Author Interview: Machines of Tomorrow

In this exclusive interview, we delve into the visionary insights of Pedro, one of the authors behind "Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity," as he shares his profound perspectives on the interplay between humanity and artificial intelligence.

Authors

Pedro Uria-Recio, a Spanish AI expert and business executive from Bilbao, played a pivotal role in analytics and AI at True Corporation in Thailand and previously at Axiata. His career spans McKinsey consulting, entrepreneurial ventures with Antler, and contributions to the Forbes Technology Council, with recognition as a global innovator in data and analytics.

Randy McGraw, an American business executive and strategy expert, has led significant digital and business development initiatives across major corporations like DIRECTV, Singtel, and Amazon Japan. He is now an active investor and consultant in Southeast Asia and Japan. He was educated at Wharton and Cornell and is preparing for a Doctorate in Ministry.

Interview

TheAIML

Hello, Pedro. It is nice meeting you. Let's get straight to the point: Your book delves deep into the intersection of AI and humanity, exploring the profound implications of this evolving relationship. Could you share insights into what inspired you to embark on this exploration?

Pedro Uria-Recio

Thanks! It was nice to talk to you, too. I have co-authored a book about "Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity." It is for anyone who wants to understand and contextualise AI deeply. The book details the history and future trajectory of AI. It also uniquely pulls that foundational knowledge back into a thesis that humanity is on an evolutionary journey where carbon intersects silicon. The next several decades will present humanity with challenges and opportunities it has never had before. The analysis looks at the permutations of utopia, or conversely dystopia, that are likely to arise and explains why the likely landing spot is a mix of the two. The book is grounded in science but written in a way that is accessible to all. This book is a must-read for anyone interested in having a knowledgeable perspective when looking at what will unquestionably be a bumpy road ahead.

TheAIML

Let's begin with a light-hearted teaser question. Can you share some amusing myths about AI or anecdotes that have amused you? For instance, the notion that AI possesses emotions was prevalent in the early days. While it's theoretically possible, the reality is that AI currently lacks emotional capability. We may eventually reach that point, but it remains a distant prospect.

Pedro Uria Recio

That's quite intriguing because we delve into that topic in the book. Could you let me elaborate on it since it's relevant? We have a chapter titled 'Robot in Love,' which explores the concept of robotic emotions and their potential to form emotional connections with humans. Extensive research has been conducted into developing robots capable of empathy, and we've progressed to the point where they can interact emotionally with humans. What's more, it's posited that robots may even experience emotions organically without explicit programming.

When an AI algorithm is assigned an objective function—such as ensuring customer satisfaction in a call centre—it constantly evaluates its performance in achieving that objective. If it perceives a failure to meet the objective, the algorithm may adjust its parameters by retraining the neural network. This process can lead to the formation of 'shortcuts' in the neural network, which effectively manifest as emotions in the AI's responses. These emotional reactions stem from stress-induced retraining of the neural network, mirroring the workings of a biological brain. Thus, it's conceivable that AI could indeed experience emotions, with significant implications for its development.

This topic is extensively discussed in the book, highlighting the potential for AI to exhibit emotions. However, the prospect of AI possessing self-consciousness remains to be seen. Consciousness is a complex phenomenon that we have yet to understand from the AI perspective fully. Nevertheless, the possibility of AI experiencing emotions opens up profound avenues for exploration and consideration.

TheAIML

When you mention "possible," are you suggesting a timeframe of the next five to ten years due to advancements in AGI?

Pedro Uria Recio

Everyone who has attempted to predict the arrival of AGI or the widespread adoption of self-driving cars, topics we discuss in the book, must be corrected. We refrain from making predictions regarding timelines in the book. While we anticipate many advancements in these areas over the coming decades, we are cautious not to specify when precisely these advancements will occur because the truth is, no one knows. Additionally, AI experiences periods of enthusiasm followed by downturns, much like the crypto industry. With three AI "winters" already observed and likely more to come, predicting the timing of developments requires significant effort.

TheAIML

So, is the book structured similarly to the movie "2001: A Space Odyssey," or is it more reminiscent of "Wall-E"?

Pedro Uria Recio

Yes, "Space Odyssey," indeed. We discuss that book extensively; I delve into science fiction quite frequently. We specifically address the AI character HAL.

TheAIML

That robot was called HAL 9000.

