LIVE
Impressive working video from OpenAI-supported humanoid robot company 1X
1X, a humanoid robot company supported by OpenAI, focuses on chaining tasks with its robot called Eve. A robot that can perform sequential tasks is approaching full autonomy.
Just like the artificial intelligence developments experienced after the release of ChatGPT in November 2022, it seems that a similar development has occurred on the humanoid robots side. The first ChatGPT or similar ones before it worked only in question and answer format. However, today we can give them a task and make them fulfill the requirements of that task. That's exactly what humanoid robot firm 1X, backed by #ChatGPT developer #OpenAI , is trying to incorporate into its robot, Eve.
While #1X wants to provide physical labor through safe, smart androids, the steps it takes also serve this purpose. The new video published by the company shows the humanoid robot Eve's ability to complete autonomous tasks one after the other. However, the company also points out that this is just the beginning of the journey.
The company had previously developed an autonomous model that can combine many tasks into a single target conditional neural network. However, when multitask models were small (<100M parameters), adding data to correct the behavior of one task often negatively affected the behavior in other tasks. Although increasing the number of parameters comes to mind as the first solution for this, this time the training takes longer and delays which indicators need to be collected to improve the robot behavior.
So how can we quickly iterate over data while creating a general robot that can perform many tasks with a single neural network? 1X's answer to this is quite clever. The firm says it separates the ability to quickly improve task performance from the process of combining multiple capabilities into a single neural network. To achieve this, the firm created a voice-controlled natural language interface to chain short-term capabilities into longer ones across multiple smaller models.
