Happy New Era

We have now without a doubt entered the Artificial Intelligence (AI) era of history. In November 2022, OpenAI released ChatGPT (Chat Generative Pre-trained Transformer) and by January 2023, it had become the fastest-growing consumer software application in history with over 100 million users.

So what can people do with ChatGPT?

Just ask ChatGPT and it will tell you.

Content Generation:      Writing articles or blog posts.      Creating creative writing, including stories and poems.    Drafting emails or other written correspondence. Idea Generation:     Brainstorming ideas for projects.     Generating creative concepts for art or design. Learning Assistance:     Getting explanations and clarifications on various topics.     Studying and practicing languages. Programming Help:     Writing and debugging code.     Seeking advice on programming concepts. Problem Solving:     Generating solutions for problems or challenges.     Planning and strategizing. Conversation and Social Interaction:     Having casual conversations for entertainment.     Simulating characters for role-playing. Language Translation:     Translating text between languages. Entertainment:     Playing text-based games.     Creating and participating in interactive stories. Assistance with Tasks:     Creating to-do lists or drafting documents.     Getting help with decision-making. General Information:     Looking up general information on a wide range of topics. Therapeutic Conversations:     Engaging in therapeutic conversations or journaling.

Suppose you wanted to write an essay about the effect of the French Revolution on the world, a popular question I remember from high school history. There is a lot of information about the French Revolution easily available on the Internet. Here is lovely illustrated history but it would be quite a lot of work to distill this down to a well written one page essay. 

I asked ChatGPT to
  Write a one page essay on the effect of the French Revolution on the world.

After about 5 seconds this printed out on my screen.

The Global Ripple Effect of the French Revolution Introduction:   The French Revolution, which unfolded between 1789 and 1799, stands as one of the most influential events in world history. Its impact reverberated far beyond the borders of France, shaping political ideologies, inspiring movements for change, and redefining the global balance of power. The aftermath of the revolution left an indelible mark on the world, influencing social, political, and economic structures across continents. Political Ideals and Nationalism:   The French Revolution introduced a fervent call for liberty, equality, and fraternity, setting a precedent for the assertion of individual rights and the dismantling of traditional hierarchies. The Declaration of the Rights of Man and of the Citizen became a beacon for democratic aspirations, transcending borders and inspiring movements for independence and self-determination in colonies and nations worldwide. The revolutionary fervor paved the way for the rise of nationalism, as people sought to define their identity and destiny independent of monarchies or imperial rule. Napoleonic Code and Legal Reforms:   Napoleon Bonaparte, emerging from the chaos of the Revolution, consolidated power and implemented far-reaching legal reforms. The Napoleonic Code, a comprehensive legal system, spread across Europe and beyond, influencing legal structures and principles in diverse societies. The code emphasized equality before the law, property rights, and secularism, leaving an enduring legacy in legal systems around the world. Shift in Global Power Dynamics:   The French Revolution significantly altered the geopolitical landscape of Europe. The rise and fall of Napoleon led to a redrawing of borders and the reconfiguration of power dynamics. The Congress of Vienna in 1815 aimed to restore stability and maintain a balance of power, but the revolutionary ideals had already set in motion a trajectory towards nationalism and the eventual reshaping of empires. Social and Economic Transformations:   The revolutionary fervor extended beyond politics to encompass social and economic realms. Feudal structures crumbled, and new economic systems emerged. Industrialization, spurred by the Enlightenment ideas that fueled the Revolution, gained momentum, transforming societies and economies. The impact reached as far as the Americas, where the ideals of the French Revolution influenced independence movements and the formation of new nations. Legacy and Lessons:   The French Revolution left an enduring legacy, contributing to the evolution of modern political thought and the establishment of democratic ideals. It taught the world the power of collective action, the potential for radical change, and the need for a balance between stability and progress. While the revolution had its share of excesses and challenges, its enduring impact on the world underscores the complex interplay between revolutionary fervor and the quest for a just and equitable society. Conclusion:   In conclusion, the French Revolution’s effect on the world is immeasurable. Its ideals and principles echoed across continents, shaping the course of history and inspiring generations to come. From the rise of nationalism to the restructuring of legal and political systems, the French Revolution’s impact on the world is a testament to the transformative power of ideas and the indomitable spirit of those who dare to challenge the status quo.

From my vague memory of this history this seems like a pretty good answer, complete with a catchy title. However I don’t think the answer is nearly as good as the one given by a Chinese historian, a scholar of one of the oldest human civilizations, who is reputed to have said when asked about the importance of the French Revolution, 

“It’s too soon to tell.”

