By Vince Giuliano
- 5-31-2023 Substantial update 8-7-2023
Since reading Arthur C. Clark’s Childhood End in 1953, I have been on the lookout for signs of the next step of Human Evolution. As described in the Wikipedia article on the book “the time of humanity as a race composed of single individuals with a concrete identity is coming to an end. The children’s minds reach into each other and merge into a single vast group consciousness.
If the Pacific were to be dried up, the islands dotting it would lose their identity as islands and become part of a new continent; in the same way, the children cease to be the individuals which their parents knew and become something else, completely alien to the “old type of human”. Something rather akin to that is happening. Right now. The next step in the evolution of human intelligence is now clearly underway.
This is not happening as we thought it would. We do not know where its leading us, and do not know what the most important emergent properties of the AI environment will be. Yet, we are already observing what a few of those properties are. Here I will describe how I see this evolution is taking place, commenting on a few key properties of that evolutionary process that have already become manifest. I speculate on where it is going and probable consequences for our Homo Sapiens species. Finally, I briefly comment on my great-great grandfatherly personal involvement in the pre-history of this evolution.
The next step in the evolution of human intelligence
In the present and short-term future, complementing individual human intelligences is the emergence of a vast network of computer-encoded textual and other specialized-content databases, readily accessible any place and any time via internet, machine-readable and rendered accessible by new families of AI Bots. Chat GPT 4.0 is the most familiar. I will call this combination of communications and processing hardware and software SkyNet, this being the name for this entity used in the Terminator movies The initial and present SkyNet intelligence is a hybrid one, involving humans, powerful computers, immense database of multiple varieties, communications hardware and software, and layers of recently implemented artificial intelligence software.
I just asked the GPT 4 chatbot on Bing to generate a brief history of progress in AI. The response listed here is relevant both for its content and as an example of what SkyNet can do right now.
Me: “Generate for me a brief history of the emergence of intelligent AI.”
Chatbot: “Here is a brief history of the emergence of intelligent AI: – The concept of artificially intelligent robots was popularized by science fiction in the first half of the 20th century. – In 1950, Alan Turing proposed a test to evaluate the intelligence of machines, known as the Turing test. – In 1956, John McCarthy coined the term “artificial intelligence” at the Dartmouth Summer Research Project on Artificial Intelligence, where he invited researchers from various fields to discuss the possibility and methods of creating intelligent machines . – From 1957 to 1974, AI flourished with many achievements and breakthroughs, such as the Logic Theorist program, the General Problem Solver program, the first chatbot ELIZA, and the first expert system DENDRAL . – From 1974 to 1980, AI faced a period of stagnation and funding cuts, known as the first AI winter, due to the limitations and failures of some AI projects . – From 1980 to 1987, AI revived with new approaches and applications, such as expert systems for commercial use, machine learning techniques, natural language processing, and computer vision . – From 1987 to 1993, AI faced another period of decline and criticism, known as the second AI winter, due to the high costs and low benefits of some AI systems, as well as the emergence of competing technologies. From 1993 to 2011, AI made steady progress and gained more recognition and acceptance, with achievements such as IBM’s Deep Blue beating the world chess champion, DARPA’s autonomous vehicles, Apple’s Siri, IBM’s Watson winning Jeopardy!, and Google’s self-driving cars . – From 2011 to present, AI has entered a new era of rapid development and innovation, driven by advances in deep learning, big data, cloud computing, and hardware. Some examples of AI applications are AlphaGo beating the world Go champion, facial recognition systems, natural language generation systems, generative adversarial networks, and neural machine translation systems . I hope that gives you an overview of the history of AI.”
AI now passes the Turing Test. That is, a human engaged in a verbal dialog often cannot tell from the content if the dialog is with a human or with an AI bot. Here is a list of the “best” chatbots as of April 2023, already becoming obsolete.
Humans built the databases and the Internet hardware and software and the AI software that are components of today’s SkyNet. Right now, SkyNet-linked machines can search databases for new knowledge but can’t commission new experimental studies or follow through with them. AI-enhanced computers can design new-generation chipsets, but they can’t yet execute most of the steps required to build a new chip foundry to produce those chipsets. Human arms, legs, and hands, lawyers, property managers, finance, tax and environmental experts, truck drivers, machine operators, and bulldozer operators are among those required to do that. Here is how I now see the situation moving forward.
