OpenAI’s GPT-3 is a revolutionary language engine that uses 175 billion parameters to generate text virtually indistinguishable from that made by humans on virtually any topic.
Game developer Modbox has taken the next obvious step and connected Windows speech recognition, OpenAI’s GPT-3 AI, and Replica’s natural speech synthesis to create real artificially intelligent non-playable characters you can converse with as if you are talking to the real person.
The main give-away is the lengthy pause between sentences and replies, but that is mainly due to t cloud delay, and could be solved with local processing.
Unless Microsoft implements the idea however the demo will remain just that for the near future – through its partnership with OpenAI Microsoft owns the exclusive rights to the source code & commercial use of GPT-3, and training a similar model is prohibitively expensive for others currently.
It is clear however that it is only a matter of time before we can expect to converse naturally with anyone we meet online, be they humans or bots.
via UploadVR