Artificial Intelligence Explained

Featured Image: Ancient technological artefact, detail from Tony Stewart, Hovering Christ 2011.

  Breadtag Sagas ©: Author Tony,  6 September 2024

This is article one in a series on AI. The second article in the series is Artificial Intelligence History, Toby Walsh.

Artificial Intelligence (AI) Explained Simply

Main Points

• Definition of Artificial Intelligence
• Preamble
• Visual overview: Go, AI companion, sex robots, driverless cars, killer robots, art
• Timeline
• Expert System
• Algorithm
• Big Data
• Traditional or Narrow Artificial Intelligence
• Machine Learning
• Broad or General Artificial Intelligence
• Generative AI
• Neural Network
• GANs and VAEs
• Deep Learning
• Alignment
• Comment

1 Definition of Artificial Intelligence

Artificial intelligence, or AI, is technology that enables computers and machines to simulate human intelligence and problem-solving capabilities. (Britannica)

2 Preamble

I’ve thought about recent AI in a desultory way for at least the last five years. I began using chatGPT in a similarly casual way in December 2022. I am surprised at how quickly generative AI has begun to impact our lives. Whilst I don’t think the greed of Wall Street is on a winner in the short term, I think that we are at the beginning of a rapid evolution, which will change our lives radically.

In 1996, when I first began using the new thing called the Internet or World Wide Web, I accepted the change without much thought or analysis.

I missed the Google launch in 1998 and only noticed Google when it became useful. In 2001 when iTunes launched, I was an early adopter, but I didn’t pay too much attention when Apple stole the music industry. In 2007, when the iPhone smart phone launched, I waited for the third iteration. I became a user rather than a researcher of the trend.

This time I lay no claim to be at the forefront of understanding AI and the trends. I did know about expert systems from the late 1970s and while I have been waiting expectantly ever since, I’ve only marvelled at the mind-boggling applications, from Chess and beyond without much thought.

Now, I want to learn more about AI and the artificial intelligence revolution, before it becomes an everyday commonplace, more than it is already. By this I am probably mostly referring to generative artificial intelligence, rather than to narrow artificial intelligence (definitions later).

The current article and further articles on AI that I write are part of a learning process for me. I hope they will also be useful to anyone else who wants to know about AI in a general rather than a technical way.

I’ve been using Perplexity (a generative artificial intelligence program like ChatGPT) to help write this article. It is making the process quick and much less onerous than doing it without help.

3 A Brief Visual Overview of Artificial Intelligence

I recently watched a good Australian documentary on Docplay by Justin Krook called Machine, which came out in 2019 (1h 25 min). It is also available elsewhere. Although Machine is five years old, there aren’t to my knowledge newer and better examples regarding artificial intelligence. Most of the topics covered are current and haven’t changed in general terms over five years, except for generative AI. There are also a few new and exciting applications of narrow AI, but they don’t change the picture overall.

Machine explores the rise of artificial intelligence and its potential impact on humanity. The film has strong visuals and interviews with some world-renowned experts on AI.

The film examines the practical applications of AI and the ethical questions raised by these new developments. I agree with the reviewer who said: it leaves you wanting more.

I am using the documentary as a visual entry into the topic, rather than doing the usual current listing of some real world applications and anticipated additions. The latter are inevitably misleading and frequently unimaginative.

Lee Se Dol 2016, photo LG Electronics, Wikimedia

Visual Metaphors

3.1 The Game of Go

The documentary begins showing the Korean world expert on Go, Lee Sedol, being beaten by a machine AlphaGo in 2016. It is heartbreaking for the champion.

It reminded me of Gary Kasparov being beaten by Deep Blue in 1997 but worse. Deep Blue was had been beaten by Kasparov the year before. It was not an elegant system, that is, much the same as existing chess programs and merely driven by brute force computing power and a massive database.

Alpha Go was also an expert system, but it used neural networks (see below) for decision-making, as well as a vaster database.

