Empathy for Science

Being a physicist talking about climate change feels like this:

True or False? Murder is bad because 2+2=6.

  • False (100%, 1 Votes)
  • True (0%, 0 Votes)

Total Voters: 1

Loading ... Loading ...

Big Business and Customer Service

I get the frustration with big business, especially in ecommerce, and at one level I sympathize. But on the other hand, when I buy something from Amazon, I get it or I get a refund.

I recently bought something from Menlo House, and they simply didn’t ship it. They took my money and didn’t send the product. Customer service is a Kafkaesque nightmare. I had the same experience with Babbitt’s Online. They just don’t ship what I paid for, and both of them avoid and delay returning calls or emails.

Amazon will at least email me back. Menlo took weeks to reply, and then they lied to me. Babbit’s still hasn’t sent me my order or a refund.

Green Energy

I think they’re going to be able to power the world from the vigorous gyrations of Milton rolling in his grave. Do it for the environment, people.

Work vs Luck 2

Another point is that I can’t just go to work. I can only submit proposals when a contract is open.

So if you have a job, you go to work on Monday, work your eight plus lunch, and go home. Do that again, hopefully a reasonable number of times a week. Same for homework. You’ll get more homework.

If a contract RFP is open, I can go in and apply. But if one isn’t open, I can’t apply. The RFPs I’m looking at are on a finite number of sources, so there’s only a moderate amount of hunting possible before I’m endlessly refreshing a browser tab. Going back to that homework analogy, it’s not like I have a steady backlog of RFPs assigned. I chase one and then I wait. They’re bespoke, so there’s limited prep without a RFP to work.

That’s the job. I can backfill with grad school and writing, but the landlord needs to get paid. It’s just very uncertain in a way normal employment isn’t.

Work vs Luck

There’s a sliding scale between work pays off and luck pays off. Depending on what you’re doing, you may be farther toward one side or the other.

Homework, at least well constructed and well assigned homework, is far on the work side. If you do ten hours of homework on what you expect to be a ten hour assignment, you should be done or close to it. Maybe not, but if not, you should be close. If I get some assignment and think ’10 hrs’, 10 hours later I should be close. Maybe it’s really a 12 hour assignment, or I chase some formula to find out I should be using a different one, but I should be close.

Starting a business is far on the other side, but certainly not pure luck. It’s more like earning coin flips. It takes me about 40 hours to submit a contract proposal, and if I don’t get it, I have nothing. I’m not 90% of the way to a new contract. I can port over about 50% for the next proposal, but that’s included in the 40 hours. They’re really about 80 hour jobbies, but I do a lot of them. The thing is, if I get no contracts, I have nothing. So think of it like earning a coin flip. To achieve some goal, pay rent, I need ten heads, and it takes me 40 hours to flip a coin. One might think that fine, that’s 20 flips for 800 hours or about 20 weeks, but that’s neglecting how big a role luck is. I could get 18 tails out of my first 20 flips, and then I still don’t have rent.

This is what entrepreneurs mean when they talk about business being somehow different. The nature of success in a startup is different from the nature of success as an employee. I don’t get to bill those 40 hours to anyone, because there’s no money there.

If I make the contracts, then I have money, and it seems to flow like water. But if I don’t, the time is wasted, and there’s just nothing there.

And my landlord needs to get paid.

It’s not harder or easier, it’s not better or worse, it’s not good or bad. It’s just different, but that different is worth keeping in mind.


So I just realized how I can do signal identification via a set of processing tricks. Up until now, I’ve been playing with full signals, sometimes downsampling, but generally running whole sample rates. This is because I haven’t been able to figure out how Matlab AI or TensorFlow took some sets of data. I’m going to send the network trainer just signal characteristics I want to explore.

Like, I knew it was possible, but what arrays to pass, how to call the cells, yadda yadda, was unknown. Those sorts of things are a lot more complicated than one might think, if simply because they seem so easy. Maybe they are, but it took me a bit.

So tomorrow, Friday, I’m going to redo a AI function and pass it prechecked data. I think this should work, and I’m guardedly excited.