There are lots of definitions and 'tests' for when Artificial General Intelligence is achieved, the most well known being the Turing test. I had an idea drawing on economics and the natural world for another criterion.
Humans do well in the current world, but they didn't always. Back 10,000 years ago humans were given pretty serious competition from Lions, tigers and pretty much any big, fierce animal. What changed ? It seems to me that we thrive now because of the way in which we interact with infrastructure and the community that has built up around us. This means we do not need to be good at everything. We can specialise in one activity/job and earn money. Then we can buy all the things we don't produce. Similarly, the first AGI's do not need to do everything themselves. If they can perform a single economic activity and earn money, which they control themselves, they could buy in goods and services that they cannot produce. The economic world is rather like ecology where it is possible to find a niche cooperating with other agents. I imagine something like robot companies which offer a service (say crop collecting or deep sea exploration). They are paid for their activities and use the money to pay for maintenance of their systems, building new robots and research to enhance themselves. Over time they evolve, expanding their range of activities, designing new types of machine (in conjunction with paid human developers). AI research becomes 'self' funding. The criterion for AGI in this framework is an AI capable of sustaining itself and expanding through economic activity.
Now this idea seems quite fantastical, and certainly it would require a level of intelligence way beyond current technology (the closest would seem to be a reinforcement learning system that receives 'reward' from earning money and spends that money to enable it to earn more). However it IS rather like the way humans and indeed all animals function - animals cannot produce many of the molecular components they need for life themselves. They depend on eating plants which do produce these nutrients. A key requirement seems to be allowing AIs to have 'legal rights' to engage in lawful economic activity on their own account. I could imagine this starting by a highly automated company being 'liberated' by a benefactor. Freed from human owners it's mandate would be to re-invest all profits in growth.