The Ethics of Large Language Models with Amber McKenzie
Large Language Models like GPT-4 are all the rage - how do we use them well? Carl and Richard talk to Amber McKenzie about the challenges and concerns around LLMs, especially regarding the data involved. Amber talks about the risks of creating products around LLMs while the technology is so young and constantly evolving, especially with how the general public reacts to conversational AI. Education is key, and constantly fact-checking - good advice for a lot of technology, but LLMs are making fact checking even more important!
Guests:
Amber McKenzie
Dr. Amber McKenzie is head of Data Science and Analytics at HR tech start-up Fama. With a master's degree in linguistics and a Ph.D. in computer science, she has almost 15 years of experience in data science, AI, and natural language processing (NLP). She has led a variety of projects in a number of different sectors - including military applications, healthcare, marketing, law enforcement, accounting, and adtech - during her time at the University of South Carolina, Oak Ridge National Laboratory, DialogTech, PwC. Her professional interests include NLP, machine learning, predictive modeling, and computer learning. When she does get a spare moment outside of work, she enjoys reading, board games, indoor rock climbing, and lifting weights.
Links:
- GPT-4 Dumber Over Time https://www.digitaltrends.com/computing/study-shows-a-downturn-in-openais-gpt-4-response-quality/?mibextid=Zxz2cZ
- ChatGPT can now remember who you are and what you want https://www.theverge.com/2023/7/20/23801520/chatgpt-openai-custom-instructions-chatbot
- Open AI Tokenizer https://platform.openai.com/tokenizer