Artificial General Intelligence (AGI) is a term thrown around a lot, but it's been poorly defined. Until now! In today's episode, I review the "Five Levels of AGI" paper by researchers at Google DeepMind that puts clear thresholds around A.I. capabilities.

Key Context:

• DeepMind's paper categorizes AGI development into five distinct levels, providing a clearer progression toward the achievement of Artificial Super Intelligence (ASI).

• The levels range from Level 0 (no AI) to Level 5 (superhuman performance), with each representing a significant step in AI's capabilities.

The Five Levels:

• Level 1, Emerging AGI (equal to or somewhat better than unskilled human), akin to OpenAI's GPT-4, Meta's Llama 2 and Google's Gemini.

• Level 2, Competent AI (at least 50th percentile of skilled adults): no AGI examples exist today (could in 2024!) but Competent Narrow A.I. examples include smart speakers like Apple's Siri and Amazon's Alexa.

• Level 3, Expert AI (at least 90th percentile of skilled adults): narrow AI (but not AGI) examples exist today such as grammar checkers (Grammarly) and image generators (DALL·E 2).

• Level 4, Virtuoso AI (at least 99th percentile of skilled adults): narrow AI (but, again, not AGI) examples exist today such as IBM's Deep Blue (chess) and DeepMind's AlphaGo (the board game, go).

• Level 5, Superhuman AI (outperforms 100% of humans): when generalized, this will be Artificial Super Intelligence (ASI), but today only narrow AI examples exist like AlphaFold, which predicts protein structures.

So what?

• The paper's framework could significantly influence how we develop, assess, and anticipate the trajectory of AI technologies.

• Understanding these levels aids in forecasting AI's future and preparing for its broader impacts on society.

The SuperDataScience podcast is available on all major podcasting platforms, YouTube, and at SuperDataScience.com.