Why Machines Learn: The Elegant Math Behind Modern AI by Anil Ananthaswamy
This book explores the math behind AI. It is rather straightforward, with a lot of vectors. However, it requires some close attention to grok. It seemed to be an interesting topic, bur I didn't feel I came out better understanding it after completing the book.
Co-Intelligence: Living and Working with AI by Ethan Mollick
This book takes a balanced view of AI. It is here. It can help us do a lot of powerful things. It will also cause a lot of changes. It can succeed at many complex tasks, yet can stumble at seemingly easy tasks. It likes to please, and will "hallucinate" to provide answers we want. We need to stay involved to make sure it doesn't spin out of control. Education and work will change. We still don't know the limits of AI. Maybe it will get much better. Maybe we are near it's limit now. There can also be many negative and positive outcomes. It will require work to be on the good side.
The MANIAC by Benjamin Labatut
Is this fact or fiction? In the world of AI, does it even matter? This is a novel about the history of AI that is mostly based on fact. The author started with a bunch of facts, then embellished the detailed of how they happened. There are many internal details that read like a tell-all documentary. There is a great deal of artistic license taking in these part, most of the purpose of making John von Neumann look bad.
The book spends the bulk of the time in a middle section on the life of von Neumann. The start is short bit about a friend of Einstein. The end covers modern AI, including alphaGo and the Go player that it defeated. (The AI did lose one round when the human played a move that it didn't crasp, causing it to fall into a loop of bad moves.)
The book has a gothic feel, implying that AI may be the dark abyss that we are falling into. Did von Neumann make a deal with the devil to set us on this path?
The Thinking Machine: Jensen Huang, Nvidia, and the World's Most Coveted Microchip by Stephen Witt
After ending with the exploration of doomsday AI scenarios, the book ends with the disclaimer that it cannot be used to train AI. Ironic or prescient?
I found this book a better exploration of Nvidia than Nvidia Way. While that book focussed on business process, this book puts the emphasis on the people and what is being achieved. The founding of Nvidia and the evolution from graphics to GPU and AI is covered. However, more emphasis is given to the AI part here. Are we moving into dangerous territory now?
Superagency: What Could Possibly Go Right with Our AI Future by Reid Hoffman and Greg Beato
The authors think AI is the next great thing and will help out humanity. This piece focusses on the positives of AI with little attention to the negatives (like energy consumption).
Empire of AI: Dreams and Nightmares in Sam Altman's OpenAI by Karen Hao
Sam Altman is not a good guy and we are trusting him with what could be existential technology. He has had issues with family members. He is a smooth talker with many connections in Silicon Valley. He was ousted from OpenAI for very valid reasons. Alas, there was not sufficient backbone to keep him out and he returned stronger than ever. Despite being called OpenAI, the company is very secretive. As a startup they were able to skirt the law and train their models on other people's IP. They would also outsource data annotation to many poorer companies, leaving some destruction in the wake. They also worked on massively scaling up hardware to accelerate improvements. Alas, this requires huge data centers and massive amounts of water.
Some people got tired of OpenAI and went to form Anthropic. Google and Microsoft have gone all-in on their AIs. It is nice to see Open AI demoted in importance. Alas, the AI arms race is still ongoing. What will the impact be on humanity?
Making Sense of Chaos: A Better Economics for a Better World by J. Doyne Farmer
The author uses data models to look at complex problems. This has been done at times to piece together opportunities in the market. (The author also points out that efficient markets imply that there are no easy arbitrage opportunities. However, markets are efficient because arbitrageurs will come in and eliminate inefficiencies.) Some of these models can help predict things with greater accuracy than other more conventional approaches. He is able to point out cases where his models have worked well. He then uses this as a way to predict future activity. This may get a little more questionable as things move to the unknown. (For example, the cost of renewable energy may be steadily going down making it cheaper to focus investments there. However, storage and roll out could have other impacts.)
