Has conversational AI crossed the quality threshold to become the primary way people interact with computers?

Noah Desai Weiss · Jun 16, 2023

OpenAI's smartest recent strategic move was packaging large language models in the smallest possible product — an input box and a text output box — to give the consumer world a glimpse of a new model of human-machine interaction

They recognized something magical in a technology that was sitting in research labs and academic papers, and the minimal packaging let people envision a very different future, then executed on it

47:23 20Product: Slack CPO Noah Weiss on How to Master Product-Led-Growth, The Biggest Mistakes Founders Make When Scaling Into Enterprise & What Needs to Change with your Product, Team and Processes when Scaling From PLG to Enterprise

David Luan · Jun 24, 2024

The gap between the transformer and ChatGPT's consumer breakthrough existed because two conditions had to coincide: models crossing a minimum viable smartness threshold and being packaged so consumers could play with them

The GPT-3 API shipped over a year before ChatGPT but only developers could use it, so there was no viral consumer moment until chat tuning and packaging arrived

7:55 20VC: Why Foundation Model Performance is Not Diminishing But Models Are Commoditising, Why Nvidia Will Enter the Model Space and Models Will Enter the Chip Space & The Right Business Model for AI Software with David Luan, Co-Founder @ Adept

Des Traynor · Nov 15, 2023

Chatbots have moved through three generations — rigid if-this-then-that IVR logic, 'fuzzy AI' intent matching, and now LLMs — and the LLM era is generationally different rather than incremental

Fuzzy-logic bots had broader coverage but pissed people off because the logic wasn't good enough and you still had to hand-build rules; LLMs can actually infer the specific question and find a precise answer

8:14 Llm era represents a generational leap beyond rule based and fuzzy nlu chatbots

Brad Lightcap · Apr 15, 2024

ChatGPT's adoption comes from being the first genuinely human experience people have had with the technology and from one tool serving radically diverse use cases, which is what translates into business impact.

The same tool powers a research scientist's productivity, a software engineer's code and a new parent's daily questions; something that fundamentally diverse and accessible is bound to have large adoption impact

28:05 First human ai experience and diverse use cases drove chatgpt adoption

Bret Taylor · Oct 2, 2024

We have entered the era of conversational software: with GPT-4 conversational AI crossed the quality threshold of effectiveness, the way multi-touch did for smartphones

You can now have a conversation with a computer that actually works — it understands nuance and sarcasm — just as the iPhone's multi-touch plus big screen crossed a quality threshold that made BlackBerry keyboard loyalists switch

39:24 Gpt4 crossed the conversational quality threshold like multitouch did for smartphones

Bret Taylor · Oct 2, 2024

Conversation (text, voice, multimodal) will become the main way people work with computers, because consumer adoption is driven by convenience and low friction rather than by being the theoretically best interface

You can converse in the car, the kitchen, or WhatsApp without installing an app or reading an instruction manual — the same convenience logic by which touchscreens came to dominate despite typing being easier on a keyboard

43:17 Convenience not interface optimality drives mass adoption of conversational computing

Alex Lebrun · Jun 19, 2023 · hedged

After two decades of the technology not being ready, market timing for conversational AI may finally be right now

The product is finally good enough, and users themselves have changed over the ten years since Siri; earlier bots were embarrassingly unintelligent, as when a 2004 rail-company bot greeted the deputy CEO as 'hello, wrong on the reservation'

4:10 Market timing for conversational ai may finally be right given improved tech and changed users