Video / Short
Voice Bots
Voice bots are part of the internet's shift from browsing and app APIs toward AI agents that can complete tasks on a user's behalf.
Video summary
The speaker describes voice bots as part of a shift from browsing websites and calling app APIs to asking AI agents to complete a task. In that model, a person explains what they need in ordinary language and a system works through the necessary steps.
The video also notes a practical limitation: voice bots can misunderstand what was said and may continue as though they have the right information. The point is that useful automation needs a clear task and checks for accuracy, not just a convincing conversation.
Video transcript
Read the spoken content without loading the YouTube player.
Voicebots, fad, failure, or future? Well, it seems obvious that you could get AI to make hundreds or thousands of phone calls on your behalf, or to use a voicebot as a receptionist. Well, there's some key challenges which AI is trying to address, but it hasn't completely mastered this yet. Namely, when you're making phone calls, it's not the same for AI because it's going through a bandwidth-limited channel. You probably don't notice this, but when you're on the phone, it's not the same as talking in person, it's not the same as talking to a computer because the bandwidth is limited. What happens then to the voicebot is it tends to misunderstand what's being said.
It doesn't realize that it's perhaps missed a word or misinterpreted something, and it makes things up. LLMs are very good at selling you on their capability, not always being right. Their reward system isn't necessarily geared toward being correct, and this can be a problem if you're running voicebots to try and get accurate information over the phone.