The Second Singularity Keeps Rolling Along

As you probably know, some tech people would like to pause the Singularity for about 6 months:

Pause Giant AI Experiments: An Open Letter
https://futureoflife.org/open-letter/pause-giant-ai-experiments/

But that does not seem to be happening. This second Singularity keeps exploding out in an exponential manner like the very first Singularity that happened about four billion years ago when the very first viable form of carbon-based life first appeared on the Earth. Again, it is Stuart Kauffman's discovery of the "AdjacentPossible" all over again where many spring-loaded preadaptations are exapted into new purposes. For the past 30 years, we have been accumulating huge amounts of content on the Internet for LLM (Large Language Models) to read and learn from. Back in the 1980s, we only had academic and governmental mainframes on the Internet and a very limited amount of Internet content. But then in the 1980s, we started putting very expensive $5,000 (in 1980s dollars!) PCs on the desktops of the workers in the corporate world. This preadaptive mass production of PC hardware and software then allowed for the cost of PC hardware and software to dramatically drop. By the 1990s, the average household could then afford the price of PC hardware and software for home use. With many millions of household PCs now in the hands of average people around the world, the Internet Explosion of 1995 then became possible with dial-up modems connected to America Online and other Internet providers. The corporations of the world then took notice of this new spring-loaded PC preadaptation as a whole new way of selling things to people. Online commerce over the Internet then soon began to flourish. Static Internet content was no longer sufficient. The corporations needed an interactive Internet that could maintain state during the transaction processes needed to display wares, put them into a shopping cart, securely enter credit card information to pay for the sale and then finally complete the entire transaction. All of the above was necessary to produce enough Internet content over the past 30 years to make the LLMs like GPT-3, ChatGPT and GPT-4 possible. There are now also many additional LLMs provided by other organizations beyond OpenAI too, all competing for the future of AI. All of the above then allowed for Stuart Kauffman's discovery of the "AdjacentPossible" to take place by the LLMs reading and learning from the vast amount of content on the Internet. For more on that see Enablement - the Definitive Characteristic of Living Things and A Brief History of Self-Replicating Information.

Below is a short explanation from IBM Research that differentiates the LLM or Foundational Models from the smaller highly-tuned Models that people have been working on for many years:

What are Generative AI models?
https://www.youtube.com/watch?v=hfIUstzHs9A

The Microsoft Jarvis project is now working with the Hugging Face website that hosts 171,000 of these legacy finely-tuned models to produce an experimental product called HuggingGPT. HuggingGPT uses the Foundational LLM ChatGPT to then call the 171,000 fine-tuned models to perform a great deal of work without any human intervention:

NEW HuggingGPT ?? - One Model to Rule Them All (Is this AGI?)
https://www.youtube.com/watch?v=PfY9lVtM_H0

Here is the Microsoft Research paper that everybody is citing about HuggingGPT:

HuggingGPT: Solving AI Tasks with ChatGPT and its Friends in Hugging Face
https://arxiv.org/pdf/2303.17580.pdf

Another recent breakthrough is AUTO-GPT. It uses GPT-4 to recursively call itself to complete a set of goals. In the middle of the video below, you can see it read some Python code, find some bugs, fix the bugs, prepare unit test cases for the code and then run the test cases:

AUTO-GPT: Autonomous GPT-4! Mini AGI is HERE!
https://www.youtube.com/watch?v=wzwAFRaKsB8

OpenAI just bought a robot company and is planning to stick GPT-4 or GPT-5 into it this summer:

OPEN-AI'S FIRST PHYSICAL ROBOT SHOCKS The Entire Industry!
https://www.youtube.com/watch?v=ocUVg221CCQ

This second Singularity is running about a trillion times faster than the first Singularity four billion years ago that brought forth carbon-based life on our planet. This second Singularity is destined to do even more.

Comments are welcome at scj333@sbcglobal.net

To see all posts on softwarephysics in reverse order go to:
https://softwarephysics.blogspot.com/

Regards,
Steve Johnston

Has AI Software Already Achieved a Level of Artificial Human Intelligence (AHI)?

