Reading Period: May 24 - Present
1. Discrete Mathematics and its Applications (P), by Kenneth Rosen
Link: https://www.goodreads.com/book/show/1800803.Discrete_Mathematics_and_its_Applications
In preparation for a placement exam, which if I passed I would have tested out of a Discrete Math course at the University of Chicago, I ended up reading a majority of this book. Unfortunately, the exam did not work out and I ended up having to take the course anyway, during which I re-read nine of the thirteen chapters multiple times. This meant that I did hardly any other reading over the summer, and also that I now have a pretty good handle on the topics covered in discrete mathematics. I would categorize this book as tough, but fair. The material is fairly challenging, but it is clearly an excellent introduction into the world of computer science. I would guess that there aren't any "better" books on the subject, but I also think it is clear that there are better ways to learn the subject than through a textbook.
2. Situational Awareness (P), by Leopold Aschenbrenner
Link: https://www.goodreads.com/book/show/214290546-situational-awareness
Probably the most thought-provoking book I have read in quite a few years. Leopold is clearly intelligent, and in general I share his outlook in broad strokes. I think he is too overconfident, as it seems fairly likely that scaling LLM architecture is not going to result in AGI. I listed to a podcast recently with Francois Chollet, the creator of Keras, who insists that LLMs struggle to generalize regardless of scaling. Leopold strongly disagrees, without proof, and claims that AGI by 2027 is the most likely scenario. Leopold claims to be in the "inner circle" of AI capability development, and he that he knows everyone of importance in the AI race, or at least is separated by no more than one mutual connection. He sees his previous investing prowess (longing NVIDIA in 2023 and shorting the market before COVID) as proof of his ability to call events, and he leans on his one-year tenure at OpenAI on their superalignment team for insider credibility. Leopold was let go by OpenAI and is now starting his own investment fund, at the ripe age of 22. He is too brash and overconfident, to a degree that I think harms his greater points. I don't trust someone who worked at an AI lab for a year when draws a straight line on a few data points and tells me that a curve is exponential, and I would guess that readers not ingrained in the same quirky social groups as Leopold won't buy many of his claims either. Personality quirks aside, what Leopold did with this book is extremely impressive.
First off, I think he legitimately changed my opinion on a few things, especially his points about the importance of avoiding a close AI race between the US and China. Leopold states that "superintelligence is a matter of national security, and the United States must win." Also, he rightfully points out that AI research will be the first main target of automation. Once we get AGI, why waste time with any area of development in the world except AI research, if more AI research will make smarter models that have better judgement? In addition, Leopold find the idea of Silicon Valley CEOs deploying superintelligence as rightfully ridiculous, and has extremely insightful views on the competitive dynamics within geopolitics and how history will likely repeat itself. I find his views as innovative as they are terrifying, but I think he is more or less correct. He is also fearful of AI use by totalitarian regimes, as he states:
"A dictator who wields the power of superintelligence would command concentrated power unlike any we’ve ever seen. In addition to being able to impose their will on other countries, they could enshrine their rule internally. Millions of AI controlled robotic law enforcement agents could police their populace; mass surveillance would be hypercharged; dictator loyal AIs could individually assess every citizen for dissent, with advanced near-perfect lie detection rooting out any disloyalty."
Leopold is a geopolitical realist, and he has his head firmly grounded in the history of war and competition we have seen play out over humanity's lifespan. I think perhaps the only downside of this book from a utilitarian perspective is that it could "wake up" China to the race dynamics happening, but it is probably more important that the dynamics of the new cold war be realized by those in the US early, even at this risk. What I find strange about this book is that I think Leopold might have crazy foresight. Sure, it could all be brash, youthful arrogance, but some of these ideas make almost too much sense. I am not going declare this book prophecy. But if Leopold is as smart as he thinks he is, he might as well be a prophet.
3. The Picture of Dorian Gray, by Oscar Wilde
Link: https://www.goodreads.com/book/show/5297.The_Picture_of_Dorian_Gray
"To get back my youth I would do anything in the world, except take exercise, get up early, or be respectable." One of the wittiest books I have ever read, and an altogether pleasant read. Lord Henry is certainly the funniest character, who goes on frequent, thought-provoking tirades such as the following:
- "Men marry because they are tired; women, because they are curious; both are disappointed."
- "Women, as some witty Frenchman once put it, inspire us with the desire to do masterpieces, and always prevent us from carrying them out."
- "The only way a woman can ever reform a man is by boring him so completely that he loses all possible interest in life."
- "Anybody can be good in the country. There are no temptations there."
Lord Henry believes that only the rich and interesting have social value, as "even the cardinal virtues cannot atone for half-cold entrees." He mentors Dorian Gray, a beautiful and empty man who says things like: "I have never searched for happiness. Who wants happiness? I have searched for pleasure." The book is half witty banter between Lord Henry and others, and half Dorian Gray doing bad stuff both because of and in spite of his beauty. This book took me a while to get through despite its short length, but I think it is well worth the read.
No comments:
Post a Comment