The past several decades of American economic growth and technological development have seen signs of stagnation. Total factor productivity growth is stalling, research productivity is declining, and the advances we do see in the world of bits often feel shallow. This is in contrast to the breakneck pace of world-altering innovation in the early and mid 20th century. Much of the progress in this time came out of focused research organizations within large and profitable American corporations like Xerox or AT&T. This decline in corporate sponsored research coincides with stagnating economic and technological growth. Large and profitable corporations in the 21st century are once again researching, inventing and commercializing the cutting edge technologies of our time.
Past
The most important inventions of the 20th century were invented, developed, or commercialized by large research labs run by American corporations.
“Much of the early development of the automobile culminating in the powerful Chevrolets and Buicks of 1940-41 was achieved at the GM corporate research labs. Similarly, much of the development of the electronic computer was carried out in the corporate laboratories of IBM, Bell Labs, and other large firms. The transistor, the fundamental building block of modern electronics and digital innovation, was invented by a team led by William Shockley at Bell Labs in late 1947. The corporate R&D division of IBM pioneered most of the advances of the mainframe computer era from 1950 to 1980. Improvements in consumer electric appliances occurred at large firms such as General Electric, General Motors and Whirlpool, while RCA led the early development of television.” -Robert Gordon
The research and development done by corporations in these and other fields were major contributors to high productivity growth in the 20th century. The mutually reinforcing relationship between scientific progress and productivity growth is known, but translating basic scientific research into a commercially viable product is extremely difficult. Corporate research labs have an advantaged position in this respect because the basic research that they do can be aimed towards solving practical problems from the outset and they can foster close collaboration between basic research and applied engineering. During the development of the transistor at Bell Labs, physicists and engineers famously tinkered back and forth; using nascent information theory and computer science to build experimental devices, and using the properties of those prototypes to build and edit the mathematical theories. Andrew Odlyzko, a senior scientist at Bell Labs said:
It was very important that Bell Labs had a connection to the market, and thereby to real problems. The fact that it wasn't a tight coupling is what enabled people to work on many long-term problems. But the coupling was there, and so the wild goose chases that are at the heart of really innovative research tended to be less wild, more carefully targeted and less subject to the inertia that is characteristic of university research.
Present
The corporate research labs of the 20th century did not last forever. Xerox PARC and Bell Labs have been spun off into separate companies and into irrelevance, DuPont’s Central Research & Development Lab closed in 2016 after a slow fall from the largest producer of chemical engineering research in the 1960s, and while fortune 500 firms won 41 percent of R&D magazine’s R&D 100 awards in 1971, only 6 percent were awarded to large private firms in 2006.
The 21st century has brought a new set of huge firms and a new set of technological goals that have reignited the corporate lab as a furnace for progress. The internal combustion engine, the transistor, and chemical engineering were innovations that defined the 20th century. Today, quantum computing, bio-tech, and machine learning are at our modern day technological frontier.
Research sponsored by or done within some of America’s largest corporations is at the forefront of these emerging technologies.
Google
Google’s DeepMind team is a global leader in AI research.
They developed the first neural net that could beat the world’s top players at Go, and they have developed a generally capable game-playing model that can learn thousands of games without large training data sets.
Beyond games, AlphaBrain’s neural net has been successfully applied to problems across dozens of domains including language models and translation, self-driving, and data center cooling.Google has also made significant contributions to bio-tech.
DeepMind applied their learning model to protein folding and released accurate predictions of hundreds of millions of protein structures to the world. Dennis Hassabis, the head of DeepMind said “This is a lighthouse project, a major investment in terms of people and resources into a fundamental, very important, real-world scientific problem.”
DeepMind has also helped to develop several diagnostic algorithms used in the British National Health System to catch early signs of blindness, breast cancer, and kidney disease.Google’s quantum computing project has advanced the field several times and it houses the world’s most powerful quantum computer. Recently, they have been able to isolate an extremely strange form of matter called a time crystal inside a quantum computer.
Google’s quantum computing team is also the furthest along towards commercializing the technology. They have already performed calculations in mere minutes that are impossible for traditional algorithms to do, and they plan to build a quantum computing cloud to sell compute power by 2029.Google’s research output is far larger than what I could summarize here. They have over 7,000 publications in a wide range of field. 192 health and bioscience papers, 259 general science papers, 89 papers on quantum computing, over 3,000 papers on machine intelligence and perception, and many hundreds more in other fields.
In addition to tons of basic research, google also houses a ‘moonshot’ team under the name ‘X’ which tries to bring projects from the theoretical fringe in all of the above fields into physical reality.