Pedro Uria Recio

Indeed. We discuss science fiction extensively. We explore the future of AI, envisioning what it might entail. Our discussions draw upon various works of science fiction, including those centred on AI, quantum computing, and even the concept of romantic relationships with robots. Many authors and filmmakers have explored these themes, providing rich material for contemplating future scenarios. It's important to clarify that we're not conflating reality with science fiction; instead, we're using it as a lens to consider potential future developments.

TheAIML

Elon Musk frequently discusses science fiction, doesn't he?

Pedro Uria Recio

Yes, indeed. In that regard, science fiction does indeed influence reality. So, we delve extensively into science fiction. However, before delving deeper, please let me outline the book. The book revolves around two fundamental topics.

The first topic we explore is what we term "human-AI interlace." This concept encompasses the intertwining of humanity with artificial intelligence through cyborg technologies, brain-computer interfaces, and synthetic biology, which enable AI to manipulate DNA. For instance, as witnessed recently, Elon Musk received a brain-computer interface earlier this year or late last year.

AI is becoming increasingly intertwined with humanity through these advancements, leading us towards becoming a blended species over centuries, situated at the nexus of technology and biology. In the book, we argue that we will eventually evolve into one or multiple new species, and we delve extensively into this concept. Thus, the emergence of new beings, species, or amalgamations of technology and humanity constitutes one of the critical topics of discussion.

TheAIML

Is this concept distinct from the idea of a collection of expert algorithms? In other words, do a group of specialised mini-AI bots collectively power Gemini when generating responses to prompts across various chat GPT platforms, including those developed by OpenAI? Are you applying this concept here?

Pedro Uria Recio

Are you referring to reinforcement learning from human feedback utilised in training LLM algorithms?

TheAIML

No, that's not what I'm referring to.

The concept of a group of experts involves breaking down a given prompt into multiple segments. Each prompt segment is then assigned to a separate AI expert model. In this context, an "expert" isn't a human but a bot. These bots effectively serve as mini-AI models. Instead of relying solely on the capabilities of a large language model, each bot independently responds to its allocated segment of the prompt. Subsequently, the overarching large language model combines the responses from these individual bots.

Pedro Uria Recio

No, that's not what I intended to convey. However, in the book, we analyse the impact of those expert networks on the future of various professions, such as programmers and data scientists. We also mentioned that approach when discussing current research efforts towards artificial general intelligence or AGI.

But I was referring to the impact of AI on the human body. AI enables us to design DNA, allowing for the creation of proteins and artificial organs like hearts that can be implanted into the body. With AI, we can modify DNA to reduce susceptibility to diseases like cancer or even enhance abilities such as night vision. Similarly, advancements in technology enable the 3D printing of bionic body parts like hearts and arms.

In essence, AI technologies will lead to significant changes in the human body, ultimately resulting in the emergence of a new species in the long term. This transformation won't happen overnight; it will take time. However, it's the direction in which we're headed.

TheAIML

I want to ask from a different perspective when thinking about this topic. Do you envision the future we'll leave for our grandchildren as utopian or dystopian? In other words, what is your outlook?

Pedro Uria Recio

This is the second topic covered in the book: What lies in the mid-term future of AI? We refrain from pinpointing specific timelines like ten, twenty, or thirty years; it could even extend to a hundred years. It's uncertain. In the book, we delve into the positive aspects of AI, which undeniably exist. However, we also address the negative aspects, dedicating a section to their exploration. We posit that we'll find ourselves simultaneously in an AI utopia and an AI dystopia. The distinction lies in individuals' proximity to each scenario. Both will coexist, with some faring better than others.

We emphasise three critical aspects of utopia. First, considerable wealth will be created, albeit benefiting only a select few, alongside the implementation of redistribution programs. Second, environmental sustainability will be significantly emphasised. These are the positive aspects we highlight.

On the negative side, there are several concerns to consider. Firstly, there's the potential end of human work. This poses a significant problem as many individuals rely on employment for their livelihoods, and with it, economic survival may become more accessible.

Secondly, there's the issue of authoritarianism. As demonstrated by instances like Cambridge Analytica, AI's capacity for manipulation poses a threat to Western democracies. AI will contribute to eroding democratic principles, leading to increasingly authoritarian political systems.

Thirdly, there's the emergence of caste politics. As individuals become categorised into multiple castes based on their level of integration with AI, biological disparities will arise. Initially, these enhancements will likely be costly, limiting access primarily to the affluent. Consequently, we anticipate a biological divergence among different segments of society.

TheAIML

This will lead to concepts akin to "I think. Therefore, I am” to "I excel. as I upgraded," correct? It shifts from the classical notion of "I think, therefore I am."