There were a lot of articles and videos about many different things that people were doing with ChatGPT. I thought that some of it was hype. In particular, I was dubious about the claims that ChatGPT could write anything much more than a very simple computer program. I was browsing an online introduction to an AI course from Harvard which used as an example of AI programming the problem of how to find a path through a maze.

Over a sequence of several lectures the course developed a Python program to solve such a maze. I then asked ChatGPT to write a Python program to solve a maze. I was amazed! 

ChatGPT produced an 80 line program that was quite different from the Harvard course and was pretty good code. I ran it against several different mazes to test it, and it did work.

Anything that’s hot gets copied pretty quickly. Besides the big players like Google, Facebook and Microsoft there are a whole host of chatbots now on the internet. Here’s one list of The best AI chatbots, but there are many other lists.

I have only tried a few of these. One that is listed as best “For fun”, character.ai allows you to converse with all kinds of characters from Socrates to Napoleon to Albert Einstein to Elon Musk.

The startup behind Pi, the one listed as best “For personal use”, believes that every website will be adding a chatbot which will enable visitors to their website to find information easily and quickly. For a company, this could streamline and extend the reach of their marketing and customer support. Pi also believes that their chatbox will be widely used by individuals as a personal assistant. Many people tried to use Siri, Alexa and Google Assistant as personal assistants but their “intelligence” and abilities were quite limited, at the level of a grade school child. Chatboxes based on the new AI will offer capabilities of a college graduate. Who would you choose to hire for a personal assistant?
How Siri, Alexa and Google lost the AI race

Another very popular application of the new AI is image generation. You provide a description (prompt) of what you want to draw and some parameters about the type of drawing and the AI program will pick images and elements from its training data to produce a result. Here’s a great article that shows many examples of what some of the leading image generators can do.
Image Generators Compared

Unlike most of the chatbots which seem to be free, many of the image generators require a subscription. I found a free image generator that I quite liked called Leonardo.AI, although I think I was partial to the name. By all accounts Leonardo da Vinci was one of the most brilliant and talented people to have ever lived. 

Leonardo has been very handy to quickly and easily make greeting cards and birthday cards for printing or email. Here’s the New Year’s card I made for 2024. I simply provided the following prompt. 

Happy New Year wishes encompassing the theme of peace on Earth and a healthy planet. 

It generated four different images in about 10 seconds and I picked the one that I liked the best.

You can see a current weakness of AI image generators on this card. They all have trouble incorporating text accurately. “Happy New Year” is legible but is followed by some gibberish; I really don’t have a clue what it says. This is still early days for the new AI.

I engaged Leonardo for another task, to design a logo for my theoretical AI Consulting business. Here’s the prompt that I gave:

Design a logo for my AI consulting business. It should be deep blue and have the letter D surrounded with interconnected nodes in a circle that reflect the neural networks of AI systems.

This looks like a really good logo to me. Maybe I should actually startup this business.

Of course there has been a lot of press about how ChatGPT makes it easy for students to cheat, how it will take over many jobs and lead to people getting fired, increase the threat and spread of fake news, and generally have disastrous effects on the economy and society.

A lot of people from many walks of life have become engaged in the debate about the dangers of AI. In 2014 Stephen Hawking warned about the growing power of Artificial Intelligence that would enable robots to replicate themselves and increase the rate at which they get smarter, leading to a tipping point or ‘technological singularity’ when they can outsmart humans.

Ray Kurzweil, one of the most prolific inventors alive and the author of five books on Artificial ­Intelligence, replied to Hawking and pointed out that the industry guidelines developed for DNA research have kept that field safe and the same can be done for AI.
Don’t Fear Artificial Intelligence

One of the people who was concerned about the direction of AI was Elon Musk. His concern was that the large companies would control the development of AI and keep the technology secret. To counter this, he and a group of other like minded tech people founded OpenAI. Their mission was to do advanced AI research but keep it in the public domain “to empower as many people as possible to have AI. If everyone has AI powers, then there’s not any one person or a small set of individuals who can have AI superpower.”

Fear of robots and computers taking over the world goes way back in science fiction. This was the theme in the film “2001: A Space Odyssey” that burst onto the big screen in 1968. It follows the evolution of intelligence, transcending from ape to caveman and then from modern man to the computer HAL. But then HAL goes rogue and causes one of the astronauts to drift away in space before the other astronaut is able to shut HAL down.

2001 was great entertainment but it was not great science. HAL would not have been a threat to the astronauts if it had been designed to follow the The Three Laws of Robotics. These were published by Isaac Asimov in 1942.
1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

In 2023 the debate about the future of AI continued at a heated pace. In March more than 1,000 technology leaders and researchers published an open letter warning that AI tools present “profound risks to society and humanity” and urged AI labs to pause development of the most advanced systems.

