[DataFramed AI Series #3] GPT and Generative AI for Data Teams
Wed May 10 2023
00:00 - 06:12
- Generative AI is democratizing fields that needed specialized knowledge before.
- Generative AI can generate code and manipulate it to make it more efficient.
- Generative AI can analyze data sets and identify key features.
- Generative AI can help with non-data use cases such as writing presentations, summarizing data, and suggesting visualizations.
- Editing code written by GPT or other generative AI models is similar to editing code written by colleagues.
- Data analysts and those in code-heavy roles will likely use generative AI to generate a lot of code.
- Generative AI can be used for project management, creating reports, and developing lesson plans.
- Half the job of being a data scientist involves working with stakeholders, getting projects into production, writing business cases, and making presentations.
05:48 - 11:42
- AI can be used for structure presentations and paraphrasing things.
- Drawing a shield with different things in the quarters to represent your life is a good exercise.
- GPT language models are the runaway success story and still the favorite of many data practitioners.
- There are other models like Dali that can draw images and Ritrado that generates different images of you in various art styles.
- Managers have more responsibility to think about AI ethics and use it safely for appropriate purposes.
- Writing a good assessment question has all the same key features as writing a good prompt.
Ethics and Evaluation
11:18 - 16:55
- Writing a good assessment question is similar to writing a good prompt.
- Generative AI should not be used for state secrets or personally identifiable information.
- Ignoring Generative AI's existence is not an option.
- Banning Generative AI would be counterproductive in the long run.
- Asking good questions and critically evaluating information are important skills for data scientists using generative AI.
- Being specific and detailed in asking questions leads to more specific and detailed answers from generative AI.
- It's important to critically evaluate responses from generative AI, test code, and look up sources on critical information.
- Templates can be used for routine communication with plug-ins as needed.
- There is fear that some jobs will be completely replaced by generative AI.
AI and Jobs
16:40 - 22:33
- Not all jobs will be completely replaced by AI, some tasks will be automated away.
- Jobs will shift rather than completely disappear.
- AI and humans working together is a productive approach, as seen in medicine studies.
- There are no cases where AI should never be used, but it shouldn't do everything.
- Privacy or security problems are the biggest blocker to using AI.
- Be careful about putting information about people into AI models.
- If you wouldn't want something on the front page of a newspaper, don't put it in public-facing data.
- Think about the impacts if something becomes public or if you get a wrong answer when using AI.
- AI has implications for education and can help people learn new skills.
AI in Education
22:25 - 28:11
- AI has big implications for education, acting as a private tutor and providing tailored learning.
- Personalization, interactivity, and adaptability are key features of AI-powered learning.
- AI can quiz learners on job interviews and evaluate their responses.
- Asking the right questions and critically evaluating information remain important skills in data analysis despite the use of AI.
- The use of AI reduces rote memorization but requires deeper understanding of natural language processing and deep learning for those who want to work with it at a higher level.
Skills and Training
27:43 - 33:54
- Generative AI requires NLP and computer vision skills.
- Awareness of generative AI is important for data professionals.
- NLP and deep learning will become more important in the future.
- Fundamental skills like logic, creativity, and adaptability are more important than specific technologies.
- Prompt engineering involves asking the right questions and giving context.
- Prompt engineering is a skill that everyone needs, but may be a specific job in some organizations.
- Kubrick trains data professionals in machine learning engineering and other related skills.
AI Ethics and Assessment
33:33 - 38:25
- AI ethics is a big issue when it comes to using AI for assessment.
- Using chat GPT on an exam is closer to plagiarism than automating something.
- It's important to set expectations around the use of AI in assessments.
- Collaboration with humans is more beneficial than copying from them.
- There are opportunities for developing new computational chemistry with large language models in life sciences.
- Automating document classification and supply chains can be more efficient.
- The barrier to entry for adopting AI has never been lower, so data managers and teams should try it out.
- Keep perspective on the potential of technology revolutions.