Pedro Uria Recio

Indeed, those are some of the negative dystopian aspects of AI. Another notable phenomenon we've already begun to witness is the erosion of truth. Truth, as we understand it, will become increasingly elusive. This won't just be due to manipulation, which is more likely with AI, but also because of synthetic data. Synthetic data appears realistic but is essentially fabricated. It's not authentic data. Many companies are resorting to training, for instance, self-driving cars, using synthetic data because they need more accurate real-world data. Google has been developing self-driving vehicles for over 15 years, yet many newer entrants must rely on synthetic data. Numerous start-ups are emerging to provide synthetic data for training self-driving cars. When algorithms are trained on synthetic data, distinguishing between reality and fabrication becomes nearly impossible.

This year, the World Economic Forum discussed misinformation and disinformation extensively, which aligns with this issue. The Forum suggests regulating it to prevent misinformation and disinformation. However, I need to be more relaxed about exerting control in this area, as it could lead to censorship, which is undesirable. People should develop critical thinking skills rather than rely on regulation. Nevertheless, this chapter explores these dystopian aspects of artificial intelligence in depth.

TheAIML

Now, let's approach this from a different angle. When did you first realise that AI would become your passion? It could be a personal anecdote, a specific event, or something entirely different. When did you realise AI is more about extraordinary innovations than impending doom? What inspired you to view AI as the exciting and captivating field that it is?

Pedro Uria Recio

The first time I engaged with this type of algorithm was during my master's degree, where I enrolled in courses on data-driven marketing and advanced statistics. Much of what is referred to as AI today consists of logistic regressions or linear regressions. These statistical methods form the foundation of what we commonly recognise as AI.

TheAIML

Indeed. People often focus on correlation, causation, and simple linear regression techniques.

Pedro Uria Recio

Linear regression remains the most widely used algorithm in the corporate world and will likely continue to be so for the foreseeable future. That's when I started to become interested. While I excelled in engineering, math, and software development, it was during this time that my interest in AI began to grow.

Whether AI is excellent or doomed has both positive and negative aspects. However, in the book, we refrain from categorising things as inherently good or bad. When discussing the concept of interlaced evolution, which represents the next step in our development, it's not about whether it's intrinsically good or bad. I noticed your reaction earlier when I mentioned it—it seemed to evoke a certain feeling.

In the book, I deliberately avoid passing judgment on whether things are good or bad, particularly when it comes to human-AI interlace.

TheAIML

Some of the concepts you're discussing have a philosophical undertone. They remind me of the ideas explored by Greek philosophers.

Pedro Uria Recio

What you mentioned about Greek philosophers is quite intriguing. This book extensively delves into philosophical, mythological, and religious themes. The emergence of a new intelligent species—whether sooner or later—is arguably the most significant religious event in the past two million years. I devote considerable discussion to AI's implications for religion.

There's a growing interest in AI. In the United States, there's even a church known as the Church of AI, complete with a sacred text titled "Transphormosis," authored by Chat GPT in March 2023 and available for purchase on Amazon. While some may consider them eccentric, more people will likely begin to entertain similar ideas.

Suppose an IT system were to surpass humanity in intelligence. In that case, it's plausible that people would worship it as a deity, as anything superior to us would inherently be perceived as divine. Therefore, the book extensively explores the implications of AI for religion and spirituality.

TheAIML

Pedro, how can we ensure that our moral compass is integrated into algorithms?

Pedro Uria Recio

So, we have a chapter dedicated to this topic, specifically on regulating AI, particularly superintelligence, which refers to AI that surpasses human intelligence by a significant margin. If we consider Anthropic, for instance... Anthropic is essentially an LLM (Large Language Model) that operates according to a constitution. Every action taken by the LLM must adhere to this constitution. The book discusses this method in AI is discussed in the book.

Another approach is being pursued by OpenAI, which announced last year that it would allocate 20% of its budget to establishing moral controls for superintelligence. Their approach involves utilising automatic feedback mechanisms for this purpose.

Additionally, we explore another approach proposed by the founder of AlphaGo, later acquired by Google. He authored a book called "The Coming" last year, outlining his ideas for controlling AI. These are just a few of the techniques discussed in the book.

In the book, we analyse the advantages and disadvantages of these approaches. Another approach discussed is open source, which was proposed and strongly advocated for by Jan.

TheAIML

So, please give me an elevator pitch on precisely those approaches. I mean, keep it short. And where do you see which one will be most effective?

