After 15 years of work OpenAI pushed the idea of AGI over the horizon for me

What OpenAI has done is incredibly impressive, and nothing else to say about it. Back in 2008? when I got interested in AGI, I thought about it as people working on a secret project in a basement like they worked on the atomic bomb. What OpenAI has done, if they pulled it off with just a relatively small team, it would be similar in scope to the Manhattan project, but much more impressive. <br>Right after AlphaGo, I thought it would take Google about 3-4 years for AGI and after that it could be anywhere between 5-15 years for Google to do something similar and 10-20 years for others to get similar results. <br>Right after the announcement of DALL-E 2, I thought OpenAI was the closest to AGI, closer even than Google (they did DALL-E 2 with a model that was out for a year or so). <br><br>These results make me think that OpenAI is much more advanced than I thought and I now change my prediction to “never” or “at least 15+ years” for AGI. <br><br>AGI will be a long, long way out.