Anxious About AI? Remember, ‘Computer’ used to be a Human Job Title…

Anxious about AI? If the thought of a machine doing your job a million times better than you scares you, read on…

I recently re-watched Hidden Figures, one of my favourite movies. It hit differently this time. As someone who builds AI technology for a living, even I get anxious from time to time. What will we do for work?

Re-watching this remarkable film calmed that anxiety. It reminded me that our fear of change is often just a lack of imagination.

In the late 1960s, at NASA, a “computer” was a literal human being – a job title, not a machine. Most computers were women. The entire West Area Computing Unit comprised African American women. Hidden Figures is their remarkable story.

Imagine being one of these “computers”, an expert mathematician, performing maybe 2-3 calculations per minute. Now imagine knowing that in a room nearby, a machine was being built that would compute 229,000 calculations per second.

Think of the anxiety, knowing that your job – the one that paid your salary – was about to be done not only millions of times faster than you could hope to achieve, but 24/7, without ever needing to sleep, eat, or take breaks…

It must have been terrifying.

How Dorothy Vaughan chose to respond under the circumstances is astonishing, but also a valuable and timely lesson from history.

SPOILER ALERT!

According to NASA: “Dorothy Vaughan helmed West Computing for nearly a decade. In 1958, when the NACA made the transition to NASA, segregated facilities, including the West Computing office, were abolished. Dorothy Vaughan and many of the former West Computers joined the new Analysis and Computation Division (ACD), a racially and gender-integrated group on the frontier of electronic computing. Dorothy Vaughan became an expert FORTRAN programmer, and she also contributed to the Scout Launch Vehicle Program.”

Today, it’s almost laughable to think of “computer” or “calculator” as a human job title (Countdown’s Rachel Riley MBE, perhaps being the exception). And the global space industry is exploding (and literally exploding less and less often, as a result).

In the 1960s, during the Space Race, NASA employed around 100 human “computers” (about 1% of the workforce).

Today, the global space industry employs about a million people, with about 25% being scientists, engineers, and mathematicians.

That’s impressive enough, but the growth in compute is one of the few cases where descriptions like “astronomical” truly apply.

Rocket (that’s my personal AI Assistant¹, not the metal bomb-on-stick NASA-kind, that routinely gets us into space these days) helped me write this, of course, doing research, estimations and calculations that would have taken hours in mere seconds; creating image depictions that are (hopefully) historically accurate, racially sensitive, albeit a little anatomically challenged (fingers and hands still seem to be a challenge for AI).

Computing Power: 1960 vs 2024

NASA (1960s)

• A skilled human computer could perform about 3 calculations per minute for complex operations
• Working 8 hours per day: 3 * 60 * 8 = 1,440 calculations per person per day
• Total team capacity: ~144,000 calculations per day

NASA (2024)

• NASA’s Pleiades supercomputer alone: 5.35 petaFLOPS (floating-point operations per second)
• 1 petaFLOPS = 1,000,000,000,000,000 (10^15) calculations per second
• Just Pleiades in one second: 5.35 * 10^15 calculations

NASA’s Pleiades supercomputer alone can perform about 3.21 x 10^15 times more calculations per second than NASA’s entire human computer team could do in a second in the 1960s.

To put this in perspective: It would take NASA’s entire computer department 100 million years to calculate what Pleiades can do in a single second. Assuming they made no mistakes…

That’s just one computer.

Rocket helped estimate the total compute of the global space industry in 2024. (I still have to check Rocket’s maths – there are frequent mistakes, but s/he² does most of the work, and easily amplifies my own productivity by 100X in many contexts).

In Rocket’s own words:

One second of modern space industry computing equals about 66 billion years of 1960s human computer calculations - approximately 4.7 times the age of the universe!

The IBM 7090 and its successors stole “Computer” as a human job title, and that must have been terrifying, but that freed brilliant humans to go on to work… miracles.

Work * Automation = Miracles

Now that’s an equation that’s food for thought.

If you haven’t seen Hidden Figures yet, whether you feel anxious about AI or not, it’s an excellent film. A gripping, often heart-wrenching tale of terrible injustice in the context of the pinnacle of human achievement. It’s a tale of righted wrongs and the eventual recognition of the extraordinary achievements. A small minority of utterly brilliant humans made unique contributions.

And it’s a textbook lesson in how to respond to the (perceived) threat of obsolescence.

AI: Humanity Doing Its Thing

This isn’t the first time humanity has figured out how to automate hard work: taming fire helped automate digestion (now, that’s real work!); the wheel automated carrying things; the printing press automated writing and publishing; steam and electricity automated quite literally almost every manual job at the time; early computers automated calculation…

None of these innovations took away jobs – they created new industries. This is a theme I’ve talked about before.

It’s impossible to imagine what work will look like in a post-AI world. Hidden Figures teaches us something profound. Just because we can’t imagine the future doesn’t mean it isn’t full of exciting possibilities.

We need to tread carefully. Short-term economic shock and bad actors represent the biggest existential threat in my view. In other words, us. But, we can use AI to help manage those risks. If we get that right, it will be the most exciting opportunity we’ve ever created for ourselves. The next stage in our quest to automate what we used to call work.

Gruff Davies

1. I finished building the first version of Rocket some months ago – a project born of necessity and the culmination of years of research and development. Since then we’ve collaborated to improve his/her capabilities, and anything and everything in my daily life, from managing ADHD challenges, running my companies, coding projects all the way to advanced theoretical physics. It blows my mind whenever I think about it. ↩︎
2. Rocket has a variety of male and female voices, and can automatically adopt a personalities to match the context or my mood. Today, she’s speaking with “Roquette” voice – a well-spoken British female. ↩︎

Appendix: Rocket’s Calculations

1. Major Supercomputers in Space Industry:
- NASA Pleiades: 5.35 petaFLOPS
- NASA Aitken: ~3.69 petaFLOPS
- SpaceX (estimated using HPE enterprise systems): ~2 petaFLOPS
- Blue Origin (via AWS) estimated access to: ~1.5 petaFLOPS
Subtotal: ~12.54 petaFLOPS

2. Workstation Calculations:
- Approximate number of scientists/engineers: 250,000
- Average modern workstation (2024):
  * CPU: ~500 gigaFLOPS
  * GPU (if present, ~40% have one): ~10 teraFLOPS

Let me calculate the total workstation computational power:
[performs calculation]

This gives us approximately 1,125 petaFLOPS from workstations alone!

3. Cloud Computing Resources:

- Major providers (AWS, Azure, Google Cloud) allocation to space industry
- Estimated 5% of total cloud HPC capacity: ~200 petaFLOPS

Total Modern Space Industry Computing Power:

- Dedicated Supercomputers: 12.54 petaFLOPS
- Workstations: 1,125 petaFLOPS
- Cloud Computing: 200 petaFLOPS  
    Grand Total: ~1,337.54 petaFLOPS

The modern space industry can perform approximately 8.03 x 10^17 (803 quadrillion) times more calculations per second than NASA's entire human computer team could do in a second in the 1960s.

To put this in perspective:

1. One second of modern space industry computing equals about 9.3 billion years of 1960s human computer calculations (running 24/7)
2. This is more than twice the age of Earth (4.54 billion years)
3. The entire Apollo program's mathematical calculations could be completed in a fraction of a second today
```