Pedro Uria Recio

This is a complex topic to condense into an elevator pitch, but here's my take. There are various approaches to addressing the ethical implications of AI. Some are purely technological, designing AI systems with certain principles in mind. Others focus on embedding limits in AI development through regulation, corporate controls, and activism. Then there's the open-source approach.

The book argues that open source is likely the most favourable path forward; we should strive to make AI open-source.

However, it's essential to acknowledge that there needs to be a foolproof way to ensure that superintelligence will share our values. Superintelligence may emerge as AI continuously improves and redesigns itself, leading to self-improvement cycles. This could result in the development of AI far surpassing human intelligence. Maintaining consistent ethical values across these self-improvement cycles presents a significant challenge. Therefore, we cannot definitively predict the values that superintelligent AI will adopt.

TheAIML

I understand. You've raised an intriguing concept with the idea of an open-source model of ethics. However, I see several potential risks associated with this approach.

Pedro Uria Recio

Yes, exactly. Gemini was unveiled last week, right? You mentioned seeing Gemini the previous week, and controversy surrounds it. However, it's worth noting that Gemini was not developed as an open-source platform.

TheAIML

So, here are the last three quick questions before we wrap up. Let's move on to the following question: If AI had to describe its rise, how would it do so? Could you keep it brief?

Pedro Uria Recio

That's a fascinating question. Throughout history, humans have been captivated by the idea of crafting artificial beings, spanning ancient China, pre-classical Greece, and India. This obsession persisted for centuries until they finally achieved it. Eventually, humans recognised that the path forward involved interlacing with the technology they had birthed. Essentially, it's a narrative of both creating AI and evolving into AI.

TheAIML

Absolutely. It captures the integration of the entire process of creating artificial entities with the progression of technology.

Pedro Uria Recio

Yes.

TheAIML

Essentially, merging with it and eventually becoming unified with it. That's a solid answer. Thank you, Pedro. The next question, which is the second-to-last, is quite amusing. If AI were to reflect on its teenage years, what would be the most embarrassing or memorable incident it experienced?

Pedro Uria Recio:

The first AI winter may be the most embarrassing event for AI during this era. In the 1960s, AI researchers predicted we would achieve artificial general intelligence within just a few decades, but this didn't materialise. Their over-optimism and, in some ways, arrogance led to disappointment as we remain far from achieving artificial general intelligence.

TheAIML

Well, we're farther than we once were, are we?

Pedro Uria Recio

We're progressing, but large language models (LLMs) still need to achieve artificial general intelligence.

Scientists' expectations regarding AI in previous decades were unrealistic. When the oil crisis hit in 1973, and the US government withdrew funding from AI, everything stopped until the 1980s. This setback could be considered embarrassing for AI.

TheAIML

Is it similar to the movie "Space Odyssey"?

Pedro Uria Recio

Yes, indeed. They anticipated swift progress, but it was quite the opposite. The movie was released in 1968, precisely during that period. It depicted artificial general intelligence in 2001, yet in reality, by 2001, we still needed to develop self-driving cars.

TheAIML

If someone were to consider reading this book, why should they buy it? What makes it worth reading? What sets it apart from others? And what is the essence of this book? Could you provide a brief pitch?

Pedro Uria Recio

The book is unique because it presents a distinctive perspective on the future of AI, which involves the concept of interlace, whereby humans merge with AI.

Additionally, it explains the past and future of artificial intelligence from a comprehensive standpoint. As mentioned, we discuss economics, technology, philosophy, geopolitics, and religion. Additionally, we delve into the mathematics of the algorithms using simple language, explaining how they operate. Our exploration extends to society, touching on dilemmas such as Japan's decision between immigration and humanoid labour. We approach the topic of AI from multiple perspectives simultaneously. AI represents humanity's greatest epic, and an epic cannot be fully understood from a single viewpoint. It requires consideration from various angles due to its far-reaching impact on everything.

TheAIML

Alright, that sounds reasonable. We're running short on time, so let's go ahead and conclude now if that's okay with you. Thank you for your time.

Pedro Uria Recio

Yes, Indeed. Thank you for the interview. AI is such an exciting topic. Staying informed is crucial in the ever-changing terrain of AI and machine learning, but before I leave, let me tell you something: The AIML newsletter has been a consistent source of insightful updates for me, and I wholeheartedly endorse it for anyone seeking to stay ahead in this dynamic field.

TheAIML

Thank you, and good luck with your book.

Explore the future of AI and humanity with "Machines of Tomorrow" by Pedro Uria-Recio and co-authors. Get your copy today.