Jonathan Bartlett

Senior Fellow, Walter Bradley Center for Natural & Artificial Intelligence

Jonathan Bartlett is a senior software R&D engineer at Specialized Bicycle Components, where he focuses on solving problems that span multiple software teams. Previously he was a senior developer at ITX, where he developed applications for companies across the US. He also offers his time as the Director of The Blyth Institute, focusing on the interplay between mathematics, philosophy, engineering, and science. Jonathan is the author of several textbooks and edited volumes which have been used by universities as diverse as Princeton and DeVry.

Archives

The Backdoor to Control the Internet

We almost lost the Internet last week, but open-source developers saved the day.
Few people are aware, but over the last several days, a perceptive developer foiled a multi-year plot to install a remote backdoor into, well, the entire Internet. Two years ago, a programmer known as Jia Tan (JiaT75) started helping out with a lesser-known compression library, known as xz. For those who don’t know, software today is not a monolithic entity. Every piece of software you use it built from a collection of tools, known as libraries, that make programming easier. For instance, most programmers never have to write the specifics of a sorting algorithm, because, somewhere, there is a library which performs sorting for them. This leaves programmers to focus on higher-level tasks, like actually making the software do what the users want. However, these libraries don’t

Autonomous Vehicles Are Catching Up Fast

There are winners and losers. GM's Cruise was kicked out of California over safety issues but Alphabet's Waymo, which emphasizes safety, is still chugging along
Waymo wisely focused on Level 4 urban self-driving, which is an engineering problem, rather than Level 5’s total autonomy—that’s a philosophical problem.

Framework for AI Legislation

Unfortunately, current calls for AI legislation seems to be largely motivated by fear of the unknown rather than looking for specific policy goals.
I am going to lay the groundwork for what I think good AI legislation will be. However, before I do that, I want to give some cautionary advice.

Copyright in the Age of Artificial Intelligence

What exactly is a human and how does a human differ from a computer?
On December 27, The New York Times Company sued Microsoft and OpenAI for violations of their copyright. The Times contends that training chatbots on its content in order to create an information competitor is a violation of its copyright. This suit is sure to bring up a number of old copyright issues that were never resolved, plus some new that need to be worked through. The fact is, the big search engines have been violating copyright from the very beginning. All search engines are in fact derivative works of the sites that they crawl, index, and dish out. Most search engines even provide excerpts from the sites they scan. However, most copyright holders have turned a blind eye to this for two main reasons — nobody has deep enough pockets to fight big tech and most of the people

Directed Goals in Living and Evolving Systems

Nearly every action that an organism does is for something.
 Can this teleonomic behavior of evolution simply be a byproduct of non-teleological forms of evolution? Information theory suggests that this is not likely.

Why Build Process Automation Matters

Automated build processes allow for the standardization and systematization of your development pipeline.
Whether your development organization is a single individual or a large team, automated build processes provide numerous benefits to your group.

Of Infinity and Beyond

What are the problems and solutions with infinity in mathematics?
The concept of infinity has plagued a great many proofs, both formal and informal. I think that there are two foundational problems at play in most people’s thinking about infinity that causes issues. The first problem people have with infinity is that they treat it as if it were a single value. Because infinity is bigger than all possible natural numbers, people assume that it is bigger than any number, and therefore there is nothing beyond infinity. Therefore, people have the concept that if I have two infinities, then I still have the same number.  They believe that 2 * infinity = infinity. However, using that logic can quickly lead to contradictions. This problem is exacerbated by much mathematical notation. People often will use ellipses to indicate that

Why Is Object-Oriented Programming Popular?

This method makes programmers think more systematically about their code
Programming practice has gone through several evolutions in its lifespan. The first phase might be considered the “exploratory” phase, where there were no rules but a lot of imagination. People wrote code that was simultaneously amazing and terrible—amazing at what people got their slow computers to do, but terrible in that no one but the author would ever be able to maintain the programs. The lessons learned from the exploratory phase led to what is known as “structured” programming. The goal of structured programming was to be able to write programs that someone else had a chance of reading and understanding. Structured programming favored having really well-documented inputs and outputs to every function, very clear entry and exit points to each function, and

The Microservices Controversy from a Software Management Perspective

As projects get bigger, so do the reasons for having a microservice architecture
A new report by Amazon has caused a bit of a stir on the Internet. In it, the Amazon Prime video team reported that changing their architecture from a microservice architecture to a monolithic architecture resulted in a 90% cost savings.  While the report itself was very mild (its only claim was that this architecture helped in this specific situation), it has caused the people who disliked the microservice trend to make some noise of their own. Here, I wanted to take a moment to reflect on what I see as the benefits of the microservice approach from a software development management perspective. If you are not familiar with microservice architectures, you can find out more information in my book, Cloud Native Applications with Docker and Kubernetes. First of all, I do

The Raspberry Pi Phenomenon

A Raspberry Pi is a full computer that is not much larger than a credit card, but still packs enough power to be usable as a desktop computer
For the uninitiated, the Raspberry Pi is a single-board computer that runs the Linux operating system. It can be either operated as a desktop computer or as an embedded system (i.e., a custom electronic device), or both. Historically, computer systems were either general-purpose computers or embedded systems. General-purpose computers required too much hardware, too many chips, and too much power to work inside an electronic device. However, as manufacturers packed more and more functionality into less and less space using less and less power, eventually it became possible to have a computer that was small, cheap, powerful, and not especially power-hungry. The Raspberry Pi came about right as this was happening. A Raspberry Pi is a full computer that is not much larger than a

Aren’t US Treasury Bonds Supposed to be Safe?

