Except building something the second time around is often quicker and with the current gains of hardware capabilities and interest in the space… maybe it wouldn’t be a year behind.
There are also a ton of ~first mover advantages you can't benefit from, be they of untapped markets for demand or the exploitation of underpriced labor, capital, or IP. If Sam started a new company he would not get as good a deal on hardware or labor, he would get much more scrutiny on his training sets, and he would have to compete against both OpenAI and its competitors.
I agree. Anthropic and Mistral are good examples. Both companies have key people from OpenAI and they fairly quickly developed good models, but I don’t think either are thinking too hard about real AGI, but instead are trying to create useful and ethical tools.
