Changes

I’ve been thinking change this past weekend. Change brings the opportunity for renewal and growth; it requires direct active engagement, not avoidance; and it is never too late to start dealing with change – sometimes being late to the game is an advantage because it is so clear what to do.  I'll touch on each of these implications a little as I wander through some of the changes we are all dealing with.

National leadership

The Trump administration has torn apart our national science funding, our Department of Education, our renewables investments, our diversity programs, our international relationships, and on and on. Most of these policy changes are ill-considered and deeply harmful.  

But let’s be honest — it is not like everything was going great in those areas.  For example:

• Education: Our educational spending is way up and our reading and math outcomes are way down.  
• Energy:  As will be discussed below, we are far behind in building out the electrification technologies that modern industry demands.   And we are caught up in this stupid political war around renewables, when we need every bit of power we can find to realize our goals.
• Diversity:  Despite all our diversity investments, our society continues to have persistent gaps in achievement — for instance, wealth distribution.

Despite good intentions, our programs in these areas (and others) have failed, and we must be realistic about these failures.  We can use this opportunity to rethink our goals, reinvent our programs, and shed old commitments and ineffective programs.   It is an opportunity for leaders to start from first principles, re-articulate our goals, and redefine our national efforts to achieve those goals.  I am waiting to see who steps up.

Reading

A bit of an aside, but while on the topic of education, we desperately need to figure out how to get our kids to read — we've been on a three decade decline in reading skills.  That same article has a set of simple proposals for how we turn this around — the ‘promote “bad” books’ idea really resonates with me.  Comic books were my gateway drug — I read a stupid amount of Superman, of Archie, of Scrooge McDuck.  Some Scrooge McDuck scenes are still seared in my head, and maybe that is a waste of brain space, but it got me hooked on reading.

Any book that gets kids reading is a good book. Any changes to our educational programs should prioritize making a real difference in reading and literacy.

Tech leadership 

Back to the primary topic of change, the Diff discusses tech leadership and suggests that the best tech leaders are excellent students of history and have used that ability to hone their strategic leadership and guide their companies through changing times.  My experience is that the best tech leaders really embrace Andy Grove’s “Only the Paranoid Survive” thinking.   Certainly, leaders are aware of history and recognize that success in one wave of computing doesn’t guarantee success in the next — IBM, the Bunch, DEC, Compaq, HP, and Intel — they are all gone or shadows of their former selves.  That recognition, that success is ephemeral, is the background that drives leaders to be paranoid, hyper-focused on competitors and customers, and to be on the lookout for even the slightest slippage in usage or the slightest trend moving against them.  

When Windows 95 launched, Microsoft was arguably at the height of its powers, and yet many of us were already focused on the coming Internet wave and had been so for months.  The Windows vs Internet discussion became heated at times in the company, and the confrontation was encouraged — because internal strife is far less painful than a failed product or a failed business; the marketplace is a tough teacher.  The best way to survive an industry transition is to lead the change.

Electrification

Noah Smith has a good article on the Electric Stack.  The modern industrial stack has shifted hard towards electrification, and the US is woefully behind in most of this stack.  As Noah points out, we don’t need to waste a lot of time thinking about what to do; the direction we need to take is blindingly obvious:

In this situation, we don’t need to tear our hair out asking which industries the government should promote; as in the era of the railroads, the answer is bleedingly obvious. A core set of “winners” has already been picked for us — it’s just batteries, electric motors, power electronics, and chips.

There is an advantage to being a laggard during a period of significant change — you don’t have to sit around trying to figure out what to do; just fast follow the taillights of the leader.

Generative AI Slop

Chris Williams on generative AI content:

AI is a tsunami of mediocrity. It's a rush to the middle and an oversupply of average writing. … It's the "meh" tidal wave.

Smart words.  AI is an amazing and transformative set of technologies and will have an incredible impact on our lives.  But we write to learn and to connect to people.   AI is great as a research assistant, as a critic, as an idea spark — but you need to write yourself to really understand what you are learning.   And you need to write yourself to really connect with other people — readers can see through AI slop, and won’t connect with you at all.

Generative AI Design

Feynman’s talk on “There’s Plenty of Room at the Bottom” helped set the goal for the semiconductor industry, and the industry has been striving for decades to create ever-smaller computing elements.  The complexity of designing and building a modern chip is staggering — an Nvidia Blackwell chip has over 200 billion transistors.  Every aspect of modern chip design is massively complicated.

And so this is a great use of AI — to design a massively complicated microfluidic cooling system.  Humans could not have designed this cooling system in a reasonable amount of time, and it is fascinating that it ends up very bio-inspired — designs that took nature billions of years and trials to achieve.  

Generative designs are going to find their ways into all kinds of materials and devices — structural materials with breakthroughs in strength and weight, engines with performance breakthroughs, novel chip designs, and so on.   Every industrial product will benefit from and require AI design assistance.  

Fifty years from now, when we slice into any modern manufactured product, we will be amazed at the organic-looking structures within, at every scale, in the product.  The designs would have confounded the engineers of a century ago.

Generative AI Capacity

An interesting problem is how to best utilize off-peak inference capacity.   As the industry races to build out AI capacity, there is a lot of inference hardware sitting idle at off-peak hours.  Certainly, the industry will become more creative in how to ramp up and down capacity (for instance, the Microsoft chip cooling idea above, which was partly motivated by the desire to overclock chips during peak times).   And there is plenty of room for pricing and product designs to shift usage to offpeak hours — for instance, The Diff points out that ChatGPT's "Pulse" feature is really about time shifting usage ($):

Another more prosaic reason they're doing this is that there's different demand for inference capacity at different times of day, and a batch-processed once-a-day task, distributed at massive scale, is a good way to get more utilization out of the same hardware footprint. When you hear about OpenAI product decisions, think about capex, and when you hear about their capex moves, think about product.

This is just one example — ChatGPT’s scheduled queries are another excellent example of trying to shift demand to off-peak hours. We will see a lot more experimentation with offers and pricing.

"The only constant is change"

We are in the midst of so much change, and it can be both overwhelming and exhilarating.

Recomendo recently had a link to "40 Harsh Truths I Wish I Knew In My 20s" by Daniel Pink. I particularly resonate with "action beats planning".  Obviously, planning is necessary, but oftentimes, you just need to dive in and do something, and the experience of doing so will make any planning exercise 1000x more informed. And many more of his truths are super relevant in times of change – don't wait for permission, you will regret inaction, don't waste time on cynicism.

He has a strong bias for action. Change creates the opportunity to reconsider and re-establish our priorities, and requires that we engage directly and with open minds. Wishful yearnings to return to some simpler time of yesteryear are a waste of time.