This week we will explore a new theory about how gravity created light in the early universe. We meet an AI scientist that is now able to make some scientific discoveries on its’ own using known scientific theory. We investigate a new way to use laser beams to fix construction problems early in the build. Finally we discuss the latest edition of Packy McCormick’s Not Boring in which he addresses the issue of what will happen to our employment when all of these new AI systems start to be used more deeply in the workforce.
Gravity can create Light
Gravitational waves are literally waves of Gravity traveling through the universe. See this video explainer of gravitational waves. We have measured gravitational waves that were created when black holes collide or when they rip stars apart as they are consumed. The detectors that measure these waves need to be able to measure distances less than an atomic nucleus. In the extremely early days of the universe these waves were likely much stronger and larger.
The initial universe was characterized by a short period of intense inflation. After the Big Bang the universe went from basically nothing to something many orders of magnitude larger. This event is thought to have taken place during the first second of the universe’s existence. At the end of this inflation the universe was a very messy place. Gravitational waves sloshed back and forth through the cosmos.
We have all been on a swing. We want to go higher we swing our legs. We have to swing at the right rhythm to gain height. If we go off this rhythm the swing stops going higher. This is know in physics as parametric resonance. This type of resonance happens all over the universe every day when different types of waves collide.
At the end of the inflation of the early universe the gravitational waves stopped traveling but they stood still due to the rhythm of the overlapping. The places of the greatest overlap represented an exceptional amount of gravity.
This is thought to have had major consequences for the electromagnetic field in early universe. The regions of the most intense gravity may have excited the electromagnetic field enough to cause it to release some of its’ energy in the form of radiation, creating light.
There is nowhere in today’s universe where gravity could cause this process to happen. The early universe was a very different place.
AI Scientist
We have spoken a lot about AI and in all likelihood we will be speaking about AI more and more in the future. Teams at IBM, Samsung and the University of Maryland have developed an AI that has been able to reproduce the work of American chemist Irving Langmuir.
In 1918 he published a paper examining the behavior of gas molecules sticking to a solid surface. He developed a series of equations that describe how much gas will stick given a certain pressure. The AI also rediscovered Kepler’s third law of planetary motion which calculates the time it takes one object in space to orbit another given the distance between them.
The new AI scientist is called AI-Descartes (remember it was Descartes who postulated “I think therefore I am”). It is one of a new breed of AI tools such as AI-Feynman that are being used to speed up scientific discovery. The systems are built on a process called symbolic regression which finds equations to fit data. Using the basic operators, addition, subtraction, multiplication and division, the systems can generate hundreds of millions of possible equations. The system will then search for the one that most accurately describes the relationships in the data.
The most distinctive new feature in AI-Descartes is the ability to logically reason. From multiple candidate equations that fit the data the system can identify the equation that best fits with scientific theory. This ability to reason sets it apart from other generative AI systems (ChatGPT is a generative AI).
The system is particularly well suited to real world data where errant data can trip up traditional symbolic regression programs. It is also well suited to very small data sets, sometimes with as little as 10 data points.
The team is now working to create new data sets that contain real measurement data and associated background theory to refine their system and test it on new applications. Currently human experts are needed to write down theories in machine readable terms. If the human gets anything wrong the system will not work.
Eventually the team would like to train computers to read scientific papers and construct the background theory themselves. This would allow the machine to construct the theory, assemble the data and work on solutions.
Using Laser Beams to find Construction Problems
A team in South Australia have developed a new eye tracking technology that will help identify building defects early in the construction process. This will save significant time and money in rectification work. Some estimates of the cost of rectifying mistakes are as high as 60% of total cost.
.The new technology is embedded in 3D headsets. It is designed to help construction workers complete a more through checklist cutting down on rectification requirements later in the building process. The tool combines building information modeling and eye gaze data that is captured in a standard building inspection. The augmented reality headset sends a laser beam out of the bottom of the user’s eye level to track where they are looking. This is matched against a 3D model of the building and a checklist validates the completeness of the inspection.
The earlier that errors are identified the quicker and cheaper it is to fix. In the future it will be interesting to see this type of technology applied earlier in the construction process. On going AI based validation of work as it happens so that errors are not made in the first place. Over time robotics will take over more and more construction work and these type of systems will become an integral part of the construction process.
Excess Intelligence
I am going to do something a bit different this week. Rather than add another story about a new technological development, I will talk about one of the other newsletters that I read. Packy McCormick in his Not Boring newsletter and podcast addressed the issue of Intelligence Superabundance.
We have all seen the recent rise of what appear to be intelligent AI’s. This increase in the available (cheap) intelligence has led to intense speculation that all our jobs are going away. Some estimates are up to 50% of current jobs will disappear. I have always doubted that this will be the case. In the past when technology changed we adapted and new jobs emerged. When the car was invented it did away with buggy drivers and horse whip manufacturers however a massive new industry emerged and there was tons of new employment.
Packy addresses the issue of this new abundance of intelligence by looking at Induced Demand. When a new freeway is built it initially speeds up traffic as there is more road space than cars wanting to use the space. Slowly but surely the traffic increases as the Induced Demand for travel fills up the road and eventually we are back to where we started.
I for one, have never heard a complaint about the excess of intelligence in our society. In fact it is always the opposite. Additionally whenever we have increased the intelligence of society, for example by increasing the number of University graduates, the available work has increased to make use of that intelligence. This is an example of latent demand being filled when the supply is increased.
In the 1980’s approximately 16% of the US workforce had a University degree or higher, in the UK it was 10% and in Australia 8%. By the 2010’s the numbers were US 32%, UK43% and Australia 30%. A huge increase in the educated workforce however the unemployment rate did not vary much. In the 1980’s the unemployment rate for University graduates ranged between 2 and 4%, in the 2010’s it ranged from 2 to 6%. The situation was similar in the UK, 80’s 5 to 10%, 10’s 4 to 10% and Australia, 80’s 2 to 5% and 10’s 4 to 7%.
Despite the huge increase in the available intelligence of the workforce there was a minor variation in the employment rate. These variations were more due to economic circumstances than the excess availability of intelligence. BTW where did I get this data, ChatGPT (it is so much quicker when you know how to prompt it verses a google search, AI will make us all productive if we know how to leverage it).
Packy goes into far more detail than I have here. He examines a number of economic theories and the potential impact of an excess of intelligence from AI. The essay is well worth a read or listen.
Paying it Forward
If you have a start-up or know of a start-up that has a product ready for market please let me know. I would be happy to have a look and feature the startup in this newsletter. Also if any startups need introductions please get in touch and I will help where I can.
If you have any questions or comments please comment below.
I would also appreciate it if you could forward this newsletter to anyone that you think might be interested.
Till next week.