Don’t worry if some concepts are unfamiliar. The aim of the article is to explain them to you.

3.2 The Weird Stuff First

Matt McMullen and Harmony, photo Eduardo Contreras, San Diego, Union-Tribune, 2017

Replika is an AI companion designed by Eugenia Kyda following the sudden death of her mentor and partner, as a way to deal with her grief. It is a current AI chatbot very similar in principle to ELIZA designed in 1965 and supposedly Replika is used by millions as an AI companion. (I signed up to give it a go, but I found the Avatars offered and the constant attempts to sell the paid version, made it too creepy to use.)

There have been other AI companions, I heard a radio program about an early one. A young woman had been using the AI as a close companion and friend for a year and a half, when the owners changed the algorithm. She describes in anguished terms how she felt as if someone had died and went into a process of grieving.

Replika in the documentary is shown as text messages and then an amazing cascading multi-level array of complexity to visually portray neural networks and the data handling capability of AI.

Sex robots or androids are also shown in this segment with a commentary by Matt McMullen, the founder of Realbotix, now Simulacra Corp. McMullen is the early adaptor of silicon rubber skin-covered sex robots beginning in 1997.

McMullen says the first question potential customers ask is if the robots are anatomically correct.

The documentary shows some of McMullen’s production line, which is also visually creepy. However, it is not surprising that one of the first uses of anatomically correct sex-humanoids with some form of AI control, will be the sex industry.

3.3 AI (Artificial Intelligence) Apocalypse

Kate Darling MIT, photo Mickie Winters, Winters Photography Co 2024

Kate Darling (MIT media lab, Robotics ethicist) says:

AI and robotics is heavily influenced by science fiction and pop culture, so people have this image of what this is and it is not always the correct image. So we tend to massively overestimate or underestimate what the technology is capable of.

The segment is about humanoid or animaloid robots. There is a NASA robot that begins to walk and then fails. Another robot leaps onto a plinth and does a somersault. Humanoid and dog-like robots appear scary. The makers tend to consciously or unconsciously make them so.

Glen Weyl (to be introduced in a later article) comments on the robots designed to scare whereas other non-humanoid robots in other cultures (e.g. Japan) can be made to seem gentle and practical as opposed to confronting. The Kate Darling quote above is pertinent.

Various experts talk about the complexity of the software needed for robots to do the most basic things autonomously like, for example, walking.

Rodney Brooks (Australian Roboticist) says:

When I see people who say robots are going to take over the world, it is so far off from the capabilities of these robots.

Rodney Brooks, photo Christopher Michel 2021

He says:

What can our robots do today really well? They can wander around and clean up a floor.

The visuals show a vacuum cleaner cleaning around furniture and getting stuck.

He also says that when AlphaGo beat Lee Sedol, Alpha Go was supported by a team of 200 engineers whereas Lee Sedol received a cup of coffee.

He also says that if a human Go player was presented by a 17×17 board instead of a 19×19 board they would immediately be pretty good at it, whereas AlphaGo wouldn’t be able to play at all.

3.4 Driverless Cars

The same issues that plague robots are problems for driverless car software. The documentary shows excellent visuals of driverless car technology with comments by several experts. Driverless car software is good on freeways, which the experts say gives people unreasonable and dangerous confidence in the technology. Some drivers believe driverless car technology is so much more sophisticated than it is and use it in inappropriate scenarios.

But, the software can’t cope in slow traffic with bicycles, pedestrians, public transport, children and animals. The expert consensus is that true driverless car technology is a long way off.

A local newspaper in August 2024 says that a team at ANU in Canberra has developed a new approach to the slow traffic problem for driverless cars. Driverless taxis do exist but there have been problems. I have so often been disappointed awaiting technology that I’ll believe it when I see it. But, I could be wrong!

