Shane Legg, Chief AGI Scientist and co-founder of Google DeepMind, has been talking about artificial general intelligence, or AGI, for more than a decade. So what’s changed?
Shane Legg says that AGI is more about a field of study than a precise definition. It started out as just a way to distinguish AI that was different from AI focused on a single task. This other thing would be AI focused on general abilities across varying tasks.
Legg didn’t mean it in a sense of a marker on a development graph. Later…people asked questions and he tried to come up with a definition. But he wasn’t able to make any definitions really work… “They always had significant problems,” he said.
Now that we are a number of years into AI development it’s relatively easy to get lost in the weeds so to speak because the change is continual and dynamic.
Which explains the ubiquitous current attempts to understand how good AI is now and how good AI will be next year ..or five years… or 10 years …or ever. Which, as Legg explains in the video, this pursuit of a clear single definition of AGI can be so convoluted as to be not really worth asking the questions… until we get better questions to ask.
One way to find better questions is perhaps to go back to the initial days when GPT chat was first released. The questions were kind of obvious then, along the lines of what what can this do and what can’t it do?
We’d like our questions to be that simple now… which is why there’s all this focus on terminology like AGI, which few can agree on a definition for.
That’s because there’s just way too many questions about what it can and can’t do and different answers as well. So we have a term that defines a category of questions., and is useful as a handle for a category of questions and a category of answers. Beyond that.overwhelming complexity sets in.
Because AI will always be many different things doing many different tasks and accomplishing many different goals in many different ways… yet we still want to know, how good is that AI we have now…and how good will it be tomorrow… and the next year etc.