Google DeepMind CEO Demis Hassabis spent an hour talking to reporters about Gemini’s prospects, the pace of development of strong AI (AGI), and the overall growth of neural network self-awareness. He paid a lot of attention to the Project Astra model, which is currently in preliminary testing. Astra recognizes users and remembers their history of communication with them – these features will soon be available in Gemini Live.
Image source: 9to5Google
Hassabis noted that the promising Project Astra model is distinguished, first of all, by an increased amount of memory. In particular, it remembers key details from previous conversations for better context and personalization. There is also a separate “10-minute memory” of the current dialogue. These capabilities will presumably appear soon in Gemini Live. Hassabis emphasized that Google DeepMind “is training its AI model called Gemini not just to show the world, but also to perform actions in it, such as booking flights and shopping online.”
Hassabis believes that the real time frame for AGI is 5-10 years, and that it will be “a system that really understands everything around you in a very subtle and deep way and is kind of embedded in your everyday life.”
When asked if Google DeepMind is “working on a system today that will be self-aware,” Hassabis said that it is theoretically possible, but he does not perceive any of today’s systems as self-aware. He believes that “everyone has to make their own decisions when interacting with these chatbots.”
When asked if self-awareness is “your goal” (in developing AI), he said it could happen implicitly: “These systems could gain some sense of self-awareness. It’s possible. I think it’s important for these systems to understand you, themselves, and others. And that’s probably the beginning of something like self-awareness.”
«“I think there are two reasons why we think of each other as conscious. One is that you exhibit the behavior of a conscious being, which is very similar to my behavior. But the other reason is that you operate on the same substrate. We are made of the same carbon material with our squishy brains. Machines obviously operate on silicon. So even if they exhibit the same behavior, and even if they say the same things, it doesn’t necessarily mean that this sense of consciousness that we have is the same that they will have,” Hassabis concluded.