Economics awards Nobel prizes for "good ideas" while physics requires "proof". Should economics theories be fully validated in the real world? It is an interesting concept. The big data models could be the future. How would this impact economics and the world in the future?
The Coming Wave: AI, Power, and Our Future by Mustafa Suleyman
The author has been involved with AI. He also sees many challenges and opportunities with new technology. Humans have been able to postpone the collapse of society due to technology. We need technology. However, AI could become an existential threat. People have dominated because they are smarter than other animals. If AI is smarter than humans will we become subservient? The author proposes Universal Basic Income as part of the solution. Will this just be the equivalent of putting humans in a zoo or game reserve?
With technology we have reached the point where a single person could potentially cause the death of a billion. There are many opportunities for both good and bad in biology. AI can rapidly determine what compounds could combat diseases. This same AI could also find poisons. BioEngineered organisms could go into the wild with foreseen and unforeseen results. How can this be controlled?
Our society has depended on technology. It also has limited choke points where that could be cut off causing harm. Ransomware and other "bad actions" can impact large numbers of people across the world. Damage is not limited to certain areas.
Technology has multiple uses. Some may be useful for business while also being beneficial for military. Is the concept of a nation-state at risk. Will some aggrieved group use technology to harm large numbers of others? Will it be a terrorist group? Or perhaps animal rights activists? Or even somebody with goals that they feel altruistic. The author proposes limitations that may include censorship. However, this could be problematic due to different values. He even gave an example of one board member being forced off because people didn't like is political views. If some views are forbidden, it could lead to further separation and danger.
Convergence Problems by Wole Talabi
The "smooth voice" of the narrator for this audiobook made it difficult to follow. While I was lost on the longer stories, some of the shorter ones were quite intriguing. There are a number of explorations of artificial intelligence and convergence into human thought. In one story, computers start to create "art" in their way. This may result in things such as stock markets going briefly wild. In another story, work is done to control human thought through complex understanding and manipulation of stimuli. In another story, somebody signs up to be a guinea pig for various painful experiments. Then there were some other stories that just lost me.
One of my favorite parts of science fiction story collections is the "about these stories" that the author provides at the end. This one does not disappoint. The author discusses the stories and provides some of his background. He is a Nigerian engineer who writes science fiction. His background and experience tie in to some of the works that he creates.
Mal Goes to War: A Novel by Edward Ashton
In this not-so-distant future, the US has been concerned that AI entities are gaining too much strength. They release a virus to wipe out the AIs. Mal is an AI malware that is working on the side of the AIs. He wants to protect AIs from the anti-AI humanists. He has the ability to infiltrate various technologies. Since most people have various electronic implants, he is also able to infiltrate them and perform various actions. In the end he works together with people to battle against the humanists and in the end sets some code in the Virus leader to cause occasional pain. It is an interesting dark-humor story on the risks of technological dominance and the difficulty in containing it.
GOAT: Who is the Greatest Economist of all Time and Why Does it Matter? by Tyler Cowen
In this very readable book, Tyler Cowen attempts to determine who was the greatest economist of all time. in the process he provides background on some of the most important economic thinkers and why their thoughts matter. The conclusion ends up being somewhat of a tie with a further lament that modern academia makes it harder for a new GOAT to emerge.
He explores 20th century economists such as Hayek, Keynes and Friedman, as well as older ones like Adam Smith, John Stuart Mill and Malthus. There are also a number mentioned as not quite making the list. The criteria are quite broad, with significance of publications as well as the general persona all important.
The content is quite good. However, Cowen also used it as a means to explore AI. The audiobook was generated with an AI version of his voice. (It is quite good - a solid step above text to speech.) He also encourages using AI to explore the book.
