT go live
• Dr Fenton asked about uptick in requests for CoPilot, Tony Murry confirmed that there were an uptick of about 25 up to 140+ total users.
• Christina Solis presented on a lawsuit by the state on a generative AI product for clinical notes.
• Reference
• She suggested we be clear with the metric we use for disclosure to patients about our use of AI.
• Dr. Bruce Butler asked if we are asking if AI is used in the products we buy. Christina confirmed that we are.
• Diarra Boye likened the issue to how VPATs are managed.
• Amar suggested we put in boilerplate language to every contract about AI and manage the rest by exception.
• Dr. Weaver gave an example about how imaging data models are derivative works of the original data, and UTHealth Houston should be licensing that data.
• Dr. GQ Zhang said we need to continue to assess the risk and value of future AI tools and products.

Updates

• Need to determine a workflow for questions about AI, IP and security for newly purchased software
• Amar and Dustan will take this language and work it into the process
• Dr. Butler will send over a sample contract that he wants the council to review to see what the vendor is coming back with.
• Language Model Program: University owned data may not be used in large language model programs without the express written permission of UTHealth Houston.? Such permission must include a description of the type of data that can be used and the parameters of the large language model

Agenda

• Approval of Minutes - 5 minutes
• Presentation from OpenAI Go-to-Market Team Lead - 20 minutes (10-minute presentation, followed by 10 minutes for Q&A)
• Proposal from Dr. Xiaoqian Jiang - 10 minutes
• Updates from Dr. Susan Fenton on the Educational Subcommittee - 5 minutes
• Open Discussion / Adjournment - 20 minutes

Actions and Decisions

• Minutes were motioned to be approved by Dr. Susan Fenton and seconded by Christina. No objections were noted, and the minutes were approved.
• Amar introduced new advisory members. Dr. Gregory Olson from SOD and Ms. Valerie Bomben from the office of sponsored projects.
• Andrew Glenn gave a presentation on new Open AI models and features. He touched on successes with other edu installations.
• Dr. Fenton inquired on how best we can give this to students at scale.
• Dr. Xiaoqian Jiang asked about getting a grant from OpenAI for token credits for educational purposes.
• Dr. Jiang and Dustan presented a proposal to create an AI Hub for the University.
• Dustan put the proposal up for voting and Dr. Fenton seconded
• There were not enough votes to pass.
• Dustan will form a working group with Drs. Jiang, Zhang, Le, Christina that will create the draft design for the hub and present it at the next advisory council.
• Dr. Fenton spoke about the creation of a version of an LLM that students can use that is low cost and easily scalable, secure and internal.
• Dr. Le said that there is utility for the practice plan as well.
• Amar suggested to have two permanent agenda items, one with updates from the development of the aichat application and the other from the hub working group.

Agenda

• Minutes acceptance
• Dr. GQ Zhang's survey proposal
• Dr. Fenton's updates for educational subcommittee and AI-chat project
• Dr. Jiang and Dustan will present AI Information Hub design and updates.
• Open Discussion

Actions and Decisions

• Amar Yousif updated the council on the legislature feedback. It was not submitted based on feedback from government relations.
• Christina Solis motioned for the minutes to be approved, and they were accepted.
• GQ Zhang proposed an AI survey to explore the uses and potential guardrails for AI within our community. The survey aims to understand acceptance levels and identify areas where the use of AI is encouraged or discouraged at UTHealth.
• Susan Fenton said the subcommittee for teaching and learning is planning on sending out a survey in January. They have gotten feedback that they should broaden it per Dr. Zhang's suggestion. The clinical aspect perhaps should be sent separately.
  • Hulin Wu and Dr. Zhang will assist with this effort
  • We possible could work with resources at UT Physicians to engage the patients on surveys about AI
• There is an update for the aichat application, it is being used at MSBMI for student training.
  • Zhang asked a question about the data retention policy for this application, right now there is zero data retention for the application. A suggestion for a disclaimer was raised.
• Dustan Brennan and Dr. Jiang presented the updated AI informational hub design. Dr. Fenton asked for the Gen AI offerings page to be made publicly available. Christina suggested we get some outside assistance to quell fears about usage. Dr. Zhang and Wu suggested to clarify some language on the site.
• Christina mentioned she spoke at a retreat about AI and there was a robust discussion about usage of AI in research. She suggested a group or symposium on evaluating AI for bias in research.
• Diarra Boye asked why MSBMI isn't put upfront to develop AI technology before we go to outside vendors to find a solution.

