Artificial intelligence has come a long way in a short period of time. While the underlying technology has existed for decades, the last few years have seen breakthrough after breakthrough in the level of intelligence exhibited by AI.

At the forefront of AI evolution is OpenAI, with models like GPT-4 and tools like ChatGPT, but they are now giving these models legs in the form of robots from startup Figure.

Sam Altman's AI Lab has invested millions of dollars in Figure, an AI robotics company developing general-purpose humanoid robots. The company is now worth $2.6 billion and has received investments from Microsoft, OpenAI, Nvidia, Intel, and Jeff Bezos (not Amazon, just Bezos).

For some time, the fields of AI and robotics have been tied together by an invisible thread. The two are clearly one field, but have developed independently of each other.

In the early days of OpenEye, the company tried to create a robotics division, but realized that hardware and software were still two distinct entities. That changed last year with the rapid improvement of multimodality.

"We always plan to return to robotics, and with Figure, we see a path forward to explore what a humanoid robot with a very capable multimodal model can accomplish," said OpenAI's Vice President of Products and Partnerships Peter Wellinder said. [We are overwhelmed by Figure's progress to date and look forward to working together to open up new possibilities for how robots can be useful in everyday life.

Multimodality essentially means that AI models can understand and interact with more than text. Google's Gemini and GPT-4v are multimodal models in that they can take input as text, code, images, video, or voice and interpret them equally well.

This is essential to allow the robot to do its own thing without a human having to write every single task it needs to perform before it can perform it.

Figure AI was established in 2022 with the goal of allowing robots to work in manufacturing, shipping, logistics, warehousing, and even retail, where "labor shortages are most acute."

While there are bots available for specific tasks in each of these industries, Figure is going the humanoid route, creating machines that can adapt to different situations as needed.

At this level of technology, two things are important: funding and computing power. The funding will bring its value to over $2 billion, meaning that the partnership with OpenAI can be used to adapt already-trained models and provide the machine with a mind.

The goal is for Figure 01, the company's first android, to be able to perform a variety of "everyday tasks" autonomously, which is where AI minds come into play.

We have already seen university researchers experiment with OpenAI's GPT-4 to help robots learn poses and movements from simple text prompts, such as Alter3.

Allowing robots to learn what to do on their own directly from a factory conveyor belt, or at least after minimal training, would be a game changer for industry.

Google has a research department that integrates AI into its machines and is working on enabling robots to learn routine tasks and analyze scenes to perform unseen functions.

Much of this is due to advances in computer vision technology, particularly AI vision, where the underlying model can take in real-world views from cameras, examine the situation, and determine what is needed. [Brett Adcock, founder and CEO of Figure, said, "Figure's vision is to commercialize humanoid robots as quickly as possible. He said the new investment will ensure that the company is "fully prepared to bring embodied AI to the world to have a transformative impact on humanity."

I think we will see the term embodied AI more often in the future. Basically, it's ChatGPT with legs, kind of like Google Gemini, but with the ability to walk around and talk to you in the real world.

If you haven't worried about AI before, imagine if ChatGPT could walk up to you and chat, or ask you why you didn't thank it for making your dog's picture.