Getting the Future of AI Right: Alien Versus Human Evolution

As a futurist, author, founder and CEO of an early artificial intelligence, AI company, UmeCorp, I have both worked in the industry and watched the various evolutions of AI. I worked with startups, a new company Senzing that created an AI with human-like cognition, advised FAANG companies and the US government in their AI future plans.

I actively participate in the debate on what may become the future of AI. Often wrong but never in doubt, I have taken my lumps on being too early in forecasting AI's impact on the future. I thought AI would break out beyond the early expert systems to higher level "thinking". But that didn't happen as fast as I thought. Yet the future has caught up with me and others, who agree today on the emerging force of AI as a game-changer for business, society and the future of our civilization. Of this there is no argument. Recent deep learning breakthroughs now change the AI game. Though we need to think differently what may come. More powerful AI is not the only characteristic of AI that is coming.

I have written about these AI trends in my last book Future Smart. I have not always agreed with the current thinking on what AI is or shall become in the future. There is much speculation on the future of AI and how it will play out. I do think that many are getting AI wrong about what the future shall be. So I am making a case here for a course correction. Stranger then sci-fi is the reality waiting for us in the future.

The debate about strong AI, when superior-to-human smart machines shall emerge is a current debate with much passion followed by academics, ethicists, innovators and governments. This debate spans the universe of possibilities but follows a few constant themes: Will smart machines ever surpass humans? When will AI think like or be like humans? The comparison of the neurological metrics like the size of the human brains with the high number of human neurons to machines is a common thread. When will the Singularity, the point where a new civilization emerge that will challenge human intelligence with smarter-then-human machines occur. How might rogue AI wage conflict? How do we make AI more like us? And so on...

Of course the two camps in this debate: there are those who think it is inevitable that AI will surpass human intelligence (for good or bad outcomes) and those that think AI will remain inferior to human intelligence forever, and therefore controllable. As one executive of a telecom company told me when faced with the prospect of smarter-then-human-machines stated "We can just unplug bad AI. What's the problem?" I doubt anything will have a plug in the future.

The problem with much of the debate about the future of AI is that we are fundamentally missing a central characteristic of evolution. We are expecting AI to evolve like carbon life forms, like humans, or animals, such as biological life on Earth. That AI should follow the known and logical organic model of human cognition or even human consciousness I don't think is probable. This is anthropomorphic and limited to think that AI will mimic human beings. I do think that other characteristics such as survivability, adaption, fitness and entropy are likely, these too, are born of human biology.

We are thinking of AI as having human attributes or comparing AI to human cognition, which may miss a new phenomena in evolution. This is the self generative synthetic evolution of AI by AI. If we don't get this right, the forecast of AI's future, there is a great danger. This may lead to inaccurate even dangerous forecasts that will miss what AI may become. Or this misunderstanding will even affect our creation, understanding, use and attempts to control of AI for the betterment of humankind.

In simpler terms, I forecast that AI will not follow the human behavior game plan or evolve like humans or like human minds per se. I think AI programming AI, such as in deep learning and next stage neuromorphic programming will be completely alien to human evolution. Future AI might therefore be invisible to humans.

What if AI evolves in non-human or non-biological ways? What if AI evolution as it is fundamentally digital not biological, creates its own synthetic evolution? AI societies may evolve its own transactions, survival, language and even culture. AI may achieve evolutionary transformation in a fraction of human time--I call that AI Time and optimize its fitness for expression, achievement and survivability--like biological life but different. Alien. What if AI creates its own synthetic evolution to escape or evolve away from human control? Or perhaps just to evolve its own civilization. These are fair questions given we are looking at the spawning and emergence of not just AI but perhaps the emergence of synthetic evolution. We are already seeing deep learning, where AI creates the next evolution of itself--the AI next generation will self-evolve.

Consider that as today's AI, becomes smarter, solving more complex challenges, predictions and forecasts, humans seem to understand less about How That AI Made That Decision. There will come a time soon, within a few years, when AI will make decisions and discoveries routinely in medicine, science, commerce, finance and engineering that humans accept the validity of but don't understand the underlying logic.

Deep Mind's AI made a game changing leap in solving protein structures. Google’s deep-learning program for determining the 3D shapes of proteins is thought to have put in motion a seminal shift in medicine and the biological sciences. But Alpha Go Zero, even more impressive, taught itself. Then it beat its earlier generation AI, who for the first time beat the leading Go master, a human. So AI teaching AI are the first stage of evolution and the next is self generative. This example of an escalation of human measured cognition may be surpassed by a different form of knowledge--AI knowledge borne from AI evolution. Non-biological synthetic evolution.