Amazon
Not to be outdone by their big tech rivals, Amazon also upkeeps a large and productive research team. They started in 2015 and already have over 1,100 publications in fields such as machine learning, quantum computing, and computer vision.
Amazon is also building a quantum computer and is already scaling up operations to begin offering quantum computing as a service, completing Amazon Web Services.
Although their internal research team is smaller than Google’s, Amazon also funds hundreds of research awards that it gives, largely without strings attached, to individual researchers at non-profits and in academia.
Tesla
The electric car company is also at the forefront of machine learning methods as part of their implantation of self-driving cars. As demonstrated in their recent AI day presentation, they have made huge forward strides by constructing a neural net that can predict and plan directly in a birds eye view of the car, rather than from several separate image feeds.
They have also developed advanced auto-labelling and simulation tools to speed up and facilitate the training of any type of neural network, including self-driving.
Tesla created a computer architecture called Dojo which is purpose built for training neural nets on billions of situations at once. This rapidly scalable cloud computing service will allow much faster and more comprehensive training and compete with Google and Amazon’s cloud computing services.
Although not directly associated with tesla, Elon Musk’s other projects are relevant to the idea of privately funded corporate research.
Elon Musk is the principle investor in OpenAI which, like DeepMind, has been at the forefront of AI and machine learning research. Their most influential work has been in natural language processing models. Using an assortment of neural nets, OpenAI can translate natural language into working code, illustrated or photorealistic images, or even extend essays and songs.
Elon Musk is also contributing to Bio-tech through Neuralink, which actually shares office space with OpenAI. This venture is creating brain-machine interfaces that can read and interpret the electrical signals created by biological brains. Their device already allows a monkey to play pong with its mind. Neuralink is also developing a surgical robot in parallel to assist in installing the device but advanced surgery robots could be used in any operation.
Similar lists could be constructed chronicling the AI and VR research at Facebook or Microsoft, the quantum computing at IBM, self driving car research at Apple, bio-tech research at Moderna and many more. To be clear, the majority of publications in these fields still come from academics at research universities. However, this was true during the golden age of the corporate research lab in the mid 20th century, so producing a majority of research output is not necessary to achieve large economic impact. Additionally, the research output of these companies has been growing over time. Google published 20 times more papers in 2019 than in 2005. Amazon has grown their publication rate by more than 10 times in just 5 years.
Just as in the 20th century, large companies are sponsoring cutting edge research in specific fields in a bid to outcompete their rivals and advance human knowledge.
Future
Our modern century-defining technologies have the potential to be at least as world-changing as those that came before, and they could likely far exceed the impact of any 20th century technology.
Quantum computing promises new discoveries in physics and further knowledge into the fundamental nature of information. Quantum computing is innovating out of the particle physic’s high energy problem where new discoveries are locked behind massive capital investments into bigger particle accelerators. Just like the fundamental physics discoveries of the 20th century, advances in our basic knowledge of the world are a necessary foundation for technological progress.
Bio-tech uses the power of bits to understand and control the world of organic atoms and molecules. The combination of huge compute power, mathematical modelling of genetic molecules, and advanced software tools allowed scientist-entrepreneurs at Moderna to synthesize a 95% effective Covid vaccine in two days. Further developments in software powered vaccines and a greater understanding of protein production will accelerate our transcendence above the constraints of biology; curing disease, extending life, and improving farming.
Last but not least: machine learning. This is a broad field with many different applications and visions for its future. In general, machine learning may allow us not only to relieve ourselves of repetitive and dangerous tasks like driving, but may also industrialize knowledge work and greatly accelerate the rate of economic growth. The fact that there is so much debate over AI alignment problems evinces the widespread agreement that AI and machine learning are powerful tools with growing importance to our economy.
Corporate involvement in the research and development of these technologies is necessary for them to start showing up in total factor productivity numbers. If the frontier of human knowledge stays within university departments then advancing it won’t show up in GDP. Google using DeepMind to make their data centers more efficient, however, is a direct pipeline from basic machine learning research into increased productivity from a fixed capital stock.
Our meandering off of the 2% growth trend of TFP is a tragedy. Understanding how this pace of growth was achieved in the past is essential for continuing it in the future. An important part of TFP growth in the 20th century was basic scientific research and subsequent translation of that research into productive uses. Large corporations play an important role in this relationship because they can bring both steps under a single organization; promoting collaboration and quick turnaround from scientific discovery to TFP enhancing product. Although the corporate research labs of the 20th century have fizzled out and will not return, new forms of private research and development are performing basic research and advancing our material well being.