The Boring Part of Bell Labs

"the" Bell Labs was effectively gone by the 1970's anyway. So the Bell Labs described in this article was "the" Bell Labs

The Bell System was, in modern parlance, fully vertically integrated. They didn't own the mines that the copper for the conductors was extracted from. And they didn't cast the copper ingots. Though they were interested in the metallurgy. Because they drew their own wire. And cables. And transformers. And vacuum tubes. And so on. It was all in-house. They even treated their own telephone poles. So something like a practical survey of various types of preservative treatments for wood was in the remit of Bell Labs just like the physics of a vacuum tube. Almost everything in science was in the remit when they had such a broad mandate. The Bell System had elements that almost resembled a kind of state within a state. (Part of why it was killed off -- antitrust violations.)

>I wonder why companies don’t do this anymore. Is it something to do with the monopoly AT&T held, is it related to corporate tax structures, is it related to how easy it is to find PhD graduates who studied similar topics of interest, or is it something else entirely?

Companies do a lot less to retain employees - fewer pensions, fewer accrued benefits and perks - so it's much less attractive to train when you are assuming attrition instead.

I don't think companies pay for you to go do an MS or a PhD full-time on company dime. At that time AT&T was a monopoly and may have had money to burn on this. There also may not have been expectations of hyper-growth from the stock market that exist today.

AT&T still pays for various MS courses (mostly MSCS, MS data science and cybersecurity) you can do on a part-time basis. It's quite easy to get the tuition reimbursement for it.

The answer to your question seems straight-forward: they still do.

1. Companies have tuition reimbursement for general employees, but it's usually nominal at $5k/year and part-time education is implied.

2. Executive MBAs and more selective executive seminars are covered by companies so the focus will be better. These aren't cheap (easily $125k/year) and might come with early exit penalties. I'd be surprised if the Bells lab offer didn't have similar clawbacks.

3. Tuition in 1970 isn't what it is now! Wayyyy cheaper!

One common theme is that companies used to treat good employees as assets. Now they treat all employees as liabilities.

What changed? A lot. The underlying theme across all companies, to both employees and customers, has been "see how much abuse they'll take before they leave", which sadly has had marvelous results because the answer is... a lot. Notice that at least half of the largest companies by market cap have no actual support at all.

Add in that tuition has exploded, it became cheaper and quicker just to import people than to train them.

The company I work pays normal salaries for their PhD students.

That pretty much standard in Germany in industrial research departments.

The Overton Window for what is acceptable company behaviour has slid very far to "maximizing near-term profits at all costs". You see this is in a lot of areas:

- Employee training and retention. Employers would rather aggressively churn the cheapest employees they can get, rather than cultivating experienced employees.

- Just-in-time inventory. Businesses want to have as little inventory as possible, which means any supply chain disruption causes ripples down the line.

- Advertising. Everything is stuffed full of ads now, including your smart appliances.

The reason companies get away with this is that all big companies have more or less colluded to behave the same way. The managerial class has said "we can have terrible customer service because _everyone_ has terrible customer service". Your product can be full of ads and break every 3 years because _every_ product is like that.

You might ask "why doesn't one competitor break the pattern and do a good job to gain market share?". The answer is that they do, and then they get bigger and more financialized. The capital required to do a good job requires them to take investment, and once the growth narrative has ended they're pressured to squeeze profits out of their customers.

People used to spend entire careers at one company so investment in them pays off. I spent 10 years working for Bell Labs in the 1980s albeit as a contractor, and the bodyshop that employed me found it worthwhile reimbursing all educational expenses for a grade of C or better.

This is still very common in Germany. Many companies, especially large ones, offer Master/bachelor thesis, which are supervised by a professor at an university, but the student is employed full time for the duration of the thesis.

> "I wonder why companies don’t do this anymore."

Because we're not in the gargantuan post-WW2 economic boom combined with the electronics-ization / mainframe computerization tech boom anymore. It's easy to treat employees generously when revenue growth just keeps happening almost on its own.

And STEM degrees or even possessing any undergraduate degree was much less common than they are today. The article says "Over 130 people signed up for the One Year On Campus program in 1970.", which is a pittance in a corporation with over a million employees at its peak. Unsurprisingly, people who specialties make them more difficult to replace get treated more favorably.

This is very common in Europe. PhDs are sponsored by a given company

I agree! Science is about experiments to verify hypotheses. Design of Experiments seems like a fundamental part of that. That's also why the quote below made me laugh.

> What if you don’t care about efficiency or causality?

"Yeah, what about if you don't care about money/time and are happy with finding a correlation only?!!?"

Totally agree! DOE is way underrated. Once you’ve had to run real experiments (especially at scale), you really appreciate how much time and money a good experimental design can save. It’s one of those areas where a bit of math makes a huge practical difference.

That slide rule got replaced by software at least two decades ago and probably far more. The engineering staff got quartered and the sales staff tripled, QA was fired, and stock buybacks are the name of the game.

Yeah the accidental discovery of the background radiation that they eventually traced to big bang predictions. It’s a good story about following through with your experimental results.

I suggest “Design and Analysis of Experiments” by Montgomery.

https://faculty.ksu.edu.sa/sites/default/files/douglas_c._mo...

1 week is fascinating. Was it like – the missing piece was modern version control/CD? What kind of testing would need that? (We have configs at work where the system interactions are so unknowable and the financial implications of reduced efficiency so profound that we have to run multi-week A/B tests to change values) Was it some kind of pathological documentation culture?

I'm reminded of a HN comment about working on the Oracle RDBMS: https://news.ycombinator.com/item?id=18442941