Stephen Wolfram

This short book contains an essay about how neural networks work in large language models. The author conveys his enthusiasm for the recent human-like performance of ChatGPT in text creation. He also asks some interesting speculative questions about what this might mean for language theory. Could there be hitherto unknown scientific laws of language that large language models have now learned to tap into? I don't know if I learned all that much about what ChatGPT is doing, but this essay is show more nevertheless an interesting read about new developments which might in the long run be more revolutionary from a scientific perspective than we have yet realized.

At the end of the book there's a shorter essay about how ChatGPT could be married with the author's own Wolfram Language. This one is quite confusing - the author just repeats again and again that great benefits could be reaped by marrying these two systems. But he does not provide any interesting examples of these benefits. The point may be valid for all I know, but I'm not sure why he felt the need to publish a bad essay about it. The most plausible explanation is that this second essay was actually written by ChatGPT. show less

Fine job of showing off features of the Wolfram language. Many of which incidentally require an active internet connection to access the Wolfram Cloud and potentially require an additional logging in for the really neat-o ones, go figure. If I hadn't had previous functional programming experience, I would have been completely lost, because the Wolfram syntax contains some of the ugliest line noise this side of legacy Perl. At times the exposition is overly terse in an effort to show how very show more simple everything is. The decision to delay introducing how to write a function until chapter 40 is a bit odd, but the philosophy of the book is to use one of the scores of existing library functions. Which would be fine if I didn't get the feeling it was all part of an Embrace, Extend, Extinguish strategy for creating a legion of programmers dependent upon the Wolfram Computational Engine, which will, of course eventually be merged with Stephen Wolfram, the person, to create Wolfram the Computeror. show less

We are barely out of the period where we thought the Earth perched on the back of a turtle, was flat, was at the center of the universe and that we had discovered everything there was to discover without even knowing about atoms. Do we now suddenly know enough to be able to propose a unified theory of physics and therefore the universe? Do we now know every particle and its role?

Stephen Wolfram has a kind of workaround for those questions. Decades ago, he got waylaid by the potential of show more computers and the joys of discovery and invention in that field. Now, having just turned 60, he is coming back to his roots (He published his first physics paper at 15), throwing the doors wide open to tackle the holy grail – the underlying, fundamental theory of physics that dictates how the universe works, why it looks and works the way it does, and where it goes from here.

In his nearly 800 page announcement, called A Project To Find the Fundamental Theory of Physics, Wolfram expresses his delight to be able to do this now. He says had he tried decades ago, he would not have had the tools to launch or succeed in this effort. He created those tools himself, for other purposes. And along the way, he picked up a better understanding of what he faces: “We won’t be able to get even close to running those models for as long as the universe does. And at the outset it’s not clear that we’ll be able to tell enough from what we can do to see if it matches up with physics.”

He drops some important hints as to what has changed in his mind:
-“General relativity and quantum mechanics are basically the same thing.” This solves a huge conflict in modeling, and Wolfram “proves” it using simplified examples of causal invariance that by definition, end up branching back to the same state. This conveniently eliminates any potential conflicts.
-“I have an estimate that says that 10200 times more ‘activity’ in the hypergraph that represents our universe is going to ‘maintaining the structure of space’ than is going into maintaining all the matter we know exists in the universe.” So the universe is largely on maintenance.
Let those points guide your own theory attempts.

The book is about the fundamentals of modeling. He shows with great simplicity that a rule for a variable produces astoundingly complex images if repeated often enough. The book is festooned with an immense variety of results, from the basic rule to the first simple iterations, the point where before or after the very next step takes it into a new level, and then a 3D model of the result. If only for the graphics, the book is gorgeous. Wolfram makes great use of color to show what has changed from iteration to iteration, how structures differ, and to highlight what he talks about.

It is also an inspiration. If such simple rules (For every occurrence of AB, substitute BA) produce such sophisticated structures, there must be a way to structure a universe somewhere in there. In typical Wolfram fashion, he is opening up the treasure chest of tools and data to all comers. His meetings are livestreamed and recorded. He is making the collection of rules experiments, called notebooks, available to all, and allowing anyone to create their own using his tools. One way or another, Wolfram would love to see someone make a major breakthrough in the ultimate quest of physics. Starting with this book is a pleasant way to begin.

It could have been dauntingly technical, but Wolfram has made it appealing to all comers. Readers do not have to absorb the formulas or ponder the origins of the universe to enjoy it. The sections are color coded with tabs on the outer corners. It is actually a fast read. My only complaints are tiny type and graphics (This is the first book I have read with magnifying glass handy instead of a search engine), and paper so thin that images from the back side of the page are clearly interfering.

For this project to work, Wolfram simplifies by saying he thinks the universe is just a computer, carrying out formulas and rules. It’s all routine and predictable. He says the only thing special in the universe is ourselves.

David Wineberg show less

There's an old warning passed down in STEM circles. Back when the loom was the most advanced machine on the planet, the leading metaphor for how brains worked was as a loom. As telephones began being strung across the country, the this metaphor shifted---now the brain was like a telephone switchboard. These days the brain is like a computer, but I'm reasonably sure it's going to stay a computer. My point is that our familiarities inform the metaphors we use, but it's worth keeping in mind show more *that these things are just metaphors.*

Wolfram falls victim to this. Not only are brains computers, but they are in fact /computation itself/. Intelligence is merely computation. Weather systems are merely computations. Thus, Wolfram says, weather systems are intelligent too. This is a neat trick of semantics, but it's ultimately useless. The word "intelligence" refers to human-like-things, and not to weather-like-things, regardless of any computational similarities they have under the surface. Wolfram eventually concedes the point, but it left a sour taste in my mouth. If he's going to argue in clear circles like this one, why should I trust his reasoning on other things where I find the causal relationships less clear?

"Everything is computation" is the claim, and Wolfram follows this argument to its limit---that Godelian proofs are thus a limiting factor in every endeavor. We're unable to predict the future because of the halting problem. We're unable to distinguish meaning because doing so is equivalent to solving the halting problem. Et cetera.

This all may be true, but it seems like grasping at straws from a man who has this to say about mathematics: "what I’ve concluded is that actually the mathematics we have today is really just a historical accident: the direct generalization of the arithmetic and geometry that happened to be used in ancient Babylon. So it’s just history that makes the particular axiom systems we’re using seem meaningful to us."

Yes, the halting problem is a very real phenomenon, but the vast majority of the time it doesn't strike in full generality; we can often approximate solutions. And, this is all based on the assumption that the universe itself is subject to Curry-Howard. Maybe, but then again the only evidence we have is that there don't appear to be any NP-complete problems in nature.

Computation and the Future of the Human Condition isn't all bad though. It's a short enough read that you can get through it in one sitting, and it'll definitely provoke interesting thoughts. That being said, it's not Wolfram's best work. A better read is his blog post Showing Off to the Universe (http://blog.stephenwolfram.com/2018/01/showing-off-to-the-universe-beacons-for-the-afterlife-of-our-civilization/) which better details his arguments and is free. show less