Part three of Anticipating 2025 will summarize the third section of the book. This section focused on redesigning artificial intelligence, with a look at six important questions and the exploration of human-machine mergers. The six questions explored in this section are:
- Can we create a human-level artificial intelligence?
- If so, when?
- Will human-level artificial intelligence lead to super-intelligence?
- If super-intelligence arrives, will we like it?
- Can we upload our minds to computers?
- Can we de-risk the arrival of super-intelligence?
Like the first two sections, this section forces us to look at disruption through a different lens. Granted, the path forward is highly speculative, and even the most optimistic scenarios are likely years away from having transformative implications. Nonetheless, it does force us to broaden our lens beyond traditional views. For example, I’ve focused on the automation of knowledge work and all its ramifications, while the authors (Calum Chace, Martin Dinov, and Elias Rut) focus on creating super-intelligence by uploading our minds to computers. They explore a human-machine merger that they see as the enabler of super-intelligence benefits realization. This merger in the author’s view is the only way to avoid creating our successor. So yeah, that’s a little more impactful than automating knowledge work.
What we have today is a limited version of Artificial Intelligence, or what is referred to as weak or narrow AI. Per Wikipedia, this version of artificial intelligence addresses specific types of problems and is not intended to display human-like intelligence. Examples include virtual assistants, caretaker robots, virtual partners, game bots, and AI systems like IBM Watson, which are state-of-the-art question answering systems that understand natural language. According to David Senior, author of a recent post on Narrow AI, there are an estimated 170 startups ready to recast themselves as AI companies. He sees AI leading the list of mega-trends this year, tied closely to the growth of big data. The need for narrow AI is driven by explosive data growth, including user-generated content in the form of 300,000 tweets, 220,000 Instagram photos, 72 hours of YouTube video content and the 2.5 million pieces of content shared by Facebook users every single minute.
The book however is not focused on what AI is likely to deliver in the next decade. Rather, the authors look further into the future at the emergence of two more radical forms of AI. The first, Artificial General Intelligence (AGI) is an AI that demonstrates human-level intelligence, and can successfully perform any intellectual task that a human can. The second, Artificial Super Intelligence (ASI), is the successor to AGI, and represents an entity which possesses intelligence surpassing that of the brightest human minds in all respects. The authors posit that the arrival of super-intelligence will determine the future of humanity. If you buy into this notion, and the author’s belief that we cannot stop the march towards super-intelligence, then you may also agree with the author’s concern regarding the super-intelligence ability to harm us either deliberately or accidentally, and their assertion that not finding a way to upload our brains would be tragic.
The authors explore the feasibility of uploading, and the preservation of our identity after uploading. Assuming the original brain is not destroyed, do you now have two of the same person? These questions underscore the critical need to understand the brain before realizing super intelligence. The authors believe that by 2025, the Human Brain Project will have mapped human brains in very great detail. This large scientific research project is based in Switzerland and largely funded by the European Union. It aims to simulate the complete human brain on supercomputers to better understand how it functions. In effect, the project is focused on reverse engineering an existing brain through destructive analysis, or non-destructive analysis using increasingly advanced scanning technologies. The EU-funded Human Brain Project announced an important collaboration with President Obama’s BRAIN project.
When is all this likely to happen? Ray Kurzweil believes that AGI will emerge by 2029, with brain uploading coming 16 years later. Various other studies have focused on the timing of AGI and ASI realization. This article summarizes the combined results of four AI researcher surveys. The key findings were:
Source: Nick Bostrom, Director of the Future of Humanity Institute at Oxford University
The second topic within this section of Anticipating 2025 is the exploration of human-machine mergers. As the exponential progression of technology continues, the authors imagine that the power of Watson-like systems will eventually be available as an app on our phone. Since these systems are likely to provide responses faster than we can process, they see humans as limiting factors. The authors explore the feasibility of improved intuitive and thought-based interactions versus external physical interface devices. They ponder what would happen if we could use these capabilities by simply thinking about the desired steps. One simple example involves an intracranial device running a scheduling application. The application picks up a thought about a meeting on Tuesday and stores it reliably. When thinking about plans for that week, the system sends the right signals to your brain to stimulate memory of that meeting.
The authors envision a time when we monitor and cause very specific complex brain changes. Implants could replace non-functional brain parts impacted by various ailments, such as stroke, aging, or traumatic brain injury. These implants could go beyond mere replacement to deliver advantages such as increased lifespan, reliability, or speed of processing compared to the corresponding biological brain areas. This merging with advanced AI systems could enable humans to be more rational, more efficient, and correct in thinking. The authors go on to explore the ramifications of such a merger. For example, what if you could replace irrational thought processes with superior thinking? Does it cause us to reconsider previously-held beliefs? Post-merger, we may come to moral, social, and philosophical conclusions so far removed from traditional norms that we could cease to be whom we were before. This human-machine merger is sure to have profound cultural, societal, and historical effects. However, the authors find it unlikely that any kind of direct merger can happen without a sufficient understanding of the brain. Any human-machine integration via direct neural interfaces, nano-machines, or any other technology, must first understand how to interface with the brain.
There is a less radical path that AI may take. If these AGI systems continue to be bound to huge external computing clusters, they are likely to be non-personal Google-like systems that serve as public question and answer systems. However, advances would enable the AGI to focus on answering more difficult, interesting, and computationally-demanding questions.
That wraps up part three of this look at 2025 based on the book Anticipating 2025. Part four will focus on re-designing society. The first posts in this series can be found here:
Futurist contributing to the book include: David Wood, Mark Stevenson, Rohit Talwar, Calum Chace, David Pearce, Sonia Contera, Natasha Vita-More, Anders Sandberg, Ben McLeish, Amon Twyman, Iva Lazorova, Maneesh Juneja, Peter Morgan, Martin Dinov, Elias Rut, Zolton Istvan, David Levy, Andrew Vladimirov, Michael Nuschke, Alex Zhavoronkov, by Riva-Melissa Tez, Victor Anderson, Jerome Glenn