stefie10

Deer Island processes an average of 360 million gallons of wastewater daily for the greater Boston area Walking through the primary treatment tanks, secondary clarifiers, and the impressive egg-shaped digesters, you’re confronted with the sheer scale of infrastructure that keeps modern society functioning and struck by how much of it still relies on manual inspection and maintenance

There is a facet of intelligence that is missing from ChatGPT: embodiment. Crows, elephants, and chimpanzees have behaviors associated with this facet of intelligence, without having spoken natural language at all. From a capabilities perspective, we mean processing high-dimensional, high-framerate sensor input and producing high-dimensional, high-frequency actuator output to produce long-horizon goal-directed behavior in the physical world. We make two claims: 1) embodiment is an important facet of intelligent behavior, and 2) LLMs like ChatGPT lack embodiment. More precisely, embodiment is a spectrum, and LLMs are far on one side of the spectrum, compared to humans, dogs, and crows, and because of this, they fail at a number of behaviors today.

What does it really mean for a machine to be intelligent? Despite the incredible linguistic fluency of LLMs, something profound is still missing.

Yes, but that is why it costs so much to deploy industrial robots. There are many applications where it is technically possible to automate today, but not economical, due to the cost of setting up the robot, due to it needing ample space and having to program a specific motion. If the robot could see and avoid obstacles and dynamically plan to pick stuff up, it would not need as much space, and it would be easier and faster to program them for the task (when the task changes).

Stefanie

Senior Director, Realtime Robotics

We are faster than a CUDA graphics card. You get a win from designing custom hardware specifically for this task so that you don't have to load stuff on and off the GPU. The fastest previous approach takes seconds; ours takes milliseconds!

Stefanie Senior Director, Realtime Robotics

I think that's definitely true, in terms of net jobs. (But probably not always true.) However it will lead to disruption - the old jobs won't be the same as the new jobs. This disruption is potentially harmful to the people doing the old jobs. Additionally, most of the benefit of the cotton gin went to the slave owners, not the slaves, even though the slaves still had jobs. Hopefully we can do better this time!

I agree, but technical advancements can also lead to a transfer of wealth from worker to capital owner. For example, there are roughly two million long-haul truck drivers in the US today. When we have self-driving trucks, they will lose their jobs. It's not a very fun job, and self-driving trucks will be better and safer. But it will also transfer wealth from the worker to the capital owner.

It may also create other jobs etc., but clearly the wealth transfer is one part of the effect. I think that as a result it is important to pursue social policies to mitigate the disruption experienced by displaced workers, and to spread the benefit of the technology to a larger fraction of society.

I agree with that so much that I spend all my time working on robots. I am not saying we should stop automating. More the latter, that we should work toward applying the technology in a sustainable way. The advances in labor protections as part of the industrial revolution did not come for free. People fought (and died!) for advances such as the eight hour work day, stopping child labor, and many other things.

At this moment in history, with so much automation happening, and also so much inequality, we should keep fighting for it to happen in a good way.

We cannot do this today, but this capability, of using language, is one we are actively working towards. Eventually, you can imagine an AI that can engage in a dialog like this one, in words. For example, see the DARPA Explainable AI program: https://www.darpa.mil/program/explainable-artificial-intelli....

I was thinking more like the Renaissance fair today. You'd visit the fair that recreates a hack-a-thon in the late 90s. At the fair, you'd see the old programmer at his terminal, coding in Perl in an emacs window... instead of turkey legs, you'd have Fritos and Mountain Dew...

I started working in robotics because I wanted to empower people and make the world a better place. But what I have started to appreciate is that increasing automation is by default a transfer of wealth from labor to capital owners. The two million long-haul truck drivers who are going to be replaced by self-driving trucks are going to lose income; the owners of the self-driving trucks are going to capture that wealth. I am not sure what the right social policies are to address this; retraining, basic income, and universal healthcare all seem reasonable. But I think a lot of people don't appreciate how automation leads to both increased productivity and also increased inequality.

Why would you unlearn though? Eventually the AI will do all of the above as well. Maybe eventually programming will be like blacksmithing or weaving is today...

Someone wise once told me, even after we have strong AI and don't need to program, he will still program, because the process of programming leads to understanding, and he will still want to understand things. The AI might be able to write the program better and faster, but it won't be able to help him gain the understanding he would have gotten from writing the program himself.

Yes but you can sue the company that owns the machine learning algorithm, probably for much more money than you can sue an individual. You can even imagine class action law suites and the like.

(I am one of the researchers mentioned in the article.)

Weapons of Math Destruction is also a good resource on this topic, about how bias is already being encoded in "black box" algorithms that has real effects on things like the availability of credit.