3.5 Killer Robots

The idea of autonomous warfare is mentioned but superficially in the documentary, merely regarding the open letter of 2015 signed by 20,000 notable people including experts and the UN forum (2013 followed by committees). The letter by the way was coordinated by Toby Walsh whose book The Shortest History of AI published in April 2025 will be discussed in the next article.

I covered the subject of killer robots in the article of Isaac Asimov’s I, Robot. I still find the threat of the military use of AI alarming, but suspect that the idea of killer robots swarming across a landscape is also for far in the future.

Lethal androids are attractive in an era when recruiting cannon fodder seems extremely difficult, and robot companies are certainly looking at the military for funding. I suspect, however, observing the Ukraine war for technological innovation, particularly low cost elements, and simpler existing forms of AI that the future may be different.

(Even though military applications are often the source of new technology. The military industrial complex is both conservative and slow to develop new things.)

Pindar Van Arman, photo Hudapi 2018, Wikimedia

3.6 Artificial Intelligence and Art

Attempts of AI to produce art are more interesting and complex than facial recognition technology and deep fakes. The last two are examples of narrow AI.

In the documentary, Pindar Van Arman an American roboticist and artist who has expertise in developing painting robots, ponders the future. The documentary shows his 3-4 year old daughter painting. He marvels at her free expression and creativity. This scene is followed by a robot, he is teaching to paint.

The key element he says is that robots (and AI) can only copy things. According to Van Arman, until a machine has something it wants to tell us, it won’t be creating art.

3.7 Super Artificial Intelligence

The final segment of the documentary on super intelligence also suggests that it is not on the horizon yet, because of ‘alignment failure’. Alignment is an import concept that will also be dealt with later.

The reason why super intelligence may be some time off, the experts thought, is because the issues of alignment, empathy and value systems need to be resolved and it won’t be easy.

Subtle values are thought to be even more difficult but necessary.

4 AI Artificial Intelligence Timeline

We have to remember that generative AI hasn’t come from nowhere. Below is a brief, summarised history of the development of AI.

1940s: Foundations of Artificial Intelligence

1942-1950s: Preliminaries and Birth of AI

1942: Alan Turing developed the Bombe machine (and later the first main frame computer) at Bletchley Park to help break the German Enigma Code during WWII.

The Bombe, Historic Photo, Bletchley Park

1943: Warren McCulloch and Walter Pitts designed the first artificial neurons, laying the groundwork for neural networks.

1950: Alan Turing introduced the Turing Test.

1956: John McCarthy coined the term ‘artificial intelligence’.

1960s-1970s: Early Development and Challenges

1965: Joseph Weizenbaum created ELIZA, an early natural language processing program that simulated conversation, paving the way for future chatbots.

1972: The development of DENDRAL, an expert system for chemical analysis, which demonstrated the application of AI in specialised domains.

The first AI winter, reduced funding and interest, due to unmet expectations.

1980s: Revival and Expert Systems

The 1980s saw a resurgence in AI, particularly with the rise of expert systems, which were designed to solve specific problems using a knowledge base.

Limitations in handling novel information led to another AI winter by the late 1980s.

1990s: Machine Learning and Strategic Milestones

1997: IBM’s Deep Blue defeated world chess champion Garry Kasparov, marking a significant achievement in AI’s strategic capabilities.

The 1990s also saw advancements in machine learning techniques, including the resurgence of neural networks.

2000s: Generative AI and Consumer Applications

The new millennium ushered in generative AI, with significant developments in algorithms and applications. Notable advancements included the introduction of voice recognition systems like Siri in 2008 and the emergence of intelligent personal assistants.

2010s-2020s: AI in Everyday Life

2011: IBM Watson gained fame by winning the quiz show Jeopardy! showcasing AI’s ability to process and analyse vast amounts of information.

2016: Sophia, a social humanoid robot, became the first robot to receive citizenship, highlighting the intersection of AI with social and ethical considerations.