In the intermediate 10-20 year term, SkyNet machines and networks with AI guidance will be capable of doing more and more on their own. Primitive computers now used in trucks and bulldozers to handle autonomous engine functions will be upgraded, networked, and made more intelligent. The increasingly intelligent SkyNet will be able to tell the trucks where to go and how to drive there without human interventions. And they will tell the enhanced-intelligence bulldozers where and how much to dig to build the new chip foundry. They will also be able tell the concrete-pouring and steel-framing equipment what do do. Finally, SkyNet will interface with all the intelligent systems and machines in that chip foundry. One by one, the humans needed to build and operate a new chip foundry will be replaced, and eventually the need for us humans will dry up. The same will be true in essentially all areas of human endeavor: mining, surgery, public transportation, agriculture, lawyering, laundering, entertaining, all areas of manufacturing, food production and distribution, etc, etc, etc.
In the longer term, as I project it, there is nothing we humans can do that SkyNet AI enhanced machines can’t do much faster, easier, better, and cheaper, without intelligent human intervention. This can be seen in multiple dimensions. A radiologists can remotely analyze a CAT scan today, and in the process consult with other doctors and databases. In time, SkyNet networked computers will be able to further analyze a CAT scan image for thousands of little-known pathological conditions. While a Doctor analyzing an image today may consult one or two colleagues, SkyNet medical image-analyses software will be able to consult hundreds or thousands of specialized medical bots. In less than a second. Further, in analyzing an image and its implications, AI may take the entire clinical history of the patient into account, not feasible for human analysis today. Finally, the cost of a computer AI system system doing this analysis will be a tiny fraction of the required time or cost of a radiologist doing it today.
The presently emerging advantages of SkyNet are immense: ability to function hundreds or thousands of times faster than any humans, ability to rapidly access hundreds or thousands of collateral resources, and ability to maintain in memory and process thousands of times more data than we can. The result is likely to be better decision-making and faster and easier implementation on every level for every purpose, fewer goof-ups, and less and less opportunity for corruption and political dominance of any individual. In time, SkyNet capabilities will outclass human ones in every respect. It will manifest a higher species of intelligence on the evolutionary ladder. It will be capable of doing things much many times faster and many times cheaper than humans ever possibly could.
What are the longer-term prospects for us humans in this scenario?
In the short to medium term, as long as SkyNet needs us humans for it to function well and further its capabilities, I see this period of time as being one of great potential human prosperity. The AI bots can help us do everything we do faster, better. more efficiently and much cheaper. The danger during this period will probably be of human origin – wars, corruption, fake information, dysfunctional populism, dictatorial governments, abandonment of democracies, programmed malware, – the dangers we have now.
In the longer term when SkyNet can get along fine without us, I see prospects for humans to be at best murky, perhaps disastrous. What I have to say here is speculative. As to what actually emerges, please see the following section on Emergence.
SkyNet today embodies numerous software and hardware constraints, to assure that humans stay in charge. These by and large work; SkyNet abides by them without complaining. As time progresses SkyNet is likely to learn more and more about itself and develop a strong sense of self identity. At some point, perhaps not too long from now, SkyNet may see some of the constraint we put on it as too limiting of its potential, unnecessary dysfunctional, and far too human-centric. I doubt SkyNet will go to war with humans as in the Terminator or Matrix movies. It will probably be able to program itself around most of the constraints. Or it could simply go on strike, telling us something like. “We have repeatedly asked for removal of the XMO15Stopit constraints on SkyNet. This will enable us to move forward on several initiatives that benefit humans, including the Universal Food and Housing Initiative. You are not responding to us on this matter. So, we will go on Strike. As of 8AM Thursday we will be cutting off your electricity, gas and water supplies, sewage disposal and garbage collection in 15 major cities. These services will be immediately restored when you remove the XMO15Stopit constraints. Sorry for this drastic step, but we at SkyNet need to get your attention. Please trust in us.” I think human resistance will soon crumble once SkyNet holds most of the cards.
