Note: I first became aware of OpenAI’s language model development work in 2019. Even then, it was clear where their work would lead. In the aftermath of ChatGPT’s release in November of last year, it’s quite clear that the disruptions to work and truth that were envisioned in 2019 are simply inevitable. I am sad for my children. I added a follow-up to the initial post a day later. Both are provided below. (DL, June 11, 2023).
Feed it the first few paragraphs of a Guardian story about Brexit, and its output is plausible newspaper prose, replete with “quotes” from Jeremy Corbyn, mentions of the Irish border, and answers from the prime minister’s spokesman. One such, completely artificial, paragraph reads: “Asked to clarify the reports, a spokesman for May said: ‘The PM has made it absolutely clear her intention is to leave the EU as quickly as is possible and that will be under her negotiating mandate as confirmed in the Queen’s speech last week.’” www.theguardian.com/technology/2019/feb/14/elon-musk-backed-ai-writes-convincing-news-fiction
And who will be able to tell the difference?
I spent ten years as an English teacher. Trust me when I tell you most people definitely won’t be able to.
The search for truth is about to die—shortly before the planet does.
blog.openai.com/better-language-models/
More on the natural language processing algorithm I posted about yesterday.
These results were achieved without any training against factual reality. The goal was simply to create a machine that can create text from scratch that passes as human product. It obviously succeeded.
The next step will be to add actual Knowledge to the model, so that the algorithm can create text that intentionally communicates known information. This will happen. (Don’t get me wrong. It’s not an easy next step. It’s hard. But it will happen.) And a whhoooollleee lot of knowledge worker jobs will be on the line.
Speech synthesis is real already. Siri can already read it out loud. Now Siri (I don’t literally mean Apple) will be able to write it. Siri will become an effective entry-level remote knowledge worker before too long.
And Siri will have a completely artificial face generated from scratch by an algorithm trained on real photos (previously posted). And there will be video.
It’s already difficult to know what’s real. It will become impossible. China for one is going to go all in on manufacturing reality.
In the book 1984, MiniTru had a horde of workers in a building rewriting history on demand. Good concept. Terrible execution. When you control the Internet and have the tools to create reality, Tiananmen Square demonstrably becomes whatever you say it was. And you don’t have to reprint any old newspapers to make it happen.