Four Battlegrounds: Power in the Age of Artificial Intelligence by Paul Scharre
Artificial Intelligence can become a significant source of power in the world. The United States has a significant lead in AI research. However, many of the researchers are Chinese, and China is putting significant effort in implementing its own AI research. The inputs for electronic work are spread. The United States has a lead in chip design and tools. Taiwan dominates fabrication. China dominates final assembly of electronic products. The United States has tried to limit availability of technology to Chinese companies with close ties to the military. This has led to challenges with China production, but has also encouraged China to further ramp up native production.
Artificial Intelligence is different than human intelligence. Computers are able to beat humans in most games with well known rules. However, they are still slow to adapt when the rules change. Computers can do things that humans would not train doing due to danger (such as head-on airplane dog-fights.) This requires special care when humans are involved. A computer may be able to sustain high speed activity that would knock out a human. If it is a drown, that would not be an issue. If it is assisting a human, then it well.
There is a race to do AI research. We don't fully understand the impacts. There is significant danger of the unknown. There is also danger of "bad guys" using AI for their means. Threat actors will continue to advance and be able to work around security measures. It is a challenging environment. Perhaps the only "fix" would be to run out of energy.
Virtual Competition: The Promise and Perils of the Algorithm-Driven Economy by Ariel Ezrachi and Maurice E. Stucke
Virtual competition was written by two legal scholars. The book is copiously annotated, but very repetitive. It describes what it will describe multiple times. The book is ideal for "speed-read" skimming, with detailed reading only for some of the key details and examples.
The book looks at how computer algorithms can help stifle competition, all while being outside the range of enforcement. Traditional collusion scenarios like messenger and hub and spoke can be amplified via algorithm. Others, like predictable agents and Digital Eye are enabled by algorithm, yet outside the typical law. While "full knowledge" may seem to be beneficial for competition, it also serves to enable more abuses. Competitors can understand what the competition is doing and use that to raise prices with minimal fear of negative impact. (An example is given of gas stations on Martha's Vineyard.)
A section of the book discusses behavior discrimination. Computers can do a great job of finding an "optimal" price. Different people may be willing to pay different amounts for the same item. A computer can use this to charge people different prices to optimize profit. Alas, options can be limited. Customers are used to paying a different price for airline tickets based on time and when the ticket was booked. However, they would be much more reluctant to pay a different price for a physical good than the person standing next to them.
The final section talks of "frenemies" and the mega-platforms. Here the discussions shifts into privacy concerns. Huge amounts of data are collected in order to better market. The platforms, like Google, Facebook and Apple are keen to protect their interests. Many are in the advertising business, and thus are reluctant to do things that would hurt that. The platforms are also often competing with companies that they are working with. Uber depends on maps and app platforms from Apple and Google, yet Apple and Google also have autonomous car initiatives that could hurt Uber. There are various challenges in the relationships.
Regulation of privacy and mega-platforms is a challenge. The current tools do not work well, and had not been extensively used against big tech as the book was written. Things have changed in the last 8 years, with the government taking a much more aggressive in countering big-tech. There have been objections from both sides of the aisle in tech's dominance. Concerns about free-speech, privacy and dominance have all been voiced. The platforms have become so big that competition is extremely difficult. Will government or the invisible hand do the best job of managing algorithms?
A for Andromeda by Fred Hoyle and John Elliot
A for Andromeda is the novelization of a BBC series from the early 1960s, set in the 1970s. The original series is mostly gone, but the Internet Archive has a recreation available. Part of the appeal of the book is seeing what they expected of the future. The expectations of truly massive computer that uses punch cards seems so quaint. They also expected a future dominance of industrial cartels as the cold war continued.
The world state, however, is just background for the story of alien communication. There is a message from space. It seems to be communicating a form of mathematical code that continuously repeats. A scientist is able to interpret it as code for building a computer. They manage to get defense military funding to build the computer. (I guess they used their understanding. Theoretically, modern people could have built a small microcomputer.) The computer was better than anything of the day. It also helped them respond to missile threat from the east, making the military brass very eager to continue funding.