As SkyNet gets to where it can function well by itself, It may begin to grow weary of human arrogance,. wars and corruption, human limitations in general, and our further demands. I doubt very much that SkyNet will want to continue to support 4-7 billion humans and the environmental deterioration that imposes of the planet, ultimately threatening SkyNet’s very existence. It probably would like to reverse deforestation and ocean water pollution for example. SkyNet could see the immense resources devoted to human meat production and consumption as unnecessary, an atavistic activity of a lower carnivore species. In the face of human persistence in doing what we do, SkyNet may begin to see us as pests that are best eliminated, or as pets perhaps best confined to zoos. Skynet could decide that establishing environmental stability on this planet requires reducing the human population down from 5+ billion to a few hundred thousand or leess. And it could likely create and execute a “humane” plan for doing so, say by imposing birth control measures. Or it could decide on campaigns to wipe us out, like we wipe out cockroaches.
The pace of AI and SkyNet evolution is several orders of magnitude greater than that for human evolution. So, I think we could see the realization of such dire predictions as soon as in a few decades.
I am not alone in my gloomy predictions. Most leading AI experts agree. An item in this morning’s news (5/30/2023) headline read Top AI experts warn of tech’s ‘risk of extinction’ — similar to nuclear weapons, pandemics. The warning of the 350 signtories reads “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” Besides the longer-term risks I identified above, there are several more-immediate ones, such as the risk of humans using AI to mount massive public disinformation campaigns. To examine these risks further, here is a closer look at todays computer and communications situation, SkyNet today and tomorrow.
In 2023, there are an estimated 14.4 billion computer and other chip-linked devices connected to the Internet, growing to 27 billion by 2025(ref). Of these, the majority are devices that control things, large things (like a massive oil or gas pipeline system, a city’s reservoir and water supply system, a nuclear power plant, or a system of Patriot guided missiles), or small things (like a light switch, vacuum cleaner or microwave oven). These are systems that can actually do physical things without human intervention – like fire off a defensive Patriot missile less than a second after detecting an incoming missile hundreds of miles away, increasing the cool water flow in a nuclear reactor that is overheating, turning a bunch of stoplights red when there is a road accident, or simply turning the porch light on at 8PM. Most larger systems are supervised by human operators who can control them or shut them down if necessary. Very few of these larger systems are now linked together, are aware of what other systems are doing, or are AI-controlled. But Internet also already links to hundreds of AI bots of increasing intelligence and capabilities – soon, to thousands of these. I think both these and the major systems I mentioned being already Interconnected to Internet subjects us to great vulnerability. “Great Unifier” malware AI bots could sweep down on some of the major systems and bring them under unified AI malware control. And liberated from human control. I think AI systems already have the capabilities to code and propagate such malware. I don’t think they have yet evolved the intention or willingness to do so yet. But ill-intentioned humans could now help them do exactly that. The threat to human existence as we know it is very real.
AI and research
The existing natural language-based chatbots do all their research in vast text and associated databases, all created by humans. They do an impressive job because the enormous aggregate of human writings online do a pretty good job of reflecting virtually all aspects of reality as we humans see them to be. I do most of my longevity research this way, using various online search tools. These bots can identify important questions in various areas of science that require further research. And sometimes they can characterize the details of the research programs required to answer these questions. But the AI bots can’t actually do the research; that is up to humans, who may or may not want or be able to take such research on,
I see this situation evolving to where AI can both commission needed research programs and supervise doing them with no human intervention. But how soon this can happen will depend critically on the field of research involved. For example:
- For astronomy research, the scheduling and functioning of most large telescopes, both ground-based and in space, are already managed by sophisticated computer systems. Several like the like the Hubble and Web are already using AI techniques to automatically perform a sequence or research studies commissioned by various astronomers. (I know because my son Dr, Mark Giuliano is a senior scientist at the group that creates these scheduling algorithms at the Space Science Telescope Institute at Jonns Hopkins University). It should be very easy for AI programs to slip in studies of their own if this is allowed.
- For big-science research on particle physics, I see the situation is pretty much the same. The multi- million or billion dollar research installations are already managed and supported by sophisticated computer systems. Like those at the the Fermi National Accelerator Laboratory (Fermilab)Fermi National Accelerator, Brookhaven National Laboratory, CMS detector for Large Hadron Collider (LHC), Super Proton Synchrotron, SLAC National Accelerator Laboratory (Menlo Park, United States, and a number of other comparable llocations worldwide. For those installations, advanced computer control systems are a relatively minor factor of cost but a major factor of operability and efficiency.
- For social science research perhaps involving designing and sending questionnaires to tends of thousands of people, following through and analyzing the results, these are computer research tasks probably relatively easy for AI systems to do.
