Peter has written and shared a lot about Large Language Models in recent posts and blogs. We wanted to get his perspective on various things, especially now that Voiceflow have incorporated LLMs into their tool.
How can conversation designers use LLMs now? How does Peter use the LLM as a sounding board to help him craft prompts? Where might we be using them next? How can we document dynamic conversational designs (when LLMs can allow conversations to go off-rails)?
Ben asked him these questions and more…
Have a listen!
In this episode
Contents:
0:00:00 Start
0:02:44 About Peter
0:05:44 What’s your favourite bot?
0:06:57 What Aussie lingo should bots know?
0:09:36 Pick 3 words to describe your career so far?
0:12:03 Do you draw on your former jobs for inspiration?
0:14:06 What’s it like being the first Conversation Design Advocate at VoiceFlow?
0:18:12 What’s the best way to incorporate VoiceFlow into a workflow?
0:21:24 Are you seeing more interest in conversation design now?
0:23:18 Where do you see LLMs being used now?
0:28:52 Could conversation designers become an organisation’s knowledge experts?
0:31:15 We’re all making our first steps with LLMs
0:33:43 It’s easy to fall under the spell of an LLM and think it knows everything
0:35:36 Has conversation design changed forever now?
0:37:08 Are we over-excited or being too timid?
0:40:08 Now we can design conversations that are far more dynamic
0:41:54 Are there best practices for documenting designs with LLMs?
0:44:10 How does Peter use an LLM to help write the prompt?
0:46:42 Have you found ways to avoid ‘hallucinations’?
0:49:56 How would a conversational system work without an NLU?
0:53:32 Are we about to see a new use case for LLMs?
0:57:55 How will this affect the work of conversation designers in the next 15 years?
1:03:12 Why’s there not been more shocking news about misuse of LLMs?
1:05:39 We need to constantly monitor where this tech is going
1:06:52 Outro