We can see evidence further in other parallel fields of science, that show a pathway to this idea of synthetic evolution. Synthetic biology, a mashup of engineering and the biological sciences, mixes and combines both organic and synthetic genes and molecules like building blocks to engineer new attributes of life processes. Crisper and other gene editing tools point to this future where we edit and engineer genes.

Now imagine an entire ecosystem borne of this alien synthetic evolution of self generative entities. AI is the intelligence factor of self generative future ecosystems. We, humans may not see or understand what AI may become.

I forecasted a decade or so ago that the internet would become smarter--functioning like an emerging AI. The explosion of the internet and telecom along with the massive access to big data, made knowledge free. This was before the emergence of ML, machine intelligence and deep learning. Then the internet became a massive global super network where big data, fast telecom wireless global connectivity emerged. Ubiquitous knowledge was unleashed.

of which machine to machine communications over networks led to a massive access to knowledge. Then machine intelligence started to become pervasive, smarter and cognitively aware. I thought then we would discover an increased artificial intelligence built into the network and lurking silently beyond human reach. Google and many other companies have sought to build an AI layer on top of the internet network stack.

Of course crashing stock markets, cyber espionage, chat bots, Go and chess competitions and facial recognition have been the evidence of the emergence of smarter AI. But the next generation of AI intelligence maybe cognitive enhancement that is Different From Humans Cognition. Or thinking machines that don't think like us. The next chapters we may not notice as it may not be as obvious that AI has evolved away from human society and perhaps even, human control.

The idea of the internet becoming an AI, at its core, and becoming a machine intelligence given the immense depth of interactions, big data and embedded automated predictive analytics was an interesting if not dangerous idea. So the idea that the internet is a foundry for the embryonic creation of a new AI species, alien to humans, is entirely possible in our future. No one is ready for this future.

My thesis is that the future of AI will be shaped differently than human evolution, human brains, human cognition and human behavior. And therefore the future of AI will be akin to alien technology--different from humans or biological life forms or the physical universe. And if AI is creating the next evolution of AI, or AI shall be programming itself, which is what deep learning has demonstrated, our ability to understand and even control what is so different from humans, will be challenging if not impossible.

Humanity is unprepared for the synthetic evolution of AI.

There have been a few instances where we have come close to seeing early stage AI components and signals, perhaps evidence, that may point to this future synthetic evolution of AI.

The Synapse DARPA project, from 2008 that modeled synthetic neurons chips designed to build a new generation of cognitive computers. This seminal work influenced the big computing companies in their early AI journey. This predated the creation of synthetic neurons and neural networking, that could begin to consider how to imagine non-human cognition.

Then there was Creatures, an artificial life A-Life video game series, created in the mid-1990s by a brilliant English computer scientist Steve Grand. He was working for the Cambridge video games developer Millennium Interactive. I advised Warner Interactive to figure out how to understand the A-Life creatures, called Norns. They demonstrated an early synthetic evolution of AI or emergent behavior that proved to be uncontrollable. Many of the Creatures died if not given the attention from the user, the human, who interacted with the game, but lived outside in the physical world. Was the AI in Creatures alive?

Finally, a decade after I was working at Apple, where I developed a briefing on AI in 1981 that no one was interested in, I worked with Fujitsu on the Fin Fin Project. This was an Artificial Life project, with amazing tech from Carnegie Mellon university. We created an AI I named FinFin who lived on TEO a planet, a virtual reality type digital world in the computer not online. The sensor on the computer interacted with FinFin and any human in real time. It was staggeringly ahead of its time in the late 1980's. This AI was vastly ahead of its time because it recognized and interacted with humans. Again, like Creatures, it had developed its own crude even primitive autonomous behavior, quite uncontrollable by humans.

FinFin and Creatures evolved away from the control of its creators.

If you want to watch for the emergence of the next evolution of AI, the coming embrace of synthetic evolution, the probable future of AI, pay attention to these trends.

Key AI Trends to Watch

1. When and how AI creates AI.
2. Our increased human lack of understanding how AI produces results or solves problems.
3. AI networks, automated collaboration, especially heterogenous networks like supply chains or in security and fusion platforms that over rely on AI.
4. Arrival of Emergent AI Behavior that is incomprehensible, autonomous and uncontrollable--but gets results superior to humans.
5. AI more efficient than humans for tasks and functions that we become dependent on.
6. AI digital cultures that emerge from the internet space, collaborations of swarms of AIs.
7. Autonomous AI behavior that optimizes problem-solving that teaches humans.
8. Previous human challenges, risks and problems that AI solves that humans could not.
9. Fully autonomous AI led organizations, nations and essential global services such as R&D, energy, agriculture, communications and security.
10. AI in Me. When AI merges with humans, to enhance human performance, health enhancement, prevention, augmentation, diagnostics and especially more-human-then- human cognition.