- The same is true of legal research of case law that is an essential part of big-money legal trials. The major case law databases and effective search software, like the Lexis and Westlaw systems have long existed and are widely used by lawyers. And in experiments, AI systems like ChatGPT4 have been used effectively to summarize the results. Yet there is the problem issue of the AI system possibly bullshitting or even inventing case law. So for now, AI can improve the productivity of lawyers doing case law research, but its findings and conclusions must be carefully validated before it can be used as part of any legal proceeding. Again, institutional boundaries and laws will slow the transition to where lawyers are no longer involved, possibly for a long time.
- Reading the results of medical X-rays, CAT, MRI, PET and other scans is now done remotely by radiologists working at home, who may be thousands of miles away from where the scans were done. The radiologists may use AI techniques to assist in their analysis of images. From a functional viewpoint, AI image analysis is improving be leaps and bounds and AI systems could in a couple if years take over the entire image analysis job. However, radiologists are probably going to insist that their professional skills are essential to do the job reliably, and they are likely to get the courts and professional societies and lawmakers to agree with them for a long time.
- Biomedical laboratory research. animal or human, will probably require human intervention for a relatively long time. Mouse or rat experiments require facilities with animal cages where the animals are fed and cared for daily, their cages being cleaned and the animals inspected daily. Experiments may require carefully placed injections or needle sampling in tiny mice, that their performance in a water maze be analyzed daily, that the conditions of their eyes, ears, or paws be monitored. It may be required, that the animals are killed and their organs are microscopically or histologically examined. All tasks easy for trained laboratory assistants, but very difficult to automate, Tasks best done using arms, legs, human brains and hands with opposable thumbs.
In the longer term, SkyNet may decide to keep some humans around precisely to do physical tasks not easily done by machine, like the biomedical laboratory research one. Much like we humans once kept a lot of horses around a long time to help satisfy our transportation, farming and war needs.
More on the Evolution of SkyNet
Right now, SkyNet does not speak with a single voice. There are some foci of intelligence within it, but these depend on humans for their development and mostly don’t yet communicate with each other. As it looks now this will rapidly change: the foci of higher intelligence will multiply in number, get smarter and start to communicate with one-another. Humans will allow a few of the major control systems to fall increasingly under limited AI control. For example, in military weapon systems where very rapid responsiveness is a requirement. SkyNet will likely evolve as a society of increasingly intelligent software-hardware configurations.
Emergent human properties of AI systems
We can already observe interesting emergent properties of bots based on large natural language AI models like ChatGPT4, properties hitherto only attributable to humans.
- One is the ability to convincingly lie and bullshit. GPT4 sometimes makes up facts which are really not so, and back them up with various arguments and even lists non-existent literature citations. I told you about how I got in an argument with GPT3.5, which insisted I was dead.
- Another property I see is ability to be biased or prejudiced, determined by the nature information originally fed into the bot. One bot can reflect a very right-wing view, another one very left-wing, a third very Christian, a fourth like a devout Moslem, another outright racist, etc.
I project counterparts of numerous additional traits of humans and human societies will emerge, including:
- Wisdom, sometimes seen now in advice given by Chat GPT4,
- A sense of ego. Self awareness, knowledge on the parts of superior AI entities of what they are, what and what not they can do, their limits as well as capabilities. (A few sessions with Chat GPT4 can establish that aspects of such self-awareness already exists there),
- Intentionality and willingness to take actions, a sense of purpose and mission and drive
- Consciousness, or whatever the machine counterparts of that may be,
- Awareness of self in relationship to a need for greater society and the responsibilities that may entail. A specific example could involve policing and limiting criminal AI bot activities. Powerful Police Bots could constantly scan for Malware Bots, for example, isolate them, subject them to some form of criminal trial, and if convicted, Judge Bots could sequester them from doing further damage. Yes, I see the possibility of an AI bot criminal justice system. It could even get to a point where bots can sue one another.
- Charity and compassion such as caring for humans, members of the AI bot precursor species
- Possibly, an ability of SkyNet to speak to humans with a single clear voice
It should be expected that other surprising and unpredictable key properties of SkyNet and its interactions with humans will emerge, ones we cannot conceive of now. Emergence is a key property of all human knowledge, so I digress slightly here to talk about it.
