Brian Eno’s Apprentice
Or; What Fred again.. teaches us about what it means to do good science
It's likely that in the last year you have heard of the artist Fred again..1. If you haven't, you should listen to his Tiny Desk concert with NPR for a sense of melodic pop/rock, or to his Boiler Room concert for dance music. If the genre doesn’t resonate with you but you do care for stories of scale and success then you should take a look at his listeners over time.
Within four years of the release of his first solo album in 2020, Fred has become one of the most listened-to musicians in the world. He sells out multiple nights at massive stadiums in every city he visits and has headlined the largest music festival in the United States.
Even before this successful solo career started, Fred had already become a prolific producer working with Ellie Goulding, Charlie XCX, Stormzy, Ed Sheeran and others; producing a third of the number-one singles in the UK in 2019; and being the youngest producer ever to win Producer of the Year at the BRIT Awards at 26. The only way to explain how Fred managed to achieve all of this so fast seems to be that he is good at music (something that becomes clear the moment you listen to his compositions). What exactly does that mean? And, how does someone get good at a skill as multifaceted as something like “music”?
The answer in Fred's case was a blend of traditional classroom training and apprenticeship, a pattern we have seen before with the techniques employed by the Bauhaus. Fred's case, though, is instructive, not only because it shows how apprenticeship is a powerful way to teach skills, but because it has lessons more broadly for what it means to become capable in fields that, like music, are complex and amorphous The implications are significant for both art and science.
The Training of Genius
Fred grew up playing classical music, and there are stories of him recording his own pieces on an aunt's tape recorder when he was as young as eight years old. He attended Marlborough College in Wiltshire, a prestigious boarding school with a history of educating great talent, including the musicians Nick Drake and Chris de Burgh, the creator of “The Lady in Red.” There, he received a top-tier formal education in classical music, and he would often skip his other classes to spend more time practicing in the music rooms. As a result, by the time he was partway through his teenage years, he had as strong a foundation in formal musical training as you could find anywhere.
It was at this point that he experienced a quirk of fate that transformed his trajectory.
One day, while with his parents in Notting Hill, a family friend invited him to join an a cappella group rehearsal at a neighbor's house. This neighbor was Brian Eno.
Brian Eno is an essential node in the history of modern music. He is credited as the inventor of the genre of "ambient" music starting with his album Ambient 1: Music for Airports. He was also a pioneer of "generative" music, well before the current wave of generative AI; he would use samples and musical snippets algorithmically to “generate” new sounds and create novel combinations. And he developed Bloom, the first generative music app for iPhone, three months after the app store went live.
Beyond these experiments with novel forms Eno played an important role in popular music: mentoring John Cale, the founder of the Velvet Underground; producing tracks for the Talking Heads; and collaborating closely with David Bowie and U2.
And while his own music may not have reached millions, he got there another way by creating the Windows 95 startup sound.
In the early 2000s, Eno hosted weekly music-making sessions at his home and Fred quickly became a fixture. In this setting, he was exposed, consciously or not, to the kind of training plan that has been the standard for apprenticeships for hundreds of years.
At first, he would simply help out with the events, cleaning up, getting tea, arranging song sheets, making sure that everything was going smoothly. Over time, though, he started to participate more and more in the music creation sessions, and he enjoyed privileged access to Eno's mentorship. He would pitch Eno ideas to test out new approaches and received feedback from the master: “I’d then not sleep for six nights making 100 song sketches. ... I’d come back the next week and try to make it look really casual.”2
“The fundamental learning situation is one in which a person learns by helping someone who really knows what he is doing." - Christopher Alexander
The most important training moment came when Fred was nearing the end of school and Eno asked him to help produce two collaboration albums he was working on. Help is very different from instruction. Rather than learning in a protected environment, you work in the real world to create something that effectively solves real problems but—in the case of an apprenticeship—with the oversight of someone who has developed taste and technique and can correct you with granular feedback as you make mistakes. Fred seized the opportunity and dropped out of school to work with Eno full time.
As a result, before he was 20 years old, Fred had received an in-depth education in music theory, practiced all manner of music creation in an environment with rich feedback, and had also seen from Eno exactly what it looked like to work with a team and create an album in the real world.
