📌 Industrial robots today are programmed to perform specific and repetitive tasks, which limits their versatility; to make them perform different actions, they need to be reprogrammed by a person.
📌 Incorporating #AI into industrial #robots could enable them to learn autonomously from their own actions and the environment; they would thus learn to perform diversified tasks and decide in real-time which task to carry out.
📌 Developing #AI-integrated #robots capable of adapting and learning is no small challenge, especially for complex and dynamic work environments; one of the critical aspects is that robots can cause significant damage, including harm to people, if not properly controlled.
📌 The #startup@covariantai is precisely addressing this issue, currently focusing on the warehouse #logistics sector; to delve deeper, I recommend listening to the episode of "This Week in Startups," which I'm sharing at the end of this thread.
❓ Now, considering the potential developments of a #technology like this, several questions come to mind; here are some:
🤔 If industrial #robots were to become capable of autonomous learning and real-time #decisionmaking, what are the ethical implications of granting them a degree of decision-making autonomy? How should we manage on a legal level the #responsibility for actions taken by such robots?
🤔 #Robots of this kind could reduce the need for human #labor in certain tasks and industries. What #social implications could arise? How should we rethink the right to #employment? What actions should #schools and training institutions take today to prepare tomorrow's workers for a #future like this?
🤔 Considering that these #robots can cause severe harm, how should the responsibilities of #businesses and developers be distributed in ensuring #safety and control of such machines? What kinds of technologies and approaches should we develop to prevent accidents or abuses arising from their autonomy? How should we rethink the concept of safety in the #workplace?
🤔 What will happen to #businesses that cannot keep up with this #technology if it gains traction? On what other aspects can those businesses focus their competitive advantage, and how should we reimagine the very idea of competitive advantage in a world where industrial #robots can learn and work autonomously? What kind of technological and #social divide might arise from these innovations?
🤔 A #robot capable of learning and making decisions about actions to take could be the first step toward the development of entrepreneurial robots, i.e., #AI-driven robots capable of founding and managing new #startups and #businesses. How would the #economy and #finance change if this were to happen?
@krishnadeltoso@startupnews@philosophy@cogsci Difficult questions. I have no substantive knowledge about anything related to business. However, one possible ingredient in the framework for research on such topics always pops to my mind,
The debate on the ethics of AI/robots is still in its infancy and needs acceleration, given the speed at which AI is pervading our lives. It would be interesting to explore how to structure an ethical test for AI that can effectively convey to humans the level of reliability and responsibility inherent in AI data processing, considering that AI can make mistakes and may replicate more or less subtle cognitive biases.
@krishnadeltoso@startupnews@philosophy@cogsci The field of robotics has been building learning systems for decades. While your questions highlight important topics, they haven't suddenly become more important or salient just because AI is better at language.
They are important and salient because we are doing an extraordinarily bad job at
defining allowable operating modes & regimes
imposing pre-deployment test and evaluation constraints
1/
@krishnadeltoso@startupnews@philosophy@cogsci We're doing a terrible job at this in part because we can't even do a good job of describing what, precisely, we want the robot to do and refrain from doing, and in part because robotics is not yet mature enough as a scientific field to have developed generally applicable test and evaluation tools that effectively reduce the risk of using these systems in uncontrolled environments.
2/
@krishnadeltoso@startupnews@philosophy@cogsci We have enough understanding to design the systems, but we cannot look at the performance of a given system in a given environment and predict its performance in other circumstances. Without experimental replicability, repeatability, and generalizability, our test results don't predict operational performance. We lack scientific understanding.
3/
@krishnadeltoso@startupnews@philosophy@cogsci We're developing the ethics alongside the development of that understanding, as well as the test, evaluation, verification, and validation techniques, but we're still at the one-off stage. Any specific problem can be solved with a robot given enough time and resources, but we can't generalize that testing or analysis to other systems or tasks.
4/
@krishnadeltoso@startupnews@philosophy@cogsci At this point, we can't guarantee even partial safety or functionality in the most basic sense, just like software still comes with "we take no responsibility for this working or not" disclaimers.
And if we can't say whether a system will provide the basic functionality it was designed for, how can we say whether it can adhere to any ethical or legal guidelines defined outside the design process?
I'm saying we don't currently have the ability to act on that moral responsibility because we lack fundamental understanding of how to specify and evaluate these systems.
In practice, to define morally acceptable use cases that don't impose unfeasible system design choices, we require the very specification tools we lack.
@robotistry@startupnews@philosophy@cogsci thanks for this thread. The points you raise are fundamental. My thought is that today we're experiencing the result of a hyper-parcelisation of skills and the long absence of truly generative dialogue between professionals (engineers and humanists) who should contribute together to develop new products. The questions I try to ask myself are certainly not new but today more than ever I feel them urgent, precisely for the reasons you underline.
The problem of IA is not how intelligent the device can be. Intelligence is a very elusive concept. It is the responsibility accountability what is behind them and makes all of this an “irresponsible“ step against common sense
Who is responsible for the many deaths Tesla cars provoked (and were hidden)?
@juanfal@startupnews@philosophy@cogsci This represents indeed a significant aspect within the realm of AI, underscoring the substantial disparity between the rapid pace of technological advancement and the relatively slower pace at which regulatory bodies can establish effective and sensible laws. The challenge lies in harmonizing these dynamics to mitigate potential harm without hindering access to new technologies.
@juanfal@krishnadeltoso@startupnews@philosophy@cogsci In the remake of RoboCop the anti-robot senator who more or less gets the plot going (and is named, I kid you not, Hubert Dreyfus...) asks in a hearing, basically, "when one of these armed robots accidentally kills a child, who is going to feel sorry for that child?".
No answer is given but obviously, it's "no one". As with any bureaucratic power structure everyone will point their finger at someone else.
Add comment