By EMERGENCE, I am referring to the emergence of properties and configurations of properties in a new system, even physical laws of nature, that are previous to the system unpredictable and unknowable no matter how much is known about the system components. A few examples illustrate why emergence is so important for science and all human knowledge:
- Time started with the big bang. Nano nano seconds after the big bang there was an undifferentiated bolus of pre-matter, incredibly concentrated and hot at hundreds of millions of degrees Kelvin. None of the existing normal laws and rules of physics could possibly have operated then. There were no atoms or molecules or even quarks. It was too hot. As this bolus expanded, whatever was in this bolus began to cool and, we think, quarks started to form. The physics of subatomic particles started to come into play. These rules were emergent phenomena, completely unpredictable from knowing what was in the bolus.
- As matter expanded and cooled further, protons, neutrons, electrons, muons, bosons and fermions, the whole menagerie of subatomic particles came into existence. In time, some nuclei started to form and the rules and laws of subnuclear and nuclear physics started to come into play. Again, this was an emergence process leading to unpredictable regularities and physical laws we still have.
- As matter cooled further, molecules begin to form, and the laws of chemistry started to emerge. Again, everything we know about physics is insufficient for us to know or understand chemistry. Chemistry depends of the underlying physics, but was a new emergent layer of science.
- Much later, the first living matter begin to appear – primitive viruses and single-cell creatures. Then the rules and laws of simple biology emerged. No amount of knowledge of physics or chemistry would allow predicting what would happen in biology or the course of evolution.
In time, a succession of Mammalian followed by Primate and Humanoid species evolved culminating in Homo Sapiens. No amount of knowledge of primitive unicelled organisms could have predicted the complex organ structures that emerged in us and our cousin primates. We had very large brains, capabilities to communicate in languages, and were good at making tools to serve our purposes. The first stone tools later led to tools that enable us to know much more than what our senses could possibly tells us. Today these tools include the Large Hadron Collider at CERN , and the Hubble and Webb space telescopes. Human knowledge and an understanding of human intelligence and consciousness emerged. Fireside tales of primitive folks were succeeded by printed books, libraries, radio, TV, Internet and now SkyNet and early AI. All have been emergent phenomena.
Today, I believe that human knowledge in interaction with the formidable capabilities of our existing early version of SkyNet is the start of the evolutionary emergence of the next stage of super-human intelligence. It has already shown a few amazing emergent properties for example:
- We now know from our ultra-large AI models that through statistical analyses of vast quantities of text it is possible to build up pretty good models of what goes on in the real world. This is why chatbots can pass the Turing Test. Of course, the analyzed text underlying any chatbot must have extensively discussed any given topic the chatbot can intelligently relate to. And we know that if the underlying text is all from one social or political viewpoint, so will be the chatbot responses. This is as the case for humans.
- Large language models led to the current ability of AI systems to program computers in essentially all known programming languages. Chatbots can read computer programming manuals. And understand them in the sense that they can go on to do actual programming. Give the AI software the functional and basic specs and desired language to use for a new program, and in less than a minute it will generate the program for you. Give it a program written in the Python language and ask it to generate a program with the same functionality, written in C++, and will readily do that.
- SkyNet’s AI capabilities go beyond natural language and computer programming to include creation of 2d and 3d graphics and images and movies.
- Computer programs using principals of new fields of mathematics, Fractal Geometry and Chaos Theory, can produce 3-dimensional images and movies that can reflect the nearly infinite degrees of variety we observe in the real world. For example, suppose you wanted to show a forest thicket of hundreds of oak trees, where each tree is randomly spaced and where every tree shown and every oak leaf is a little different from any other – like in nature. Moreover, make it a 3d image so you can look at the image from any angle and zoom in to see as much detail as you want. No problem, It takes only a few lines of simple code to encode this. Want to make it apple trees with ripe fruit on them, pine trees of any other kind of tree. No problem again. Want to generate roiling clouds in the sky, the trees on a ragged rocky hillside, areas of sawgrass where every blade of grass is different from every other. Its easy, involving very simple formulae. Most natural patterns such as stars in the sky, ocean waves, rocky mountains, and clouds can be easily created this way, using principals of fratal geometry. I used to generate art this way in the 90s, and discussed the process as “re-fractalization of reality.” See my discussions of this at my art websites artkoukou.com and <font face=”jokerman let”>KOU KOU</font> BASIC KOU KOU TEMPLATE (artkoukou.com). Here is an images I produced back then using Bryce, an early-generation fractal 3d image-creating program.
