In this monthly feature, we bullet point the latest developments in the use of generative artificial intelligence to benefit your career (and life). This first note comes courtesy of Ethan Mollick, author of the “One Useful Thing” substack, and includes commentary from me on how it applies to our field. The purpose of the notes below is to learn the basics of how to explain the various differences between AI tools (including hot tips for starting conversations with AI rather than writing to it).
- What is a parametric model? The parameter number represents the complexity of the AI model (the meaning of model is the same as in LLM, it is the world of data that a particular AI tool extracts). A higher number means the model will work better, but it will also consume more resources.
While the smallest models can be run without modification on a mobile phone or laptop, larger parametric models require specialized hardware. Smaller models are less powerful, but they make up for it in other ways.
- What are the advantages of smaller models? In addition to being easier to perform, smaller models are more focused and can perform tasks better in niche areas. The generalist model is broader.
So to help us with our work, we’re probably going to use a smaller model that digs deeper into that niche, like preparing section 16 forms.
- Philosophical difference between co-pilot and agent. Co-pilots are smaller models designed for limited tasks and tend to be less hallucinatory and more predictable. Agents, by comparison, are large-scale models that take full advantage of the power of LLM.
- Difference between co-pilot and agent experience. As an example, copilots, which are often built-in features on mobile phones, tend to respond well to straightforward questions with direct answers. In comparison, agent voice chat is a natural conversation with interruptions, rapid flow, and even subtle changes in tone. It feels very human (fake breathing, pacing, intonation, etc.).
Your tips: Voice capabilities will change the way most people interact with AI systems. Audio and visual interactions are more natural than text and can appeal to a wider audience. Since we will need to have conversations with AI in the future, it would be a good idea to start using voice with AI to learn how it works.
[View source.]
