Physical AI, $1 trillion from SoftBank, and the Department of Robotics?
Don’t call it a chatbot.
SoftBank bets $1T on US robotics infrastructure, China pours $888B into its own, and Tesla calls for a national robotics strategy.
At first glance, this might not seem like an AGI news roundup; but it is.
SoftBank doesn’t have a trillion dollars
It’s true, SoftBank doesn’t have anywhere near that kind of money. But, to be fair, SoftBank didn’t have the $500 billion dollars it promised to invest in the Stargate Project earlier in the year. So it does sort of seem a bit risk-seeking for it to suddenly decide to double down and invest a cool trillion instead.
Note: the Stargate Project is an OpenAI sibling firm that’s spinning up to build infrastructure for AI hardware farms. SoftBank and OpenAI are the “lead partners” in the venture.
Still, it’s apparent that we should all take SoftBank seriously. A recent analysis indicates that SoftBank has the credit to borrow at least $450 billion towards its previous investment goal of $500. And, for those who’ve been following over the long term, SoftBank has a long history of executing relatively high-risk investment strategies to support nascent technology.
These investments don’t always pay off. For example, it lost 99% of its value when the dotcom bubble burst. However, there’s an argument to be made that SoftBank’s contributions prior to that market collapse helped lay the groundwork for today’s worldwide web.
With that in mind, what could OpenAI possibly do with a trillion dollars? We think the answer is pretty clear: physical AI.
We predict one of the following two scenarios are on the immediate horizon (6 to 36 months):
A laboratory with at least as good a reputation as OpenAI’s will announce that it has developed AGI and is willing to defend its claims.
Multiple top-tier AI development firms will announce a pivot to “physical AI.”
It’s difficult to point to any evidence for number one. Despite the flurry of papers showcasing AI developer conjecture on the nature of “thought,” the (perceived) needle towards AGI doesn’t seem to have moved significantly since GPT 4.0 launched. Research is good, but it’s hard to imagine a firm with something to lose taking on the challenge of convincing the Gary Marcus’s and Francois Chollet’s of the world that they’ve stumbled onto human-level thought in a chatbot.
There’s a lot to be gained by convincing the public you’ve reached AGI, but a lot to be lost if they’re not onboard.
Read more: Google, OpenAI, Artificial Intelligence benchmarks, and AGI — Center for AGI Investigations
Meanwhile, the US faces mounting external pressure as China has reportedly invested nearly a trillion dollars in its own automation infrastructure.
Physical AI
We’re predicting that US firms toward the top of the “race to AGI” will pivot to physical AI as a means to continue showing progress toward “human-level” AI. It’s easy to imagine a torrent of papers discussing the “embodiment problem” flooding arXiv by the end of the year.
Here at the Center for AGI Investigations we have a hypothesis that says “a model without real-world agency may only be a theoretical intelligence.”
And the closer developers get to the edges of the Uncanny Valley with chatbots, the more likely users are to accept an alien modality such as humanoid robots.
We’ve reached a point where the ability to hold a conversation with a person is no longer the most impressive thing a robot could do. This indicates the general public is ready for the next “big thing.”
Department of Robotics?
In related news, representatives from Tesla, Boston Dynamics, and Agility Robotics met with US lawmakers last week to discuss the implementation of a national strategy for robotics.
The big idea appeared to center on maintaining US competitiveness against China — as mentioned above, the PRC has poured about $888 billion into its own robotics industry.
Chatter surrounding the meetings seems to suggest that tech leaders in the US would like the federal government to form some sort of central department to oversee efforts to scale the AI and robotics industry.
While that’s probably not a bad idea, we find it ironic. By the end of this year the US could have a Department of Robotics but no Department of Education.
Art by Nicole Greene