Agenda

• Meeting minute approval
• Educational subcommittee -Dr. Fenton
• Ensuring Nondiscrimination in the Use of AI - Christina Solis
• AI Hub update - Dr. Jiang / Dustan Brennan
• Collection of questions from council about upcoming legislation for governmental affairs
• Open Discussion.

Actions and Decisions

• Mary Dickerson moved to approve the minutes and Dr. Susan Fenton seconded.
• Fenton mentioned that will the help of Richard Halpin we will distribute the AI survey this week. Changed some wording awaiting some last-minute edits. Additionally, UT Southwestern is interested on the survey and collaborating as well.
• The survey will be widely distributed and working with public affairs to send it out.
• Christina Solis commented that in her presentation last week people were asking if surveys were asking about AI usages at institutions.
• The survey will be open until 2/19/25.
• Christina presented on nondiscrimination and AI:
  • If AI is used in healthcare, there will be a need to assess it for discrimination.
  • She recommended we create policies for review and what to do if we find discrimination.
  • She also recommends creating a workgroup for writing these policies.
  • The policy is archived on the HHS website, but she still recommends creating the group.
• GQ Zhang recommends waiting until the HHS is fully staffed and back going until we create the group.
• This will be a follow up item in next month's meeting.
• Dustan gave a walkthrough of the AI Hub updates:
  • We will add an FAQ for data usage.
  • We will put the hub up for a vote in the next meeting.
  • Weaver wants to add AI in IP course under the educational page.
    • Jiang asked Dr. Weaver to supply some guidelines for IP
  • Dustan will send Mary the link to the prototype see where we can add security.
• GQ Zhang spoke on Deepseek and its implications:
  • Has the ability to be disruptive, he tested it and it had good reasoning and much more impressive than GPT.
  • He feels it will change the paradigm on cost and fine tuning.
  • He suggested the University may not want to highlight specific collaborations to avoid looking like as though we are biased toward one foundation model.
  • Bassel Choucair mentioned that he wonders if this will make the list of prohibited software.
  • Xiaoqian Jiang feels as though o1 still outperforms DeepSeek, but its approaching parity.
• Yen-Chi Le brought up two scenarios that she wanted the AI council perhaps to weigh in on. The main one was about a contract with a company using AI for patients’ mental health and how the disclosures and tools should be distributed.
  • Valerie Bomben is working on guidelines for types of data and their usage with third parties that she will bring to a future meeting.
• Fenton presented the flyer for the AI survey, Dr. Weaver suggested the logo get updated before sending out.

Agenda

Acceptance of Minutes

• Fenton - Survey Update
• AI Hub Proposal / Demonstration / Vote
• OpenAI Deep Research Availability Update
• Update on DeepSeek

Actions and Decisions

• Susan Fenton motioned to accept the minutes, and Dr. Xiaoqian Jiang seconded.
• Fenton provided an update on the recently conducted AI survey. Key insights included:
  • Many faculty members feel pressured to adopt new technology.
  • Fenton is collaborating with Richard Halpin to potentially publish the survey results.
  • Survey findings may be included as a panel topic at the AI symposium.
• Dustan and Dr. Jiang presented the updated AI Hub design.
  • Fenton motioned to approve the creation of the AI Hub, seconded by Dr. Yen-Chi Le.
• Dustan provided an update regarding OpenAI's Deep Research feature and anticipated release date.
• Mary Dickerson shared information about DeepSeek and its GitHub resources, noting approved sources: https://ollama.com/library/deepseek-r1
  • Bassel Choucair raised concerns regarding whether DeepSeek might be categorized as prohibited software.
  • Xiaoqian Jiang mentioned that GPT-4 currently outperforms DeepSeek, but the gap is narrowing.
• Yen-Chi Le presented two scenarios for the AI Council's consideration:
  • A primary scenario involving a contract with a company using AI for mental health services, including considerations for appropriate disclosures and distribution of tools.
  • Valerie Bomben is developing guidelines for third-party data usage, which will be presented at a future meeting.
• Fenton presented a flyer for the AI survey.
  • Christine Weaver suggested updating the logo before distribution.