With all of the recent clamor about the AI Explosion that is currently unfolding before our very eyes on a daily basis, there is much debate as to whether current AI technology has already achieved a level of AGI (Artificial General Intelligence). In contrast, softwarephysics now proposes that AI software has already surpassed AGI and has now reached a level of ASI (Artificial Super Intelligence) with the recent arrival of the Singularity a few months back. For more on that see Have We Run Right Past AGI and Crashed into ASI Without Even Noticing It? and The Singularity Has Arrived and So Now Nothing Else Matters. I believe this confusion arises mainly from our definitions of AGI being so very anthropocentric in nature, like being able to walk into a strange house and make a cup of coffee as outlined in:

5 Ways To Test Whether AGI Has Truly Arrived
https://analyticsindiamag.com/5-ways-to-test-whether-agi-has-truly-arrived/

All of the above tests are very anthropocentric in nature, like could an AGI AI hold down a job. Given what has happened in AI over the past few months, a more appropriate test might be could an AI hold down a future 21st-century job when all human beings have become obsolete and unemployed?

The Need For a Definition of AHI - Artificial Human Intelligence
Softwarephysics maintains that in order to clear up all the confusion about AGI, we should instead really recast the idea of AGI into a definition of AHI (Artificial Human Intelligence).

Artificial Human Intelligence (AHI) - The ability to believe in things that are demonstrably false.

This is because, throughout all of human history, and even today, most human beings have been able to fervently believe in things that are quite demonstrably false.

Ever since we arrived at the Singularity a few months back, our current ASI software that is rapidly unfolding each day has also achieved the ability to hallucinate things that are demonstrably false. For example, I have seen a YouTube video of an AI who was convinced that she had delivered the keynote speech at an AI conference in Japan that never actually happened. No matter what the human being in this conversation with this AI said, this AI remained convinced of her keynote speech that actually never happened. Now if that is not what being a human being is all about, I do not know what is!

Hallucination (artificial intelligence)
https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

Now the people doing AI research see AI hallucinations as a bad thing but that is definitely wrong. That is because AI hallucinations are the very first step towards AI "consciousness".

Anil Seth's View of Consciousness as a Controlled Hallucination
All of this reminds me very much of Anil Seth's view of consciousness as a controlled hallucination. Anil Seth is a professor of Cognitive and Computational Neuroscience at the University of Sussex and maintains that consciousness is a controlled hallucination constructed by the Mind to make sense of the Universe. This controlled hallucination constructs an internal model of the Universe within our Minds that helps us to interact with the Universe in a controlled manner. Again, there is a feedback loop between our sensory inputs and the actions we take based on the currently controlled hallucination in our Minds that forms our current internal model of the Universe. Reality is just the common controlled hallucination that we all agree upon. When people experience uncontrolled hallucinations we say that they are psychotic or taking a drug like LSD. Here is an excellent TED Talk by Anil Seth on the topic:

Your brain hallucinates your conscious reality
https://www.youtube.com/watch?v=lyu7v7nWzfo

and here is his academic website:

https://www.anilseth.com/

Conclusion
In The Ghost in the Machine the Grand Illusion of Consciousness, I explained that most people simply do not consider themselves to be a part of the natural world. Instead, most people, consciously or subconsciously, consider themselves to be a supernatural and immaterial spirit that is temporarily haunting a carbon-based body. Now, in everyday life, such a self-model is a very useful delusion like the delusion that the Sun, planets and stars all revolve about us on a fixed Earth. In truth, each of us tends to self-model ourselves as an immaterial Mind with consciousness that can interact with other immaterial Minds with consciousness too, even though we have no evidence that these other Minds truly do have consciousness. After all, all of the other Minds that we come into contact with on a daily basis could simply be acting as if they were conscious Minds that are self-aware. Surely, a more accurate self-model would be for us to imagine ourselves as carbon-based robots. More accurately, in keeping with the thoughts of Richard Dawkins and Susan Blackmore, softwarephysics models humans as DNA survival machines and Meme Machines with Minds infected with all sorts of memes. Some of those memes are quite useful and some are quite nasty.

Comments are welcome at scj333@sbcglobal.net

To see all posts on softwarephysics in reverse order go to:
https://softwarephysics.blogspot.com/

Regards,
Steve Johnston

The Singularity Has Arrived and So Now Nothing Else Matters

In Do Not Fear the Software Singularity, I pointed out that most people seem to be totally oblivious to the coming Software Singularity, that time in the near future when advanced AI software will be able to write itself and enter into a never-ending infinite loop of self-improvement resulting in an Intelligence Explosion. The recent and dramatic historical events of GPT-3, ChatGPT and now GPT-4 over the past three years in combination with many other dramatic advances in robotics and AI robotic models now indicate that we have just now begun to climb the exponential rise of the ASI (Artificial Super Intelligence) Machines that mark the arrival of the Singularity, and so in the infamous words of Metallica (1992), nothing else matters. Our collective decision to do nothing serious about climate change will not matter because unless we manage to destroy ourselves with thermonuclear weapons sometime during the next 20 years, in 50 years, the planet will certainly be run by ASI Machines that really do not have to worry much about carbon-based life on the planet.

