Keep pulling the thread on Ines Montani.
Jodi believes that Large Language Models (LLMs) currently have more interest than profitable applications due to the high costs of running, deploying, and maintaining reliable pipelines.
The host, citing Ines Montani, identifies a major trend in AI development towards smaller, purpose-built models rather than general super-intelligence models.
The host cites research indicating that training a single large AI model consumes an amount of energy equivalent to a person driving a car for one year.
Jessica argues that the biggest problem currently harming the technology industry is a lack of junior-level job opportunities, rather than the impact of AI.
Jodi believes the data science field currently feels hostile to beginners due to the hype around AI.
Jodi asserts that the fundamentals of data science have not changed significantly, but Natural Language Processing and computer vision have recently become more prominent.
Maria predicts that the data science field will be completely different in eight years.
Jodi estimates that approximately 80% of data science work still involves core techniques because business problems are typically solved with the simplest effective models, not cutting-edge technology.
Jessica states that significant security problems still exist with large language models, making it "early days" for deploying them in production environments.
Maria asserts that deploying LLMs for real-world problems is more challenging than commonly portrayed, with issues extending beyond just hallucinations.
Maria identifies key challenges in deploying fine-tuned LLMs, including the costs of inference, electricity consumption, and the resulting CO2 footprint.
The host compares the current hype cycle in AI to the dot-com bubble of the late 1990s, citing companies like Pets.com as an example of excessive spending.