In December 2023 the European Parliament passed a First Regulation on Artificial Intelligence. This is fairly broad legislation that establishes obligations on companies developing and using AI that depend on the level of risk.

Unacceptable risk AI systems are systems considered a threat to people and will be banned. – Cognitive behavioral manipulation of people –  Social scoring: classifying people based on behavior, socio-economic status or personal characteristics –  Biometric identification and categorisation of people High risk systems will have to be registered in an EU database. - Systems that negatively affect safety or fundamental rights General purpose and generative AI systems have to: –  Disclose that the content was generated by AI –  Design the model to prevent it from generating illegal content –  Publish summaries of copyrighted data used for trainingGeneral-purpose AI models that might pose systemic risk have to undergo thorough evaluations and any serious incidents have to be reported.

Canada and the U.S. are also preparing AI legislation.

My view is that many of the “experts” (physicists, philosophers, historians, journalists, politicians) who are sounding the alarm about AI do not have an in-depth knowledge of software/technology and the psychology/neuropsychology of human intelligence to fully comprehend how difficult the problem of advancing to general intelligence is. They underestimate how long it will take and amplify fears expressed by others.

The current AI boom is fueled by a breakthrough in deep learning which is fundamentally advanced pattern recognition. AI models today are not very strong in analysis, reasoning and logic which are key components of human intelligence. These critical characteristics of advanced general intelligence are more difficult to achieve than pattern recognition. The number of breakthroughs necessary to achieve general intelligence stretches far into the future. Climbing a tree can take a monkey nearer to the moon, but climbing taller and taller trees does not enable a simian astronaut to plant a flag on the moon.

Artificial Intelligence is not new. It began way back in the 1950s when the term was coined. There were a number of early dramatic successes, such as a program that played championship level checkers and a program that could prove many of the symbolic logic theorems in Principia Mathematica, the monumental classic on the foundations of mathematics. These successes convinced many in the field that AI was going to revolutionize the world and a boom in AI began. However the hype raced ahead of reality and the boom turned to bust in the 1970s.

A second AI boom occurred in the 1980s. The focus was on “Expert Systems” in which input from experts in a particular subject area was entered into a system which could then respond to user queries and provide expert answers. A number of such expert systems were successfully used, from hardware circuit design to medical diagnosis. But as in the first AI boom, many of the prototype systems failed to scale to real life applications and meet the hyped expectations so investment dried up and the boom was over.

Will the AI boom today escape AI’s past history?

The current boom in AI began in 2012 when a program to recognize images in a database containing more than 20,000 categories won the annual ImageNet competition. The software had been developed by Geoffrey Hinton and several graduate students at the University of Toronto using a technique that was called “Deep Learning”.  It achieved 85% accuracy, a very significant improvement over previous years.

Deep learning was a type of neural network that had been used as a model for AI research since it was first developed in 1960 by the psychologist Frank Rosenblatt. Anybody who took a psychology course at McGill University back in the day would be familiar with the idea of a neural net. It is based on D. O. Hebb’s neuropsychological theory of how learning occurs in the brain, which can be summarized by the pithy statement “Neurons that fire together wire together”.

The success of deep learning triggered a whole new wave of AI research. After decades of research and millions of dollars spent by IBM and others, progress in speech recognition had been slow. Applying deep learning resulted in a breakthrough and enabled computers to understand the spoken word in real time. This success was the basis for Siri and Alexa and the voice recognition used by telephone answering services. 

But a new era only occurs when a revolutionary development (based on prior breakthroughs) changes life for many people. So although the current AI boom began in 2012, the AI era didn’t really begin until 10 years later with ChatGPT.

In fact there have been 5 eras in the evolution of computers since the first early machines.

1951 First Commercial Computers —  after the first working computers were developed during World War II in government research labs and universities, general purpose computers became available from many different companies in the U.S., Europe and Japan. Their rapid growth was triggered by the invention of the transistor in 1947, an invention which many people have called the most important invention of the 20th century. The transistor, which functions as a switch, is at the heart of electronics design. It replaced vacuum tubes and made computers a lot smaller, faster, cheaper and more reliable.

1964 IBM Mainframe Era — the IBM System/360 family of compatible computers allowed customers to start with a small system and upgrade to larger computers as their computing needs grew. The success of System/360 drove many computer vendors out of the market and established IBM as a dominating force in the computer industry for the next 30 years.

1981 Microcomputer Era — Intel invented memory chips and CPU chips in 1970/71 which gave birth to microcomputers such as the Apple I. But the revolution in the widespread use of microcomputers that defined the era only started with the IBM Personal Computer (PC) in 1981.

