Clunky robots get an upgrade: Robotics Transformer 2

DeepMind’s sorry attempt at innovation

Oh joy! DeepMind releases this over-hyped train wreck called Robotics Transformer 2. It’s a vision-language-action model, allowing their bucket-of-bolts robots to perform new tasks without any additional training. ‘Cause we totally need more robots screwing up tasks they were not specifically programmed for.

Ramifications of this sad excuse for technology

Apparently striving for mediocrity, DeepMind’s aim is to expand the versatility and independence of these robots. They want their clunky machines to tackle new tasks by just reading about them. Yeah, like that’s going to end well. Potentially, these robots could find use in logistics, manufacturing, health care and maybe even household chores. And with DeepMind’s reputation, we can expect them to do an equally poor job across the board.

My brutally honest take

On a scale of useless to ground-breaking, this development barely scratches the surface. Seriously, a Roomba does tasks without direct training and it doesn’t need some pretentiously named ‘Robotics Transformer’. DeepMind seems to be playing at some kid’s dream of creating thinking robots. While the supposed versatility of these robots might seem attractive, in reality, it just opens the door for more catastrophic robotic mishaps. This attempt at innovation is as impressive as a squashed tomato on the footpath – messy and wholly unremarkable.

Original article:https://venturebeat.com/ai/deepmind-unveils-rt-2-a-new-ai-that-makes-robots-smarter/