Are we entering the age of the machine?
As sci-fi-styled hardware and futuristic software advance together nearly in step, observers could be forgiven for wondering if we’ll find out in this lifetime whether human-like androids do, actually, dream of electric sheep.
Emerging and innovative technology companies are continually pushing the bleeding edge of what’s possible with transformative artificial intelligence (AI) and robotics capabilities.
AI already powers a broad class of mobile robots across industries. But with new generative AI models, the robotics capabilities can go even further.
The rapid pace of advancement has led many to believe that this particular moment in time represents the perfect intersection where a robot body with an AI mind could be a functional reality.
With warehouse suppliers in the U.S. predicting they will run out of people to hire by 2024, the timing for bringing to market an intelligent, all-purpose robot couldn’t be better.
See also: Why Sony Venture Fund is Still Bullish on Robots, and Still Thinking About Gen AI
The realities of daily life and industrial operations are generally built around the capabilities of the human form.
Hands, along with the all-important thumb, allow us to use tools, open doors, and move and grasp things. Arms and legs allow us to lift objects, climb stairs, move in a 360-degree radius and more.
The benefits of general-purpose, humanoid robots are appealing compared to the many single-purpose automated warehouse solutions already making an impact across admittedly narrow, and often industrial, verticals.
OpenAI, the company behind ChatGPT, agrees — having led a Series A funding round for humanoid robotics company 1X.
“Finally watched ex machina last night. pretty good movie but i can’t figure out why everyone told me to watch it,’ tweeted OpenAI founder and CEO Sam Altman June 1.
finally watched ex machina last night. pretty good movie but i can’t figure out why everyone told me to watch it.
— Sam Altman (@sama) June 1, 2023
“Ex Machina,” is a 2014 sci-fi thriller about a young programmer chosen to evaluate the human qualities of a highly advanced humanoid robot powered by AI.
And OpenAI’s tie-up with 1X appears to indicate that the AI pioneer wants to give its text-driven large language models (LLMs) a physical hardware body.
Humanoid robots with AI brains were central to OpenAI’s original business idea.
The company first revealed its robotics work in 2019 shortly after launching a for-profit arm, when it unveiled a five-fingered robotic hand powered by an AI model with 13,000 years of cumulative experience, VentureBeat reported.
The disembodied hand could successfully unscramble Rubik’s Cubes about 20% to 60% of the time and was able to recover from challenges such as wearing a glove or having its fingers tied together, according to the report.
OpenAI shut down its robotics division just two years later in 2021, after realizing the sheer amount of data and computing power needed to train robots when compared to building LLM-based AI models, the report stated.
Text-based environments are not only much faster for neural models to process, but there is also far greater access to data-rich text-based environments than there is to robot-relevant video data.
But given that 1X’s stated goal is to design a robot that can “explore how artificial intelligence can take form in a human-like body” of its bipedal robot NEO, the anchor investment from OpenAI could be the beginning of a revolutionary partnership.
After all, innovative firms tackling future-fit problems are typically able to make gigantic progress whenever they have access to vast amounts of data.
Given the white space opportunity for developing a workable and intelligent android, OpenAI and 1X are far from alone in their goal of disrupting the global labor market as well as the broader economy with commercially available AI-embodied humanoid helpers.
Other market entrants are coming from both tech incumbents and emergent startups alike.
One startup, Figure, raised $70 million last month to build general-purpose humanoid robots that can do almost anything a human worker can.
Electric vehicle maker Tesla is also working on its own general-purpose, bipedal, autonomous humanoid robot capable of performing unsafe, repetitive or boring tasks. Competitor Toyota is also developing its own bipedal humanoid machines.
Shenzhen-based UBTech Robotics, U.S.-based Boston Dynamics, Softbank Robotics, and Pittsburgh-based Agility Robotics are also all developing various general-purpose, AI-powered smart robots.
Still, hardware development proceeds at a much slower pace than software advances. The concept of AI-embodied robots is built around the idea that the android can be continually improved over time, learning new skills and eventually expanding into more complex tasks.
After all, a robot solution might be able to lift a box, unload a truck, or answer a question in real time, but can it successfully and cost-effectively scale while driving real business impact?
That remains to be seen.