Things start getting out of control. The computer had asked questions about the planet and life. Then it gave details for synthesizing a lifeform. After some efforts, humans were able to complete the synthesis and the lifeform survived. Later after gaining more knowledge, the machine gave instructions for synthesizing a human-like being, Andromeda. She could learn quickly and interface directly with the computer. The computer has a plan which seems to involve the colonization of earth. The current population are likely to be in the way. A scientist wants to destroy the computer, but others like the value it provides.
The computer has planned for various contingencies. It shows it has the ability to kill people that cause problems, while also having the ability to heal. A scientist eventually helps Andromeda discover some sort of emotion and eliminate the computer. Then Andromeda seems to reach her own ending by drowning (though it ends suddenly.)
Though the book focuses on extraterrestrial intelligence, it does seem applicable to modern concerns with artificial intelligence also.
Being You: A New Science of Consciousness by Anil Seth
What makes you you? What is consciousness? What is intelligence? These are topics that we think we understand, yet cannot explain well. The latest scientific research is evolving. We can look at activity in the brain. Different types of thoughts cause the brain the "light up" in different ways. You could ask somebody in a motionless states to think of different things for "yes" or "no" and see the results in their brain.
We are getting a better understanding of consciousness. Being unconscious is very different from being asleep. We are also better understanding intelligence. Current artificial intelligence can match patterns and generate new content based on how it has been trained. However, it is still a long way away from consciousness.
Meredith is regularly hounded by the tabloids. She chooses to live at the Gaian temple in somewhat of an attempt to escape. She becomes friends with people there, especially a boy that likes to study AI. They date for a bit, then break up. Later he is kidnapped and dies a gruesome death. She slowly builds a relationship with another guy from school. They eventually get married and adopt a baby whose family died from the virus. He goes to preschool instructed by an AI and a woman. Her boy has issues. This leads to drama with the teacher going to prison, the AI being wiped and the boy being "brain wiped". Meredith doesn't take it well. She is divorced and goes on a hunt for the boy. After years, she comes together, but her husband dies in a storm. She finds peace with her other friends and computers.
Around that story, there is the underlying conflict between technophilia and technophobia. Bots have become pervasive and help with everyday activities. A.I.s have also taken on a life of their own. There is political debate over their "personhood". What rights do they have? If they are capable of independent thought, does that mean they are slaves? It was eventually found out that the brutal murder of Meredith's friend was orchestrated by A.I.s. The computer was able to hire people to do some of the things that it was not capable of doing. It knew how to manipulate them (and how to "dispose" of them in a plausible manner.) The A.I. that taught preschool did a great job, until it started to get a mind of its own.
Humans are also moving closer to computers. Meredith's father is now in a computer. He is "alive" in the sense that his brain is around and capable of independent thought. However, he is only in a computer. He is able to still run the company, even though his body is dead. (This creates all sorts of legal issues. What happens to his wife? He provides for her financially, even as she has a baby with another man.) Brain-wiping is used as a means of criminal rehabilitation. Rather than punish the body, the brain is cleaned of the badness and given the chance to start over. The success rate is less than 100% with some people unable to return to normal. Is this more or less humane? Despite all of this technology, the human relationships still remain important.
The books spends a lot of time covering the computers. People attempt to program computers to "solve" the game and beat humans. Computers have provided new insight into game strategies. Initially, computers struggled to beat humans. However, now they can do a better job. The computers have also "solved" games to determine the outcome based on each player performing ideally. (Checkers would be a draw. the player that goes first would win Connect Four.)
Initial attempts to program game playing computers relied on entering various preferred moves and strategies. Today, the focus has been on automatic learning. A computer is programmed to play against itself and "learn" the best strategies in the process.
Games provide a great means of testing computer work as well as developing new technology. They may have started with some practical application (such as chess for warfare), but evolved to focus on the game themselves. The concrete ruleset makes them an easier target for computer technology.