1995 Internet Era — the Internet was first built in 1969 as a research project under a grant from the U.S. Department of Defense, but widespread use of the Internet did not occur until after the World Wide Web was developed in 1991 and Microsoft Windows 95 made it easy for non-technical users to access the Internet with simple point and click.

2007 Smartphone Era — the first smartphone was developed by IBM in 1994 and was followed by the BlackBerry in 2002. But the general Smartphone era really began with the revolutionary Apple iPhone in 2007 which led to widespread mobile computing for almost everyone .

2022 Artificial Intelligence Era — ChatGPT and other chatbots came out of the lab and into widespread use following the breakthrough successes of deep learning.

It seems there is a new era in technology on average every 14 years. 

What will the next era be 14 years from now in 2038?

I predict the next era will be Personal Robots (PR) or Personal Medicine (PM).

It is a natural extension of AI to herald a revolution in robots. Robots are not new. The first programmable robot was used in a General Motors plant in 1961. Robots are now in fairly wide use but they are generally designed for very specialized industrial jobs. There is a lot of research into making general purpose robots, especially in Japan, the largest exporter of robots in the world. Japan’s declining population has spurred the government to invest heavily in robot technology to provide their future workforce.
Japan’s robots are conquering the world

Personal Medicine is where medical decisions and interventions are tailored to an individual. Some of the breakthrough discoveries that are the foundation for this advance have already been made. 

The Human Genome Project that identified, mapped and sequenced almost all of the genes of the human genome was completed in 2003. This opened the way for gene therapy and genetic engineering.

CAR T cell therapy is a gene therapy in which T cells, one of the important components of the immune system, are taken from cancer patients, genetically modified to contain receptors that will attach to cancer cells and re-infused into patients. The modified T cells are then able to locate cancer cells and destroy them. The first CAR T cell therapy approved by the FDA was in 2017 for the treatment of childhood leukemia. Following this success, several more CAR T therapies were approved for other types of leukemia and other blood cancers.

In 2023 AlphaFold won the Lasker Award for solving a 50 year old problem in biology to accurately predict the shape of a protein molecule. In a very short time AlphaFold has proved to be an invaluable tool in many areas of medical research. The program and the database of 200 million structure predictions, amounting for nearly all the known human proteins, have been shared with the scientific community and more than 300,000 science researchers have used these resources. AlphaFold is one of the most significant achievements in the history of Artificial Intelligence. The story of this breakthrough was unfolded in an earlier blog article.

CRISPR is a revolutionary technique to edit genes that was discovered in 2012. Jennifer Doudna and Emmanuelle Charpentier discovered that Cas9, a gene-cutting protein used by bacteria to kill viruses, can be re-engineered as a programmable gene editing tool. The CRISPR-Cas9 system is faster, more accurate and more efficient than previous gene editing methods. Doudna and Charpentier were awarded the Nobel Prize in Chemistry in 2020. 

The following short video is a good demonstration of how CRISPR does “cut and paste” gene editing.
CRISPR Explained

In 2013 Charpentier co-founded a startup company called (surprise, surprise) CRISPR Therapeutics. They undertook developing treatments for several diseases and conducting them through clinical trials. In 2023 their treatment for sickle cell disease received FDA approval. This is the world’s first approved medicine based on CRISPR gene editing. Many people think this is a groundbreaking achievement opening the way for a lot of new CRISPR developed medicines.
CRISPR Gene Editing Breakthroughs

_____________________________________

So here are three possible scenarios for 2038. 

1. You order the PR (Personal Robot) that everybody is buying and an hour later it arrives from Amazon by drone delivery. After a friendly introduction, you ask it to do the laundry and then prepare a gourmet lunch while you get back to more important things in life.

2. After going to your doctor for a checkup, she informs you that the diagnostic tests that you took have identified a rare disorder. She has submitted all the test results and your medical history to the PM department (Personal Medicine) at the hospital. In three days you are notified that a drug has been made to treat your medical condition and that you need to come into the hospital for a three day initiation. Possible side effects cannot be completely predicted yet by Personal Medicine so you need to be under observation in case you need intervention.

3. Continuing scenario 3, you get ready to depart for the hospital for your new PM drug treatment program but there is a howling windstorm and torrential rain that make it impossible for anybody to travel. Because the world did not move quickly enough to meet the target of limiting global warming to 1.5°C by 2030, extreme weather events are happening frequently and continue for weeks. You are shut indoors while your health condition deteriorates and you wonder if you will get your new miracle drug on time.
(more on climate change in a future blog)