According to tech titans like Bill Gates and Elon Musk, the long-awaited robot revolution is finally here. When these machines arrive, they will require computer-powered brains that maximize the capabilities of cutting-edge artificial intelligence technology.

That looks like a lot of work; in the past, robot designs were purpose-built, which meant that their systems were created only for the particular tasks that the robot was intended to perform.

A startup named Skild AI believes it has a better idea and has received $300 million in funding to support it from major investors like Jeff Bezos.

Skild is producing a “general purpose” robot brain at a lower cost than custom-built robot software. This is essentially a prefabricated artificial intelligence model that can be installed into almost any type of general purpose robot.

These devices may be “quadrupeds mastering adverse physical conditions” or “vision-based humanoids performing dexterous manipulation of objects for complex household and industrial tasks,” according to the company’s most recent press release.

According to Forbes, two former Carnegie Mellon University professors, Abhinav Gupta and Deepak Pathak, founded Skild in May 2023. The business model is also fairly straightforward, much like many of the greatest tech concepts.

Similar to typical “chatbot” AI systems, Skild’s main AI model, the “robot brain” system, was trained by being exposed to a vast amount of real-world data, including text, images, and video.

However, the Skild “brain” was also trained on robot control tasks, such as ones in which a human remote operator guided a robot through basic tasks, teaching the AI how a human would handle physical movements, and other training tasks in which the AI was given random tasks and had to learn them through trial and error.

Interestingly, Forbes claims that because of this synthesis of various data kinds, Skild’s AI has begun to exhibit emergent behavior. It can therefore execute actions and tricks that weren’t included in its training set.

These movements to retrieve a slipped object or to rotate an object it is manipulating into the proper orientation are examples of these subtle behaviors. Nevertheless, humans almost always make these small “corrections”; this is part of what makes them so adept at general-purpose work.

In contrast to, say, an old-fashioned industrial robot in an automobile factory that works quickly and precisely but depends on the object it’s working on and all of its tools being in exact locations, a robot that could integrate some of these more subtle motor skills would be far more flexible.

Subtle adjustments that resembled those made by a human were a crucial component of the amazing humanoid robot that another startup, Figure AI, unveiled earlier this year. However, the robot serves as an excellent illustration of a specially designed and coded device that depends on OpenAI, an outside AI provider. Skild is trying to disrupt this kind of business model. Jeff Bezos of Amazon and Lightspeed Ventures, two of the biggest venture capital firms in Japan, have invested $300 million in Skild after being impressed by the company’s business plans.

In the meantime, news concerning venture capital firm Andreessen Horowitz suggests the investment firm is covertly entering the “AI chips-as-a-service” market, if you consider Skild’s robot AI brain to be “robot smarts-as-a-service.”.

As per the news website The Information, Andreessen has been covertly accumulating a “stash” of over 20,000 sophisticated AI processor chips, such as the state-of-the-art Nvidia H100 units, which have been instrumental in training numerous highly renowned AI chatbots available in the market.

The concept behind this new “Oxygen” plan is that access to powerful AI chips is currently limited due to high demand (Elon Musk, for example, has come under fire for allegedly taking Nvidia chips away from Tesla to use in his other AI projects). As a result, the VC firm will be able to provide access to startups in its portfolio. ​

When you consider the money, the idea makes sense right away. With all of Andreessen Horowitz’s wealth and connections, startups are far less powerful. The initiative’s name is further reinforced by the fact that, in contrast to some earlier technological breakthroughs, such as the .com revolution, generative AI systems rely heavily on expensive, powerful hardware in order to operate. The “oxygen” that keeps their portfolio AI companies breathing is provided by Andreessen’s AI chips.