?????????? AI. Interview with Amro from the Xpikuos Channel, Robotics and Artificial Intelligence Engineer and Digital Master. Part One.
#ArtificialIntelligence #Robotics #AIinAction #TechnologicalFuture #AIdebate #TechnologicalAdvancements #DigitalTransformation #RoboticInnovation #DigitalSociety #AIandRobotics #FutureTechnology #AIandSociety #TechnologicalChallenges #AIintheWorkplace #EthicsinAI #TechnologicalDevelopment #Xpikuos #AmroInterview #IA-ismointerview #
[Disclaimer: The following text has been written by a human but has been translated to English using AI tools.]
We inaugurate an interview section. Due to its length, as we wanted to ask Amro many things, we will divide this interview into two parts. Here, we present you with the first part.
Every day, the number of channels dedicated to artificial intelligence continues to multiply. However, back in 2019, if you wanted to learn about this discipline, there were pioneering channels like Xpikuos that specialized not only in artificial intelligence but also in robotics. Today, we are fortunate to have this digital master answering some questions for our IA-ism newsletter regarding the state of the art in Artificial Intelligence and robotics.
Amro: I am an Engineer-Master in Telecommunications with a specialization in Automation and Control Systems, currently working as a programmer in an R&D department at a German company. Since I was very young, I knew that I wanted to learn how to build robots that can speak and communicate with humans. This passion led me to develop an interest in AI, robotics, and programming. A few years ago, I created the YouTube channel "Xpikuos" with the aim of teaching people how to build robots with artificial intelligence. So far, I have focused on explaining topics related to machine learning and neural networks, as well as showcasing some of the robots I have built and their components. However, the intention is to gradually expand the channel's content on robotics and explore more interesting and useful AI topics for robot construction. Due to personal reasons, the channel has been inactive for the past two years.
Amro: An AGI is a General Artificial Intelligence whose objective is to imitate or even surpass human intellectual capabilities. It is truly the holy grail that many AI researchers ultimately pursue. However, it certainly carries certain risks that, in my opinion, are far from resembling a robot invasion that will take over our homes and annihilate us as a species. The risks will instead stem from the lack of regulation in its use by governments, leading to a massive loss of jobs within a time frame too short to allow for personal and social adaptation to its use in a constructive and collaborative manner. We can already see this happening with the emerging AIs that drastically simplify tasks for graphic designers, writers, customer service personnel, drivers, translators, etc., to the point where they can be replaced in some cases...
and those are not even AGIs. Imagine what will happen to other professions such as programmers, doctors, or lawyers when AGIs become a reality.
In fact, there are already AIs solving certain typical legal problems. Personally, I believe that in a short time, we will see general practitioners being replaced by an AI and a small device that you can buy for a low cost and have at home to perform quick blood tests, measure blood pressure, temperature, and even auscultate the sound of lungs and heart by moving it across the chest. With this data and a description of symptoms provided by the patient, the AI will prescribe the appropriate medication or refer you to the corresponding specialist, while also booking a convenient appointment for you. In my view, anything that can be automated will be automated if there are no laws to prevent or regulate it.
Regarding content creation, we are currently witnessing a boom in AI-generated content across various platforms (blogs, websites, YouTube videos, etc.). This will also have short-term consequences. For instance, I imagine that soon we will see YouTube, and by extension, Google,
facing challenges in storing an enormous amount of newly generated information within much shorter intervals than what was previously required for human-created content.
It wouldn't be surprising if these companies start reducing the rates paid to content creators. Moreover, the number of channels will increase even more, intensifying the competition among them. Initially, many creators will thrive, but in the long run, people might grow tired of generic content videos where the creator's face is not shown, and they may highly value those videos in which the creator does appear.
Personally, I think YouTube might implement a policy that rewards videos featuring the actual creator, which could disadvantage those who don't want to lose their anonymity or are not visually appealing. However, it's also possible that none of this happens, and Google sees an opportunity to increase its advertising revenue due to the abundance of content. Nonetheless, with the current advancements in AI models capable of creating animated 3D models of people and cloning their voices with inflections, it becomes increasingly difficult to distinguish between the appearance of the real creator in a video and that of their digital clone. This is why many researchers, philosophers, and influential figures in the field of AI are advocating for the
领英推荐
use of some kind of mark or label to indicate creations made with the assistance of AIs.