How can you lose money selling treasury bonds?
For context, read Bartlett’s two previous articles on the fall of SVB and interest rates. Some people are confused as to how you can lose money selling treasury bonds, since they are supposed to be “safe” assets (the government is not expected to default on its loans, and, if it does, the economy probably has bigger problems). Economist Bob Murphy put together a great explainer thread on Twitter, which I will largely follow here. Let’s say that there is an asset that always yields a 1% return every year on however much you have invested, but you never get the principal back except by selling it to someone else. Let us call this asset ABC, and let us say that it is completely riskless. If you buy $1,000 in ABC, you will get $10/year forever. You have essentially

Why Did the Tech Bubble Correspond with Low Interest Rates?

Ultimately, our economy’s deeper problems aren’t so much a result of “money” as they are bad allocations of resources.
For context, read Bartlett’s article explaining the fall of SVB here. I wanted to make a quick note about why tech bubbles tend to correspond with low interest rate environments. Interest rates essentially dictate how long someone can wait before they need to produce something of real value. In a 20% interest rate environment, it will be evident really quickly if you are failing to produce something of value. Since essentially 1/5 of your capital disappears each year, if you aren’t doing something that will generate real profits quickly, you are sunk. You can’t paper over problems with more borrowing because the cost of that borrowing is so high. Additionally, in such an environment, the payoffs for investment need to be large in order for someone to invest. No one is

What’s Going on at Silicon Valley Bank?

The bank's failure is making a lot of people nervous about their money
Many people awoke this morning to news of a bank that suddenly collapsed – Silicon Valley Bank, or SVB. While information is still developing, I thought I would provide some background information on what is known so far. SVB is the go-to bank for Silicon Valley startups. Over the last few years, the tech bubble has been growing and growing and growing, focused especially around Silicon Valley. That meant a lot of banking was happening, and it was happening with SVB. That is where the various companies put their deposits.  How does a bank make money? By lending out deposits. In 2021, at the height of the tech bubble (and, not coincidentally, at a historically low-interest rate environment). The bank did what most banks do and bought very safe

The Need for Accountability in AI-Generated Content

Just because we live in a world of AI does not mean we can escape responsibility
AI-generated content has become increasingly common on the web. However, as we enter this new era, we will need to think through the moral and social ramifications of what we are doing, and how we should negotiate the new ethical landscape. But first, a brief recap of recent history. The first major player to pioneer AI-generated content was the Associated Press. AP realized that many market-oriented articles were pretty monotonous and read like templates anyway, so they decided to fully commit and auto-generate many of them. If you read an AP story about a company’s earnings report and it sounds eerily like every other story about other companies’ earnings reports, there’s a reason for that. Templated content, while annoying, provides window-dressing to raw data, and

Whatever You Do, Don’t Ask GPT for Sources

The chatbot will give you a lot of links that don't necessarily direct you where you want to go
One of the more amusing things I’ve found from OpenAI’s GPT-3 and ChatGPT is the fact that it will very confidently provide you with sources on anything you ask—and they will often be completely made up. It will even provide fake (but real-looking) URLs for you! I stumbled across this feature when researching a previous GPT-3 article about how well it could write blog posts compared to real authors. I initially tried asking GPT-3 to include sources, and it generated complete nonsense for the sources. I decided that, for that article, sources were not the main question, so I left it out of the final queries. However, in response to my latest article about ChatGPT not being a Google replacement, someone commented that, in the future, it would be simple to expand

Why ChatGPT Won’t Replace Google

With Google, the algorithm eventually leads you to content made by real people. With ChatGPT, you never leave the algorithm
To some extent, ChatGPT is a newer, easier-to-use interface than Google.  Unlike Google, it doesn’t make you waste time by visiting those pesky websites.  It not only looks into its database for content, but it also summarizes it for you as paragraphs. There is a problem lurking in there, however.  Being computers, neither Google nor ChatGPT cares about the truth.  They are algorithms, and they merely do as they are told.  Additionally, you can’t code the human mind into algorithms.  However, there is a fundamental difference between what ChatGPT does and what Google does that will prevent content generators like ChatGPT from displacing search engines like Google: Google eventually lets you out of its system. Ultimately, the goal of search engines like

GPT-3 Versus the Writers at Mind Matters

How does the AI fare when it is asked to write on topics covered in Mind Matters articles?
In order to give a real-world comparison of the output of GPT-3 to human-written writing, I decided it would be a fun activity to see how OpenAI’s GPT-3 compares to Mind Matters on a variety of topics that we cover.  Here, we are using OpenAI’s direct API, not ChatGPT, as there is a lot of evidence that ChatGPT responses have a human-in-the-loop.  Therefore, we are going to focus on the outputs from their API directly. I used several criteria for article selection in order to even the playing field as much as possible.  For instance, I only chose articles that did not depend on recent events.  This way, GPT-3 is not disadvantaged for not having up-to-date material.  However, I also chose recent articles so the articles themselves were less likely to be