Blog

The Limitless Possibilities of Chat GPT Now and Then

Share:

What if tedious administrative tasks in a healthcare organization – like writing reports and emails or creating call schedules – could be done by Artificial Intelligence (AI) saving time and allowing physicians and staff to focus more on patients?

UCSF’s Associate Professor of Urology Anobel Odisho, MD, MPH, put it to the test when he generated a call schedule that only took 3 seconds with Chat GPT, a task that can sometimes take hours to create.

Odisho recently appeared on GU Cast to weigh in on Chat GPT, the AI online chatbot developed by OpenAI that has been creating a buzz on the internet. While it has become widespread because of its public availability, generative AI like the Large Language Model technology isn’t new, according to Odisho citing Bidirectional Encoder Representations from Transformers (BERT) that has been around since 2018.

Chief of Urologic Oncology at Michigan Medicine Todd Morgan, MD, urology trainee Jonathan O'Brien, and Liam Mannix, National Science Reporter at The Age and Sydney Morning Herald, also join in on this disruptive technology discussion on the episode: “Chat GPT This is Nuts! AI Goes Mainstream!”

Some highlights from the episode:

  • Skills must be developed to learn how to interact with Chat GPT, just like any other new technology. The output is dependent on the prompts given and can be tweaked further to give a different output each time.

  • Media outlets have been using AI such as:

    • The Washington Post has an AI tool called Heliograf used to generate content since it was first introduced during the 2016 Rio Olympics, and in that year wrote 850 news stories.

    • Bloomberg News which has about a third of its content AI-generated.

  • There can be significant bias in what the algorithms are trained on. For example, some are trained on Wikipedia text that is 90% written by men; a variety of topics are not covered or women are underrepresented, which will impact the output.

  • Algorithms need to be monitored and patched as people exploit them, for example in 2016, Microsoft’s chatbot Thinking of You (Tay) was shut down within 24 hours when it started tweeting offensive messages.

  • There are equity benefits and risks, e.g.:

    • Possible Benefit - it can be a leveler allowing someone with great ideas to communicate them if writing is not their forte.

    • Possible Risk – who will own the algorithm, and who will regulate it?

Watch the GU Cast episode “Chat GPT This is Nuts! AI Goes Mainstream!” for more insight into ChatGPT.