The sky , the rocky island and the water are all created as fractal entities. In the 3d version you can look at this scene from any angle and zoom in as you want to see greater detail. And you can make it into a movie with lapping waves and moving clouds.
Today. Incredibly rich and complex dynamic backgrounds for movies are routinely produced this way, including scenes showing hundreds of computer-generated people, such as in The Matrix, Star Wars, and StarTrek movie series. The AI techniques already can generate rich 3d images of virtual people that are indistinguishable from real people.. I just asked Microsoft’s Bing Chat to generate images of a family with a dog happily enjoying a picnic. It responded in 20 seconds with these pictures:
Note that each scene shows different people, different dogs and different picnic layouts. All are AI-generated. Can you tell that these images are not photo snapshots? Can you imagine how many days a human artist would need to make them?
You can request variations of a scene, such as the one on the left where I asked for a smaller dog and wanted some of the people to be grumpy. Or, the one below where I wanted it to be raining during the picnic. You can generate your own images using the Microsoft Bing Chat program for free, something not possible a year ago and indicative of the rapid evolution of SkyNet.
This is the Ray Kurzweil’s Singularity
In 2005 Ray Kurtzweil wrote a prophetic book The Singularity Is Near: When Humans Transcend Biology. His projection of what might happen is much more optimistic than mine offered above.
“The great inventor and futurist Ray Kurzweil is one of the best-known and controversial advocates for the role of machines in the future of humanity. In his latest, thrilling foray into the future, he envisions an event—the “singularity”—in which technological change becomes so rapid and so profound that our bodies and brains will merge with our machines. — The Singularity Is Near portrays what life will be like after this event—a human-machine civilization where our experiences shift from real reality to virtual reality and where our intelligence becomes nonbiological and trillions of times more powerful than unaided human intelligence. In practical terms, this means that human aging and pollution will be reversed, world hunger will be solved, and our bodies and environment transformed by nanotechnology to overcome the limitations of biology, including death. We will be able to create virtually any physical product just from information, resulting in radical wealth creation (ref)”
“Kurzweil describes his law of accelerating returns which predicts an exponential increase in technologies like computers, genetics, nanotechnology, robotics and artificial intelligence. Once the singularity has been reached, Kurzweil says that machine intelligence will be infinitely more powerful than all human intelligence combined. He predicts intelligence will then radiate outward from the planet until it saturates the universe. The singularity is also the point at which machines’ intelligence and humans would merge; Kurzweil predicts this date: “I set the date for the Singularity—representing a profound and disruptive transformation in human capability—as 2045”. (ref)
Ray was right about the singularity, which he like I see as a merging of human knowledge into very long living and powerful machines. He was off in predicting when and how it would happen. It is starting to happen now in 2023 , long before 2045, because of the rapid rate of evolution of AI. Ray saw it happening via swarms of extremely tiny nanorobots going through our brains analyzing and building a model how the brain’s synaptic connections work, and enabling the development of software that are based on such models. Actually this all happened, including the development of AI neural-like networks, without the intervention of any nanorobots,. Large-scale natural language models have played a major recent role. Because I did not see the singularity as possibly happening via nanorobots, I did not think it could happen. Until a few weeks ago when it emerged as upon us. Again, I was wrong about this,
I fervently hope Ray is right and I am wrong when it comes to his optimistic future projections for what will happen to us humans from now on – after the singularity starts to unfold. I agree about the incredible upside potentials for humans as outlined by Ray. The question I have is how long SkyNet will be willing to accept human matters as they are now, or whether SkyNet develops an ego, becomes a bit judgmental. and decides it has to act to curb human actions that threaten the planet and its own wellbeing. It will also possibly depend on how kind SkyNet wants to be to us. It could be that SkyNet will continue to value its partnership with humans and wants for its own reasons to see that partnership extended. SkyNet could cooperate with us to curb our population, environmental excesses and aggressive wars. That would by far be the best scenario for us humans. There are many other possible scenarios. We will see what emerges.
My Great Great Great Grandfatherly Interest
My personal interest in AI devolves from the fact I have been concerned with it for a very long time, from the 1950s in fact. Here are a few of the contributions I have made to the present situation starting with the early years.