Personally, I would like to know, for example, at the beginning of a phone conversation whether I am being assisted by a virtual assistant or a human being. The same applies to reading a text, watching a video, or viewing an image. The problem is that even such a mark could be counterfeited, leading to various legal and identity theft issues that could have devastating consequences. Imagine a cloned president appearing in a viral video declaring war on a country, or the opposite scenario: a real high-ranking official issuing a warning about an imminent catastrophe in a video that goes unnoticed due to the abundance of "fake" content to which people have become accustomed. Elections could be manipulated with false news and rumors that appear entirely real. Conspiracy theories would proliferate, and trust in anything or anyone would erode because anyone could create any kind of fake audiovisual content. Consequently, legally valid evidence such as videos, photos, and documents would no longer hold up in court.
Returning to the topic of content creation, I believe that those who are highly creative will stand out from the rest. With little money and minimal time, they will be able to produce diverse, high-quality, and visually appealing content.
In my opinion, this is truly the key to all the upcoming changes: creativity and multidisciplinarity.
Knowing how to do something or not will not be the problem.
What will matter is having different ideas from the rest, combining them with other fields, and implementing them as quickly as possible.
The downside is that these ideas will also be rapidly copied using these tools, leading to an extremely fast pace of innovation. In such cases, we might employ an AI to monitor the competition and, applying creativity techniques used by humans, immediately create a new product, idea, or content that surpasses the competition. This could be an example of using an AGI to automatically achieve a goal: creating a new product that is different and better than what the competition offers. It would no longer be necessary, as it has been until now, to iteratively write texts like with ChatGPT until the desired result is obtained. In fact, we already have examples of this, although not AGIs, such as AutoGPT, capable of generating text and code given a description, and self-refining through multiple iterations until it achieves the desired outcome.
2. In the field of Artificial Intelligence (AI), we have prominent figures like Hinton and Harari who have spoken out about the need to curb the uncontrolled development of AI, also known as the "AI arms race." Meanwhile, there is a discussion about the possibility of applying AI in the field of robotics, and whether Long Language Models (LLMs) could represent a significant breakthrough in this area. Could the implementation of LLMs bring about the expected qualitative leap in robotics? What risks would need to be faced?
Amro: As soon as companies like Google and Amazon realized that truly useful AI models required substantial resources for training and that people would be willing to pay for remote access to these models, AI shifted from being a regular research field to a highly lucrative product for these companies.
Consequently, it is subject to market laws, and the intense competition to obtain the best model becomes inevitable.
Unfortunately, unless governments regulate this development, it will not slow down unless they find themselves facing an "explosion" caused by AI in their own hands. However, by that point, it may already be too late.
Regarding the use of AI in robotics, I must say that AI and robotics have always been like two good friends, with one admiring the other. AI lacks the hardware interface that robotics possesses, while robotics yearns for AI to make it less clumsy. However, their union is complex because robotics deals with real objects that generate real measurements, measurement noise, errors, vibrations, and so on. AI does not appreciate such indeterminacy, as it makes it imprecise and prone to many errors. Nevertheless, if we manage to eliminate all these imperfections and provide AI with "clean" data that it can handle effectively, its performance improves significantly.
For instance, if we use AI to recognize objects, label them with names, and locate their positions, we can pass that considerably "cleaner" information to another AI that can perform useful actions based on it. For example, if we train an LLM in natural language with a sequence of instructions on what actions to take when encountering a specific known object, we would have solved one of the major problems in robotics: planning. Moreover, these actions could be further subdivided into subsequent actions, creating a cascading effect.
If we describe all of this using natural language and train an LLM accordingly, we have a genuinely powerful tool for generating actions.
It wouldn't be far-fetched to think that Elon Musk's team is pursuing a similar strategy with their robot. However, when it comes to refining finer movements, this approach may prove somewhat limited. That's why I believe Elon Musk, leveraging the enormous technical and economic resources at his disposal, has opted for a cleverer strategy to enhance the finer motility of his robot: creating multiple robots that interact with a real-world environment and training AI models capable of handling data imbued with the imperfections of the real world. But, the most crucial aspect here is
that what one robot learns through its interaction with the real world, the others will also learn.
Consequently, the amount of real-world data learned is enormous. With the help of their dedicated AI hardware, the team can achieve quite good results in a relatively short amount of time. They can continuously update and test their model in a real-world setting, obtaining increasingly precise results. This iterative process allows them to adjust the model as many times as needed. If this holds true, we may find robots coexisting with us sooner than we think, performing repetitive manual tasks, transporting packages within warehouses, and stocking shelves. Additionally, professions like waiters, shop assistants, and cooks (among others) would see their job security threatened, unless all these changes are regulated at the political level.
[...] To be continued.
We hope you enjoyed these answers as much as we did, and we wish you a joyful and festive weekend!
If you enjoy this content and would like to support us in bringing you more, you can contribute with a symbolic coffee. This platform allows payments through PayPal. Ko-fi! Ko-fi! Ko-fi! ????