Self-learning artificial intelligence is one of the pillars of dystopian science fiction, especially when paired with robots. In light of that, tech giant Google is making robots that can program themselves.
Google’s artificial intelligence subsidiary Deepmind is at the cutting edge of AI development. In fact, the group is seemingly the closest anyone has ever gotten to an artificial general intelligence.
On the other hand, Deepmind is dripping with controversy. The AI company has repeatedly displayed a disregard for AI ethics, firing ethicists who find holes in the company’s research.
In an even scarier move, Google seems willing to take out the human component of AI development all together. In a new blog post, the company revealed that its created a robot capable of creating its own code, repeatedly upgrading itself on a software level.
Using natural language models, Google created a robot that can write its own programming based on inputs. Instead of a developer having to write code for the robot to pick up a different coloured block, the robot can simply be told to pick it up. Afterwards, it will write code specifically to tell itself to pick up the correct block.
The self-programming AI tools work by utilising “hierarchical code generation”. This means that the software is able to learn from past created code, repurpose it and iterate on it as it continues to evolve. Essentially, it’s self-learning.
Over time, the Google robot may even be able to create its own programming language. However, at the time of writing, it is only able to use existing languages, including Python and its arithmetic operations.
For now, the Google self-programming robot is reliant on language input to work. On the other hand, the technology could be used to make the robots improve themselves without human intervention.
Of course, the introduction of self-learning robots does introduce some ethical concerns regarding the nature of limits. After all, isn’t that exactly how sci-fi dystopias start? It’s all a bit Skynet, isn’t it?