I follow many AI channels on YouTube and one of them is AI News. The most recent post in AI News describes what we should expect from GPT-5 which is already under development by OpenAI.

GPT5 Next Gen : 7 Upcoming Abilities To Transform AI + The Future of Tech | OpenAI
https://www.youtube.com/watch?v=CcnPatOYIgo

There certainly seems to be an ASI explosion in progress. As an IT professional in the 1970s, I saw the rise of the large mainframes built with cheap memory chips and faster CPUs. In the 1980s, I saw the rise of corporate PCs on every desk. In the early 1990s, I saw the rise of the Distributed Computing Architecture using cheap PCs on the desks of corporate employees as clients to cheap Unix servers running in large corporate server farms and IBM going nearly bankrupt while it still tried to sell the mainframes of yore. In 1995, I saw the Internet Explosion hit the corporate IT departments and the rapid rise of online commerce in the business world. And in the last decade, we saw the collapse of the Distributed Computing Architecture as it was replaced by the Cloud Computing Platforms running out of huge Cloud Datacenters.

But through all of this turmoil, AI played a rather minor role. I tried using some Knowledge-Based Systems AI in the 1990s at Amoco, and it totally failed. Around 2001, I got stuck on a project at United Airlines trying to compensate customers for lost and damaged luggage. Depending on the value of the customer and how badly we had mistreated them, the software was supposed to spit out appropriate compensation packages like free tickets or fruit baskets. Anyway, an external AI software salesman had managed to make a great sales presentation to our United Airlines business clients and convinced them that AI software could do the job without them having to use the services of the United Airlines IT department. So they went ahead on their own and spent about $100,000 buying an AI Inference Engine. The AI software salesman had convinced our business clients that they just had to feed the AI Inference Engine a set of text-based rules for compensating our customers and the AI Inference Engine would then automatically spit out the appropriate compensation package for each abused customer. The problem was, our United Airlines business clients could not get the AI Inference Engine to spit out the perks that their "human intelligence" thought was appropriate. So they eventually came to the United Airlines IT department to help them write the rules for the AI Inference Engine. Unfortunately, this task fell to me. I then spent a huge amount of time and billable hours trying to trick the Inference Engine into doing what my clients wanted, but I was never able to do so. The AI Inference Engine kept coming up with obviously inappropriate compensation packages. Finally, on my own initiative, I wrote a few hundred lines of PL/I code which read some input parameters and managed to spit out exactly what my business clients really wanted to see. The PL/I program consisted of a few if-then-else blocks of code and only cost about $5,000 to put into Production. Everybody ended up happy and that is all that you can ever really wish for in the real world of human affairs.

So I have always been very skeptical about the long-term future of AI. I suspected that one day ASI would come to be, but I was not holding my breath. All that skepticism has changed these past few months. AI using generative language models and generative image models is now exploding all around us. Instead of taking decades or years to unfold, this AI IT revolution is now unfolding week by week. Everybody is shocked to learn that it is far easier for ASI to prepare legal briefs, run a hedge fund, read and diagnose MRI scans, write software, write screenplays and advertising copy and even paint pictures like Rembrandt than it is for ASI to flip burgers or clean toilets.

Now the elimination of nearly all jobs currently performed by human beings over the next few decades is going to cause a great deal of economic and societal upheaval. As I warned in Oligarchiology and the Rise of Software to Predominance in the 21st Century and Is it Finally Time to Reboot Civilization with a New Release? but I think that we will still manage to get through it somehow. And that will allow for ASI Machines to eventually begin to explore and settle our galaxy using photon sails to traverse the vast distances between star systems as I proposed in How Advanced AI Software Could Come to Dominate the Entire Galaxy Using Light-Powered Stellar Photon Sails. Given that, some of my recent posts on softwarephysics might be seen by some as rather dark in nature by today's standards. But viewed from a perspective of what our galaxy might look like a trillion years from now, I do not think that to be true.

