Artificial intelligence is without a doubt the most transformative digital technology around. While the topic has been off the radar for popular science for some time, the alleged self-awareness of Google’s LaMDA created headlines a few weeks ago. However, that’s far from the only breakthrough in the path toward true artificial intelligence.
While LaMDA raises some interesting philosophical questions, the advancements toward artificial general intelligence (AGI) by DeepMinds Gato is the real news.
Last month, DeepMind unveiled a new “generalist” AI model called Gato. What separates this model from Ai models like AlphaGO and GPT3, is that where those were highly specialized models, Gato can perform more than 600 various tasks, ranging from playing Atari video games, captioning images, chatting, and stacking blocks with a real robot arm.
It is probably the most advanced AI system on the planet that isn’t dedicated to a singular function. And, to some computing experts, it is evidence that the industry is on the verge of reaching a long-awaited, much-hyped milestone: artificial general intelligence.
Unlike ordinary AI, artificial general intelligence (AGI) wouldn’t require giant troves of data to learn a task. Whereas ordinary artificial intelligence has to be pre-trained or programmed to solve a specific set of problems, general intelligence can learn through intuition and experience.
However, not everyone is convinced. the Next Web published an opinion piece, stating that Gato if anything is a sign that we will never reach artificial general intelligence. As a response, Deepminds researchers, s, Alex Dimakis, and Dr. de Freita agree, that Gato is far from being able to pass a Turing test. However, the researchers state that there is a possibility, still, that an AGI could reach a level where it knows it’s being tested and pretend that it is less sentient than it actually is. Previously, it was predicted that we would get AGI capable of human intelligence by 2035, but recently that time frame was shortened to 2028, only five and a half years away.
Others claim that Gato’s performance at its designated tasks is mediocre at best. but that misses the point in my mind, it doesn’t matter if the model at this point is dumb as a brick, after all, we humans used to live in caves and scavenge for nuts and berries. What’s important is whether this is a step toward so-called strong AI or AGI. This would be a single AI system — or possibly a group of linked systems — that could be applied to any task or problem. Unlike narrow AI algorithms, knowledge gained by general AI can be shared and retained among system components. In a general AI model, the algorithm that can beat the world’s best at Go would be able to learn chess or any other game, as well as take on additional tasks. AGI is conceived as a generally intelligent system that can act and think much like humans.
It is a stretch to call Gato a full-fledged AGI, but it is undoubtedly a major breakthrough in the inveitable path towards artificial general intelligence.