2020: The introduction of advanced language models, such as ChatGPT-3 demonstrated the capabilities of AI in generating human-like text, further integrating AI into daily life.

Alan Turing 1951, photo Elliot and Fry, NPG London

Conclusion

AI has evolved from theoretical concepts in the 1940s to a multifaceted technology that permeates many aspects of modern life. The journey has been characterised by cycles of optimism and disillusionment, yet continuous advancements in machine learning, natural language processing, and robotics indicate a promising future for AI technology.

The AI winters referred to above are typical of technology and cycles of creative excitement and disillusion. Development money follows these cycles. But, even worse, when the technology allows leaps forward in practical application to resolve really difficult problems, there is often resistance.

I can remember in the 1980s being told that OCR and verbal data entry for text were just around the corner but they weren’t. Decades later by the time OCR was extremely accurate and verbal text entry was possible, I was no longer interested. I only found by accident a couple of years ago when I had to photograph manuscript pages in a National Library. That my iPhone could turn the photograph into text immediately with virtually perfect accuracy, even when the pages weren’t flat. I have used this extensively ever since. Photographing the wall text in public galleries and using the text to write up the exhibition along with photographs of the paintings is also useful. I’ve found this use of marginal AI (the borderline) as useful to date as I’ve found generative AI.

5 Some Useful Definitions and Explanations

We’ve used a few terms above without defining them. The definitions below should provide an intuitive overview of the field of AI without understanding the mathematics or the coding technology.

5.1 Expert System 

An expert system is a specialised branch of artificial intelligence (AI) designed to emulate the decision-making abilities of a human expert in a specific domain. These systems are particularly useful for solving complex problems that typically require significant human expertise. They operate by utilising a knowledge base and an inference engine to process information and provide solutions or recommendations.

Examples:

• Medicine: For diagnosing diseases and recommending treatments.
• Finance: To assess risks and provide investment advice.
• Environmental Management: For analysing ecological data and making conservation decisions.

These systems do not replace human experts, but serve as tools to augment human decision-making capabilities.

Simple examples of expert systems and associated robotics were developed in the 1980s and 1990s. One of the cleverest uses of robotics with a very simple expert system controller early on was a welder and other industrial machines following routines set-up by human experts. For example, an automatic robot-arm welding machine was taken through a series of actions by a human welder. The expert system recorded the movements of the machine including turning the welding machine on and off. But, the machine could only reproduce that pattern exactly. If the object to be welded was not delivered to the exact place, the machine could not work.

That problem still exists to day, any slight deviation causes failure.

Nevertheless, my local chemist has a small robot delivery system that simplifies prescription delivery by selecting the drugs from an inventory of thousands. Similar, automated warehouses or logistics centres have simplified the delivery of products in a myriad of applications. They are a marvel to behold. I was gob-smacked by an automated yoghurt factory, when we did a technical assessment of it in 1998.

Expert systems are important and ubiquitous today.

5.2 Algorithm

An algorithm is a finite sequence of well-defined instructions, typically used to solve a class of specific problems or to perform a computation. It acts as an exact list of steps that conduct specified actions in a particular order, either in hardware or software. Algorithms are used as specifications for performing calculations, data processing, and automated decision-making. (The theoretical Universal Turing Machine (1936-37) is a ‘device’, which defines all possible algorithms.)

Boston Dynamics, Atlas Robot can Backflip but Fails Simpler Tasks, Sunday Times, 2017

5.3 Big Data

Big data refers to large and complex data sets that traditional data processing applications cannot effectively handle. It encompasses structured, semi-structured, and unstructured data, with a focus on unstructured data.

Algorithms and Big Data are fundamental building blocks for AI.

5.3.1 Non-AI uses of Big Data

An interesting sidelight on this is that the use of very simple algorithms with big data sets can provide illuminating answers (without AI).