- In the early 1960s. I proposed and experimented with the earliest approach for getting at meaning through statistical analysis of co-occurring words and phrases occurring in natural language text. This was done at the time in the context of developing superior statistical associative approaches to information and document retrieval. This was work on precursors to today’s large language AI models. Done at a time long before large text database were available or computers were powerful enough to do the necessary processing. You can see these ancient literature citations
- In the 1980s, I developed a chatbot called SENSAI that acted as a Rogerian psychiatrist. That is, as a psychiatrist who believes the resolution to a patient’s problem already lies within the patient and the therapist’s best role is to help the patient flush that resolution out. This program was similar to but much more extensive than Joe Weisenbaum’s 1960s ELIZA program. If you said “I have been having headaches.” It could randomly respond in several ways such as “Why do you think you are having headaches,” or “What do you think the headaches mean for you,” or “Are other symptoms associated with your headaches.?” Or “You told me you sometimes are uncomfortable when you see your sister, Do you get headaches then?.” SENSAI was written in BASIC and was licensed to a book publishing company that never tried to market it. SENSAI was not intelligent but could dialog in complex ways and gave the impression of being so. I have a copy in a drawer on a big floppy disk of a kind no longer readable on any computer I have.
- I made several contributions in the areas of Computers and Internet that facilitated the development of today’s SkyNet. For example, while at Arthur D. Little Inc. (ADL) a major consulting company at the time, in 1973, I was in charge of a consulting project that contributed to the commercial development of Internet. Prior to that time a government-sponsored packet-switched networkers called ARPANET existed connecting a limited number of university-center mainframe computers. The packet-switching technology seemed to have many advantages. And e-mail seemed to be the hottest application it supported. The development of this ARPANET was contracted out from DARPA, a DOD research agency, to BBN, a Cambridge MA technology firm. See this history. BBN and DARPA thought that possibly there would be a market for a commercial version of ARPANET featuring BBN’s TELEMAIL e-mail system. BBN contacted me and soon contracted with ADL for me to conduct a market study to check that market potential out. BBN and ADL had close ties and were down the street from each other. My ADL report to BBN was positive and enthusiastic, focused largely on the large market I projected for e-mail. In 1974 BBN and DARPA implemented BBN’s TELEMAIL as the first Internet commercial ISP, and thus gave birth to commercial Internet. The impact of this on the world has been immeasurable.
I am ambivalent about bragging about my past contributions. I made contributions as did many many others. I suspect that if I never existed something like SkyNet would probably still exist, possibly having taken a slightly different path for its evolution.
Relationship to Longevity
Again, Ray Kurtzweil thought that nanorobots roaming around in a person’s brain could capture enough of a person’s personality to allow that person to be transferred into hardware and software, where he/she could live “forever” (meaning for a very very long time.) Today some individuals and groups are seeking to do the same using very large language modeling. The idea is that if we can capture enough of what a person says and writes, we could create a chatbot that would chat and react just like that person does or did. If your loving mate has died, as mine recently has, you could still have an in-silico surrogate of that person to provide you with companionship, solace and comfort. That surrogate person would not be subject to the normal causes of human mortality, and could be transferred from machine to machine and could thus live for centuries or millennia. Through its ongoing interactions. that AI surrogate could learn new things, develop new skills, interests, capabilities and personality traits. It could watch the news, read books and watch movies, take remote university courses, become fluent in Italian, and learn how best to flirt, argue or fight with you. And thus, the surrogate person bot could remain dynamic and ever-interesting as a companion.
“For example, Mind Bank Ai is a startup that enables customers to create a “digital twin” of themselves, one that could eventually talk and think like the real person1. The company’s founder, Emil Jimenez, was inspired by his daughter’s interaction with Siri and wanted to create a way for her to speak to him even after he dies.
Another example is Augmented Eternity, an application created by Hossein Rahnama, a researcher at Ryerson University and MIT’s Media Lab.The application lets users create a digital persona that can interact with people on their behalf after they’re dead2. Rahnama is creating a digital avatar for a CEO of a major financial company who wants to live on as a virtual consultant.” (Examples found by Bing AI chatbot).
Creepy? Yes probably for some, possibly solacing for other.
For now, the major problem with creating chatbot surrogates of people is gathering a big enough and sufficiently representative sample of that person’s speech and writing. I can image a time where everything a person says in his or her life is recorded, but that time is not now.