Researchers at Google’s AI division, DeepMind, have written a report proposing a sort of ranking system for artificial intelligence, which is intended to classify both today’s AI algorithms and the future AGI systems (Artificial General Intelligence) that are expected to emerge.
The researchers drew inspiration from how self-driving cars are currently classified on a scale from 0 to 5, based on how autonomous the cars are considered to be. Google DeepMind’s ranking system for AI and AGI consists of five levels ranging from “emerging”, systems that are somewhat better at tasks than an untrained human, and then progressing through the categories “competent”, “expert”, “virtuoso”, and “superhuman”.
In the report, the researchers write that Google DeepMind’s own AlphaFold, an AI that can predict so-called protein foldings, can be classified as “superhuman” as it is considered to be able to do this better than any human can. They also write that chatbots, such as OpenAI’s ChatGPT and Google’s own Bard, could be classified as “emerging” in terms of AGI system classification.
Whether the tech and research world will adopt a common ranking system for different AIs and AGIs remains to be seen. If so, it might become easier for us humans to at least somewhat understand how competent certain AIs and AGIs are.