I was reading a relatively old 2007 book by Ian Ayres about number crunching with big data sets. He gives examples across a wide range of areas including health and the law, where using very simple algorithms and statistical correlation and even simpler statistics can yield answers that are often surprising, and are better than the intuitive life experience of experts.

Big Data in its own right is being explored but rather slowly.

I remember in the early 2000s being gob-smacked when I heard that Google was keeping a snapshot of the entire Internet every day. (Not actually entire, today I’ve heard that the searchable Internet is only 5% of the Internet: ‘Dark web’ etc.)

This soon became slightly less feasible but it is bemoaned that in February 2024 the Google Cache Internet Archive (which allowed users to check older versions of a website and was a snapshot of the public Internet but not daily) was closed down. The earlier (1996) Internet Archive/Wayback Machine is still going strong.

From a Big Data viewpoint, the two paragraphs above just show what is possible. The continuing increasing of speed and amount of storage in computing (Moore’s law), which means a doubling of power and halving of cost every two years, explains why unimaginable troves of Big Data exist now.

With AI it is not hard to comprehend that an accessible database of all human knowledge could potentially exist already, but it doesn’t quite… Nevertheless, an accessible database of everything known is just around the corner.

6 AI Itself

Expert systems are a low level example of traditional AI. Machine Learning is a higher level of traditional AI and on the borders of general AI.

6.1 Traditional or Narrow AI

Traditional AI, also known as Narrow AI or Weak AI, is a subset of artificial intelligence that focuses on performing specific tasks intelligently using predetermined algorithms and rules. These AI systems are designed to excel at a single activity or a restricted set of tasks, such as playing chess, diagnosing diseases, or translating languages.

Traditional AI can also be very powerful (e.g. playing Go and Chess) and very useful. The capability has been around for a long time and the reason that it isn’t more widely used and useful is a failure of imagination, policy, funding and a generalised human lack of creative implementation.

Dog Robot, Boston Dynamics

6.2 Machine Learning

Machine learning (ML) is a significant subfield of artificial intelligence (AI) that focuses on developing algorithms and statistical models that enable computers to perform tasks without explicit instructions. Instead, ML systems learn from data, identifying patterns and making predictions based on that information.

Natural Language Processing: Used in applications like chatbots and language translation. Computer Vision: Important for facial recognition and autonomous vehicles, are both examples of machine learning.

6.3 Broad or General AI

Artificial General Intelligence (AGI), also known as strong AI or deep AI, refers to a theoretical form of artificial intelligence that possesses the ability to understand, learn, and apply knowledge across a wide range of tasks at a level comparable to human intelligence. Unlike Narrow AI, which is designed for specific tasks and operates within limited parameters, AGI would be capable of flexible problem-solving, reasoning, and adapting to new situations without predefined instructions.

Current examples of generative AI are a step in the direction of AGI but not there yet, despite the excitement and greed of Wall Street and the other hype. Nevertheless, these generative models can be enormously useful and with machine learning and even basic expert systems are going to change our lives radically and quickly in numerous ways.

Our institutions are having to cope with the generative AI explosion already.

6.4 Generative AI

Generative AI refers to a subset of artificial intelligence that focuses on creating new content, including text, images, audio, and other forms of data, based on the patterns learned from existing data. Unlike traditional AI models that primarily make predictions or classifications, generative AI models are designed to generate new data that resembles the training data they were exposed to.

Generative AI can produce a wide range of outputs, such as stories, poems, images, videos, and music.

These models learn from vast datasets, identifying and replicating the underlying structures and patterns.

6.4.1 Learning Processes in Generative AI

The whole purpose of the learning processes involved is:

1. To harness and process massive sets of Big Data, and
2. To fine tune and hone the output to make it seem indistinguishable from human derived intelligence.

6.5 Neural Network

A neural network is a type of machine learning (or broad AI) model designed to simulate how the human brain works. It consists of interconnected nodes, called neurons, which work together to process information and solve complex problems.

