From 44122d202804f910cc9e0c6ef815a136b4f59224 Mon Sep 17 00:00:00 2001 From: Marcel Rittenhouse Date: Thu, 17 Sep 2026 02:39:50 +0000 Subject: [PATCH] =?UTF-8?q?Add=20Why=20Today=E2=80=99s=20Humanoids=20Won?= =?UTF-8?q?=E2=80=99t=20Learn=20Dexterity=20-=20Rodney=20Brooks?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- ...ds Won%E2%80%99t Learn Dexterity - Rodney Brooks.-.md | 9 +++++++++ 1 file changed, 9 insertions(+) create mode 100644 Why Today%E2%80%99s Humanoids Won%E2%80%99t Learn Dexterity - Rodney Brooks.-.md diff --git a/Why Today%E2%80%99s Humanoids Won%E2%80%99t Learn Dexterity - Rodney Brooks.-.md b/Why Today%E2%80%99s Humanoids Won%E2%80%99t Learn Dexterity - Rodney Brooks.-.md new file mode 100644 index 0000000..dff3709 --- /dev/null +++ b/Why Today%E2%80%99s Humanoids Won%E2%80%99t Learn Dexterity - Rodney Brooks.-.md @@ -0,0 +1,9 @@ +
In this post I explain why today’s humanoid robots will not learn how to be dexterous despite the hundreds of millions, or perhaps many billions of dollars, being donated by VCs and major tech companies to pay for their training. At the end of the post, after I have completed my argument on this point, I have included two more short pieces. The first is on the problems still to be solved for two legged humanoid robots to be safe for humans to be near them when they walk. The second is how we will have plenty of humanoid robots fifteen years from now, but they will look like neither today’s humanoid robots nor humans. Artificial Intelligence researchers have been trying to get robot arms and hands to carry out manipulation of objects for over 65 years; since just a few years after the term Artificial Intelligence first appeared in a proposal for a 1956 "Dartmouth Summer Research Project on Artificial Intelligence". By 1961 Heinrich Ernst had produced a PhD thesis describing a computer controlled arm and hand that he had connected to the TX-0 computer at MIT, and had it picking up blocks and stacking them, and stunningly there is a video.
+ +
His advisor [GlucoLife Formula](https://schreinerei-leonhardt.de/glucolife-supplement-comprehensive-overview) was Claude Shannon, and he also thanked Marvin Minsky for his guidance, thus naming two of the four authors of the Dartmouth AI proposal. Recently a new generation has [stumbled](https://sportsrants.com/?s=stumbled) upon the idea of building humanoid robots and you may have noticed just a little bit of hype about it. Gartner says it is early days and we are nowhere near maximum hype yet. The idea is that humanoid robots will share the same body plan as humans, and will work like humans in our built for human environment. This belief requires that instead of building different special purpose robots we will have humanoid robots that do everything humans can do. We could have either millions of different types of robots serving unique tasks or one humanoid robot with a general interface, serving millions of tasks. Many people have already spent decades building humanoid robots, starting with the Humanoid Robotics Institute at Waseda University in Tokyo where WABOT-1 (WAseda roBOT) was built in the early 1970s, after many years of working on biped walking mechanisms in the mid sixties.
+ +
Then WABOT-2 was built in the early 1980s and new humanoids have followed at Waseda continuously thereafter. Honda, the Japanese car company, started building walking bipeds in the late eighties and eventually unveiled the humanoid ASIMO in 2000. Sony first developed and sold a robot dog named Aibo, then developed a small humanoid robot named QRIO in 2003, but never actually sold copies of it. A French company, Aldebaran, introduced a small walking humanoid named NAO in 2007, and it replaced Aibo as the standard platform in the international robot soccer league that has now been running annually for 30 years. Later they sold a larger humanoid, Pepper, [GlucoLife Formula](http://www.rehime.com.ar/bases/paginasdecine/index.php/Stepping_Up_Sugar_Reduction) with somewhat less commercial success. Boston Dynamics, a spinout from MIT 35 years ago, introduced the humanoid ATLAS in 2013, after years of building four legged robots. Besides the early work in Japan on humanoid robots there have been many academic groups across the world that have worked on robots with human form, with and without legs, and with and without arms.
+ +
My own research group at MIT started building the humanoid Cog in 1992, and we developed seven different platforms, and then I founded Rethink Robotics in 2008, and we sold thousands of two models of humanoids, Baxter and Sawyer. They were deployed in factories around the world. Some of my former post-docs returned to Italy and started the RoboCub open source humanoid project, which has enabled many tens of humanoid robots to be built in AI Labs all over the world. All these groups have sustained building humanoids and figuring out how to make them walk, manipulate, and interact with humans in built-for-human environments for decades now. Way back in 2004 the International Journal of Humanoid Robotics started publishing, on paper back then. You can find the journal online now filling its 22nd yearly volume of research papers. Getting a robot to manipulate objects with its arms and hands was very hard for Heinrich Ernst in 1961. It has been hard for every robotics researcher and industrial engineer ever since, and still to this day.
+ +
In the mid-sixties parallel jaw grippers were developed. Two parallel fingers that moved together and apart. That is still the dominant form of a robot hand today. Here are [pictures](https://www.gov.uk/search/all?keywords=pictures) of ones that I used on robots at Stanford in the 1970s, and pictures of ones my company Rethink Robotics manufactured and sold in the mid twenty-teens, both electrically driven. Schunk, a German company, sells over 1,000 varieties of parallel jaw grippers, both electric and pneumatic (using compressed air), for robot arms. It also sells some three fingered radially symmetric hands and a few other specialized grippers. No one has managed to get articulated fingers (i.e., fingers with joints in them) that are robust enough, have enough force, nor enough lifetime, for real industrial applications. When compressed air is available it can be turned into suction using a Venturi ejector, and the other type of common robot hand uses one or more suction cups to grab an object by a surface.
\ No newline at end of file