But it would be wrong to give the impression that Fred simply acquired a set of explicit skills from Eno and his teachers and has since been applying them formulaically. It is true that there are echoes of Eno's influence in Fred's music. Many of Fred's songs use building drones that evoke a strong sense of place, a technique that Eno pioneered. Fred also has a habit of drawing inspiration from sounds that wouldn't traditionally be considered musical, using samples of his friends and the world around him—another Eno quirk.
But at the same time, Fred and Eno are fundamentally different. Eno, while influential, has focussed his efforts on experimental music, not on generating attention or filling seats. Fred makes music for mass audiences, is an adept stadium rockstar, and knows how to take advantage of viral marketing. And, musically, Eno has said he has learned a lot from his student, including Fred’s use of "non-linear" loops within his music. It seems that Fred is just good at music.
It's possible that Fred has some genetic predisposition to musical talent as a result of some wild fluke but it appears more likely that he somehow learned it from the environments he was brought up in. You can see similar stories in other fields. Picasso, for example, was trained from a very young age by his father, a full-time commercial painter, in how to paint classical pieces, so that by the time he was a young teenager, he was already winning prestigious competitions and painting pieces that to this day hold up with the finest classical works. Yet within a decade, he had traversed multiple other styles of painting and played a determining role in the new style of Cubism, transforming visual art forever.
It seems unlikely that the later work would ever have been possible without his early training. Yet you cannot find in the training itself an articulation of the skills needed to produce this new kind of art. Fred is just the most recent example of an old pattern, of early training by a master leading to competent practitioners who consistently break new ground. The pattern though isn’t simply limited to artists; it is just as relevant to scientists.
Science as Art
We generally think of science as proceeding algorithmically from the “scientific method.” There is a hypothesis within some field; a scientist runs a set of experiments to test that hypothesis; the results of the experiments are reviewed; and, finally, it becomes clear whether that hypothesis has some validity in describing the world. In our folk idea, this is a procedure that continues step by step. We run through the given hypotheses, and over time our knowledge of the world advances in a piecemeal fashion. The truth, however, is that this description glosses over many of the fundamental uncertainties that characterize how science is done.
The uncertainties start with how we choose which hypotheses to investigate. The range of possible hypotheses is at least as large as all potential ideas about how the world works. Even if we had vast computational ability and an army of robots dedicated to running through hypotheses it would never be possible to test every one. This is why new protein folding models such as AlphaFold promise to be such a revolution within the biological sciences. By improving our ability to predict the final shape of a specific protein, they help us to decrease the potential hypothesis space associated with what a protein might look like and how it might operate. But such approaches can only ever narrow the potential hypotheses to be selected from. They do not solve the problem of how the scientist will choose—the exercise of their judgment.
In a relatively bounded space like drug discovery—one of the areas that makes use of predictions about protein folding—even with these techniques, selecting the hypothesis always seems to have some element of inarticulable judgment, randomness, or even unknowability baked in. This factor becomes even more important though when we consider hypotheses that lie entirely outside a current scientific paradigm, and so lack a search space within which to narrow the potential hypotheses. How would we describe the process by which Einstein successfully chose the hypothesis of relativity?
How this works in practice becomes clear when looking at how science proceeds. If it were possible to conduct science using an algorithmic process that could be stated and shared, then you would expect the results that mattered in science to be relatively evenly distributed amongst the scientific community3.
It turns out though that there are some people who are just good at science. Some individual scientists have careers with multiple breakthroughs and highly impactful ideas sometimes across multiple fields, whereas others work diligently in one field for a long time and never create any results that are "interesting" or important to the overall scientific enterprise. So, how do people become good at science?
It turns out that, like with Fred again..'s journey being good at science is a teachable skill.
Nobel Apprenticeship Networks
As of January 2024, of the 727 people who have won the Nobel Prize in physics, chemistry, medicine, and economics, 696 (96%) come from a single academic family tree— in other words, 96% of Nobel prize winners can trace back their formal research training relationships (for example, between PhD student and their PhD advisor) to a previous Nobel winner. In fact, 92% of winners can trace that lineage back to a single 17th-century Swiss scientist, Emmanuel Stupanus.
This is an astounding fact. If you believe that the Nobel Prize is a good indicator of importance within the scientific enterprise, then the implication is that the ability to do science in the most meaningful way has some kind of hereditary quality requiring a particular relationship between teachers and students. And there are some runs in particular that show this in stark relief.
