Artificial intelligence (AI) software technology is a hot development target that is already being applied to many daily tasks. We have heard about self-driving cars, automatic face recognition, simple online customer service duties, and many other tasks where AI is used today, but for some reason, writing is one of the key topics that AI developers like to study. Now, OpenAI has developed GPT-2 text generator that can write amazingly well, and also answer simple questions amazingly well.
The Next Web believes the GPT-2 is already so good that it is scary. Tristan Greene argues that the technology has gone much further than anyone dared to anticipate, and it is high time that we also have a technology that can detect AI-generated content.
It is easy to try out the GPT-2 AI writing system. Point your web browser to the Talk to Transformer page, type a few words, and watch the magic happen.
Of course, I tried it. Simple questions, like “Is it safe to travel to Hong Kong?” was answered surprisingly accurately. I actually received information that I was asking for.
Then, a question that requires knowledge about recent history “If I buy a new typewriter, will I become better writer?” showed that AI machine’s memory doesn’t reach that far back in the history. It didn’t know what a typewriter is, but tried to talk itself out of it. The same happened with text snippets I entered that mentioned Brexit. This may indicate the database behind the AI system doesn’t know too much about EU or the UK (yet).
These examples only show that plenty of data is required in the AI system before it can interact with people, or write so well that people can use the information it provides. It looks like the actual AI algorithms in GPT-2 system are very advanced, so filling it up with more data is a fairly straightforward task.
GPT-2 AI text generator is being developed by OpenAI. It is a U.S. non-profit that has attracted venture capital from renowned investors. OpenAi also has a business arm that aims to commercialize the technology. In both cases, OpenAI says it is committed to safety and benefiting the humankind with its technology.