Structure: Neural networks are made up of layers. The first layer, known as the input layer, receives data. The information then passes through one or more hidden layers, where the neurons process the data further. Finally, the output layer produces the result.

Simplified Scematic of a Neural Network

The whole purpose of a neural network is the ability to handle vast amounts of data in random and controlled ways to produce an intelligent response in some form.

Expert systems and machine intelligence as defined above can also use neural networks to produce their results.

A simple example of a powerful but fixed problem of narrow AI is to teach a neural network to identify a photograph of a cat. Over time through endless iterations the program becomes better and better at identifying cats. But, ask it to identify a dog and it can’t.

To identify a dog, you start again from scratch and provide it with endless samples of dog photographs. Generative AI models aren’t limited in this way.

Similarly, chess programs only play chess. Go programs only operate on a 19 x 19 board. They can’t solve other problems.

6.6 What are GANs and VAEs?

Generative AI models use dual-learning processes like GANs and VAEs. These are complex processes. Relax. One only needs to gain a vague sense of how they work. Think of them as iterative processes (repeated over and over and over endlessly by very fast processors).

6.6.1 Generative Adversarial Networks (GANs)

GANs consist of two neural networks: the generator and the discriminator.

The Generator creates new data samples based on random input. Its goal is to produce data that is indistinguishable from real data.

The Discriminator evaluates the data produced by the generator against real data. It aims to correctly classify data as either real or generated.

The two networks engage in a game-like scenario where the generator improves its output to fool the discriminator, while the discriminator enhances its ability to differentiate between real and generated data. This adversarial training continues until the generator produces data that the discriminator can no longer reliably distinguish from real data.

GANs can be used for images, music and text, but common practical examples are generating human face images and deep fakes.

6.6.2 Variational Autoencoders (VAEs)

Variational Autoencoders (VAEs) are a type of model used in artificial intelligence to create new data that looks like the data they were trained on.

The Encoder takes the original data and compresses it into a smaller space called the latent space (a mathematical concept). This smaller space keeps the most important features of the data.

The Decoder takes samples from the latent space and tries to rebuild the original data. The goal is to create new data points that are similar to the data used for training.

VAEs are probability distribution processes, which can also be used as image generators, but are useful for signal analysis, anomaly detection of all sorts, data compression and clustering or classification.

Both 1 and 2 are dual-learning processes, whose purpose is to hone the accuracy of any output.

Two Final Definitions

6.7 Deep Learning

Deep learning is a specialised subset of machine learning that employs artificial neural networks to analyse and learn from vast amounts of data. It is a key component of artificial intelligence (AI) and is designed to simulate the complex decision-making processes of the human brain.

Real world examples are:

• Healthcare — advanced medical diagnostics and personalized treatment plans
• Computer Vision — enhanced image recognition capabilities e.g. facial recognition
• Natural Language Processing (NLP) — machines able to comprehend and generate human language e.g. virtual assistants like Alexa and improved machine translation services
• Fraud detection
• Robotics
• Games and entertainment

6.8 Alignment

AI alignment refers to the process of ensuring that artificial intelligence systems operate in ways that are consistent with human intentions, values, and ethical principles.

Audrey Tang (to be introduced in a later article with Glen Weyl) simplifies Alignment by calling it the steering wheel.

6.9 Conclusion

These last two definitions are to my mind rather fuzzy at this stage. Deep learning just appears as a subset of neural networks (and can cover the whole range of AI both narrow and general) but it may make more sense technically, which is outside the scope of my research. (I was unaware when I wrote this of the research of Geoff Hinton which helps to explain deep learning as outlined in my next article in a historical and technical context.)

Alignment is important but will have to be dealt with later. To my mind it is very close to Fred Emery’s work covered in earlier articles Causal Texture and the Search Conference and also close to Isaac Asimov’s attempts to define laws of robotics.

I suspect alignment and its lack may split AI into something like white hat and black hat applications, but much more frightening.