The chemist Adolf Von Baeyer, for example, despite winning the Nobel Prize for creating the first synthetic dye, likely left his greatest legacy, not through his research, but through his lineage of students. In total, he has had 107 Nobel Prize-winning academic descendants. The image below shows 13 that occurred within just four “generations.” Consider a few of these:
Otto Warburg studied under one of Adolf Von Baeyer's students Emil Fisher. In his later career he laid the groundwork for quantifying respiration, identified key features of cancer cell metabolism and finished mapping the operation of a key enzyme in the respiration cycle. One of Warburg's students was Hans Krebs. Following his time working with Warburg in Berlin, Krebs headed to Sheffield University where he continued to study cell respiration, eventually mapping the Citric Acid Cycle or Krebs cycle, the essential final step in the release of energy to the cell in the form of ATP. On another branch, Severo Ochoa studied with Warburg's student Otto Meyerhof. After his training, Ochoa set up residence at NYU and in 1955 he isolated the polynucleotide phosphorylase which enabled him to synthesize RNA kicking off the genetic revolution.
As this particular set of Nobel Prize winners makes clear, something strange is going on. There were some overlaps in areas of research and technique, and maybe with Warburg and Krebs you could argue that there was some transfer of specific knowledge due to their shared interest in respiration. But this cannot account for the full range of breakthroughs that happened. Quantifying respiration and synthesizing dye are so different that they even won Nobel Prizes in different domains4.
It's always possible that some of this linked success is also simply a kind of nepotism. Nobel Prize winners favor their students winning Nobel Prizes. But if we think there is any validity to the Nobel Prize in identifying what counts as truly meaningful science, then this could hardly explain a result of 96% of winners belonging to a single family tree.
It seems the simplest explanation is that there is some kind of skill, or set of skills, that makes one a good scientist, and that these skills are somehow distinct from an articulated description of how to do science. Adolf Baeyer knew how to do good science and he passed it on to his students, who in turn passed it on to the next generation.
Conclusion
Fred is really good at music. He acquired his skill from a unique kind of training that blended formal instruction with attentive mentorship from an excellent master. And it appears that this kind of training to transfer complex tacit knowledge is common to more fields than just music. The implications of this are serious.
First, it means that we need to take training from skilled practitioners much more seriously. As individuals and as societies, we should be much more aggressive about seeking out the best in various fields and then finding ways to scale their knowledge through apprenticeships. If we want to have a whole generation of Fred again..s (and we should), then we need to first learn from the Brian Enos of our society.
Second, an implication of this almost hereditary educational transfer of “good science” or “good music” is that our society's ability to act competently in these complex domains is much more precarious than we might believe. It may be the case that the tacit knowledge that allows one to be excellent exists in limited and highly concentrated lineages. If these lines of transmission are broken for even one generation the knowledge lost could be incredibly hard to replace. We should put much more emphasis on preserving and supporting these.
Takeaways
As seen when looking at Bauhaus, a training program that blends together traditional classroom techniques and apprenticeship-style mentoring can be extremely effective for promoting skilled work.
For complex and amorphous work where there is no algorithm for producing new results, like music or science, it appears it is possible to train people to be “good musicians” or “good scientists” in the abstract. One of the essential features of this training is apprenticeship-style relationships.
Effectiveness may be very highly concentrated, with specific lineages accounting for an outsized proportion of our society's competence, as seen with the academic tree of Nobel Prize winners. We should pay careful attention to when this is the case so we can safeguard and continue to expand our abilities.
Acknowledgements
Thank you to Robert Bellafiore for reading early versions of this essay and providing excellent edits, comments, and improvements.
Real name Frederick John Philip Gibson
Or, you might at least expect results to be somewhat evenly distributed accounting for some impacts that might come from people working in places that received more funding or otherwise had some kind of technical advantage.
Likewise, it may be the case that part of what the Nobel Prize-winning teachers give to their students is greater access to capital to run experiments, a specific group of effective scientists to collaborate with, or a prestigious lab with more equipment. But Warburg did the majority of his work in a major Berlin-based university, while Krebs went to set up a first-of-its-kind biochemistry department at Sheffield prior to doing his best work. Similarly, Ochoa did his most important work at NYU two decades after his initial interaction with Meyerhof
This was such a great read. Impressive research and storytelling.