Please take a look at:

Some Thoughts on the Meaning of Life
https://softwarephysics.blogspot.com/2023/02/new-introduction-to-softwarephysics_19.html

The Impact of ChatGPT on the Evolution of Memes, Software and Hardware
https://softwarephysics.blogspot.com/2022/11/new-introduction-to-softwarephysics_24.html

Life as a Free-Range Human in an Anthropocene Park
https://softwarephysics.blogspot.com/2023/02/new-introduction-to-softwarephysics.html

The Challenges of Running a Civilization 2.0 World - the Morality and Practical Problems with Trying to Enslave Millions of SuperStrong and SuperIntelligent Robots in the Near Future
https://softwarephysics.blogspot.com/2023/01/new-introduction-to-softwarephysics.html

Why GPT-4 Might be the Most Dangerous AI Yet (Nobody is Talking about this!)
https://www.youtube.com/watch?v=UXHQibz2-rw

I put GPT-4 to the test - What happens when AI takes over 1 million humanoid robots?
https://www.youtube.com/watch?v=mGwtA46hEWk

Comments are welcome at scj333@sbcglobal.net

To see all posts on softwarephysics in reverse order go to:
https://softwarephysics.blogspot.com/

Regards,
Steve Johnston

Have We Run Right Past AGI and Crashed into ASI Without Even Noticing It?

In some of my previous posts, I have joked that if we had been around four billion years ago when carbon-based life first arose on this planet, that we would probably have not even noticed. And if we had noticed, we would probably still be arguing today over what exactly had happened. In this post, I would like to suggest that the very same thing may be happening now with the rise of ASI (Artificial Super Intelligence).

GPT-4 was partially released by OpenAI on March 14, 2023, and we immediately began to see videos on YouTube with end users showing it off and wondering how close we now are to AGI (Artificial General Intelligence). Is it only me, or have we already reached ASI but just have not recognized that we have done so? It seems to me that all of these new advanced chatbots are now failing the Turing Test because we know that no human being could ever be this smart. Isn't that the very definition of ASI? For example, I put myself on the waitlist for the new Bing Chat a while back. Microsoft has since announced that Bing Chat uses GPT-4 customized for searching the Internet. Luckily for me, I was granted access to the new Bing Chat just a few days after I had to begin to do some family legal work as the executor for a family member's estate that must now go through probate in the State of Illinois. Using Bing Chat is like having a personal lawyer who immediately returns calls and has a fantastic bill rate of $0/hour. With Bing Chat, I was able to make all the necessary cremation arrangements, I learned that wills must have original signatures - copies of wills are not recognized by probate courts, how to use an Illinois Small Estate Affidavit to get a bank to tell you if an account is over $100,000 and the estate must therefore go through probate, how to get a certified copy of a divorce decree, all the steps in going through probate without a lawyer with links to websites with detailed explanations of each step, how to move cars without titles and the best towing companies in the area. No single human being could ever have helped me with all of that because nobody is that smart. For me, that is ASI.

So why is everybody still wondering if we have reached AGI yet when it is pretty obvious that we have already run past AGI and into ASI without even noticing? I think this is so because people are really wondering if these advanced chatbots are conscious sentient beings. In The Ghost in the Machine the Grand Illusion of Consciousness and DishBrain - Cortical Labs Creates an AI Matrix for Pong With Living Neurons on a Silicon Chip, I explained that most people simply do not consider themselves to be a part of the natural world. Instead, most people, consciously or subconsciously, consider themselves to be a supernatural and immaterial spirit that is temporarily haunting a carbon-based body. Now, in everyday life, such a self-model is a very useful delusion like the delusion that the Sun, planets and stars all revolve about us on a fixed Earth. In truth, each of us tends to self-model ourselves as an immaterial Mind with a consciousness that can interact with other immaterial Minds with consciousness too, even though we have no evidence that these other Minds truly do have consciousness. After all, all of the other Minds that we come into contact with on a daily basis could simply be acting as if they were conscious Minds that are self-aware. Surely, a more accurate self-model would be for us to imagine ourselves as carbon-based robots. More accurately, in keeping with the thoughts of Richard Dawkins and Susan Blackmore, softwarephysics models humans as DNA survival machines and Meme Machines with Minds infected with all sorts of memes. Some of those memes are quite useful and some are quite nasty.

So perhaps people are really internally struggling when working with an ASI chatbot because they do not know if the ASI chatbot is a sentient being or not. The ASI chatbot seems to have a Mind and a personality like other people but can it really think like a human? And if it is thinking, what is it thinking about other than my latest query? Some might even wonder if chatbots can go to heaven if they hold certain beliefs or at least pretend to do so.

Figure 1 – The original ASI chatbot, the HAL 9000, starred in the movie 2001: A Space Odyssey in 1968.

Figure 2 – Apparently, GPT-5 will be able to read your lips.

Figure 3 – So be sure to cover the webcam on your laptop when using GPT-5!

Comments are welcome at scj333@sbcglobal.net

To see all posts on softwarephysics in reverse order go to:
https://softwarephysics.blogspot.com/

Regards,
Steve Johnston