ChatGPT-like NLU startups were virtually taboo pre-2010s
It was almost impossible to get your natural language understanding (NLU) AI startup funded pre-2020-ish. Let alone early-2010s. Sure, simple parsers like wit.ai were starting up in 2013 and Google was introducing seq-2-seq models like T5 in 2014 that were starting to do GPT-ish things. OpenAI’s first GPT was not until 2018. GPT-3 2020. ChatGPT 2022.
But the VCs were way behind. Especially in Australia.
I got told nothing was happening in AI in 2014. Let alone in NLU.
LOL.
I was giving them a heads up. But they knew better.
Double LOL.
AI Winters
We had two ‘AI winters’, when little real-world products or funding occurred, fairly well agreed to be in the years:
- 1974–1980 | Followed failed promise of hard-coded early AI
- 1987–1994 | Followed failed promise of early-1980s manually-trained expert systems
But then AI, triggered by practical success of machine learning, has been on an exponential growth ramp, initially imperceptible in the 1990s, but…