7 Comment

I’m hoping the above makes as much sense to you as it now does to me. This is the foundation for a series of articles on AI which I hope will be ongoing. I don’t know where I am going next.

Although at some stage I do want to cover the work that Audrey Tang and Glen Weyl are doing on using AI to further the future of democracy. This is, I think, related to the work of Fred Emery cited in the above paragraph and to grassroots democracy and participatory methods in general.

We went to a wonderful talk by Audrey Tang and Glen Weyl at the National Press Club in Canberra on 14 August which was advertised as:

Audrey Tang, Taiwan’s 1st Digital Minister & Glen Weyl, Founder of RadicalxChange, will Address the National Press Club of Australia on “Plurality: The Future of Technology and Democracy”

Unfortunately, this talk on the ABC, Australia is going to be taken down routinely within days but if you search the two names on YouTube there are other speeches online. There is a free collaborative book on the topic also available online (but at 420 pages a video is more accessible).

Key Words: Artificial intelligence, AI, technology, computer, machine, software, hardware, simulate, human intelligence, problem-solving, Internet, World Wide Web, Google, Apple, iPhone, smart phone, ChatGPT, Perplexity, Chess, Justin Crook, Go, Lee Sedol, AlphaGo, Gary Kasparov, Deep Blue, Replika, AI companion, Eugenia Kyda, AI chatbot, Sex robot, sex android, Matt McMullen, Realbotix, Simulacra Corp, sex industry, Kate Darling, MIT media lab, robotics ethicist, humanoid, animaloid, android, Rodney Brooks, roboticist, driverless car, killer robots, Isaac Asimov, I Robot, lethal androids, military applications, facial recognition, deep fake, Pindar Van Arman, art, super intelligence, alignment failure, empathy, values, generative AI, Alan Turing, Bombe, main frame computer, Bletchley Park, German Enigma Code, Warren McCulloch, Walter Pitts, Turing Test, John McCarthy, Joseph Weizenbaum, ELIZA, DENDRAL, AI winter, expert systems, machine learning, neural networks, algorithm, voice recognition, Siri, Alexa, personal assistant, IBM Watson, Sophia, OCR, augment, human decision-making, Universal Turing Machine, big data, statistics, Google Cache Internet Archive, Internet Archive, Wayback Machine, Moore’s law, Traditional AI, Narrow AI, Weak AI, identifying patterns, making predictions, Natural Language Processing, computer vision, autonomous vehicles, Artificial General Intelligence, AGI, strong AI, limited parameters, predefined instructions, flexible problem-solving, generate new data, training data, Generative AI models, dual-learning processes, GANs, VAEs, deep learning, alignment, human intentions, ethical principles

Further Information

Perplexity, Generative AI chatbot

Perplexity Site  (This link does work but Perplexity wants to verify you are human.)

Chess and Go

Wikipedia has the following information on Deep Blue, Alpha Go versus Lee Seedol and on Alpha Go.

People mentioned from the Machine Documentary Film

Eugenia Kyda and Replika

Eugenia Kyda and the Replika Story

Matt McMullen and Realbotix

Matt McMullen and Sex Robots

Realbotix

Kate Darling

Kate Darling’s Website

Rodney Brooks

Rodney Brooks

Pindar Van Arman

Pindar Van Arman

Cloud Painter

Alan Turing

Alan Turing

Turing Test

Universal Turing Machine

Joseph Weizenbaum and ELIZA

Joseph Weizenbaum and ELIZA

Super Crunchers

Ian Ayres Super Crunchers: Why Thinking-by-Numbers Is the New Way to be Smart 2007.

Ian Ayres is a law professor at Yale Law School. the book is about how quantitative analysis of social behaviour and natural experiment can be creatively deployed to reveal insights in all areas of life, often in unexpected ways.

Internet Archive

In February 2024 the Google Cache Internet Archive bites the dust.

The Internet Archive Story

The Internet Archive/Wayback Machine