Indeed building something you want, or that someone you know wants, is necessary, but not sufficient! I'd say it depends how much time you're going to spend creating it and whether you have broader commercial ambitions at the outset.

If you're creating something you're going to use yourself anyway, that could well justify creating it (if it won't take too long). Similarly if you're creating it for someone else (as a favour, or who will pay you appropriately). Or if you can create a minimum viable product quickly to try out on people.

Also, particularly in the realm of short software projects, there's a blurry line between creating something for fun/interest and doing so with serious commercial intentions, i.e. you could justify doing it speculatively without feeling you'd wasted your time if it goes nowhere.

But if you're going to take months (or years) full-time creating something with a view to commercializing it, i.e. make a serious effort, it is remiss not to do basic research and evaluation first, to find out whether there really is a market for your thing (e.g. who customers would be, potential market size, what customers currently do instead, whether you can actually improve on that enough, how hard that might be, what customers would be prepared to spend), whether your thing should do what you think it should (i.e. its features, or indeed whether you’d be better off creating something else entirely), etc. It's far cheaper to do basic research & planning than to spend months/years creating something speculatively and only then discover much/all of that was misguided.

explicitly without concern for how exactly you are going to commercialize. Indeed, most successful companies figure out where their margins come from quite late into their lifecycle.

The exact way you commercialize or get margins can of course change - but if you can't figure out any way way of making it work on paper, the chances of it succeeding in real life are slim.

(My LW article on this FWIW: Write a business plan already — LessWrong)

Good post. On one point, I think Landmines are useful in many fields, to warn against important beginners’ mistakes/misconceptions. Though (unless for big safety reasons) this should indeed be secondary to positive advice.

Eg with a startup, don’t spend lots of time creating a product before writing a business plan. The plan should come first, or at least early on, because it’s how you decide whether to create the product! (Something I’ve written about on here)

Re being deficient in vitamins, it’s worth taking a supplement containing all 23 essential micronutrients (every few days), as almost no one gets 100% of the recommended daily amount of all of them, which is nearly impossible to achieve from a plausible diet anyway. I.e. you are probably somewhat deficient in something.

Broadly agree - I overstated my point; of course some people don’t have these concepts. But I think there is a big gap between having these concepts as theory (eg IVT in pure math) and applying them in practice to less obvious cases.

(Cf Wittgenstein thought that understanding a concept just was knowing how to apply it - you don’t fully understand it until you know how to use it.)

I have a similar comment: it seems to me the problem is not that people don’t have these fairly obvious concepts, it’s the unobvious application of these concepts to real world cases. Eg the three examples given for the Intermediate Value Theorem are not entirely obvious instances of it, and it’s not obvious that the IVT completely solves them.

Even then it’s often ambiguous what is meant. If you have a single handle (on a faucet or shower) which is rotated to adjust temperature, it’s often unclear whether it’s the position of the part of the handle you hold or the other end (which moves in the opposite direction) that indicates the desired temperature

I’ve heard it said that you should have a pair of scissors in every room in your house. (The scissors incidentally tend to differ depending on location, eg kitchen scissors vs nail scissors, but that’s a detail.)

So do you think this is part of how it generates images - i.e. having used depth estimation and much else to infer 3D objects/scenes and the wider workings of the 3D world from its training photographs, it turns a new description into an internal representation of a 3D scene and then renders it to photorealistic 2D using inter alia a kind of reversal of its 2D->3D inference algorithm???

Which seems miraculous to me - not so much that this is possible, but that a neural network figured this all out by itself rather than the complex algorithms required being very elaborately hand-coded.

I would have assumed that at best a neural network would infer a big pile of kludges that would produce poor results, like a human trying to Photoshop a bunch of different photographs together.

Indeed, airplanes are edible and delicious. Frenchman Michel Lotito (aka Monsieur Mangetout) ate a Cessna 150. Yes, he actually did!

https://en.wikipedia.org/wiki/Michel_Lotito

why are you so sure dalle knows what an image is or what it is doing? Why do you think it knows there is a person looking at its output vs an electric field pulsing in a non random way?

I don’t think it does (and didn’t say this!)

My question is how it manages to produce, almost all the time, such convincing 3D images without knowing about the 3D world and everything else normally required to create a realistic 3D image. As you can’t just do it by fitting together existing loosely suitable photographs (from its training data) and tweaking the results.

I don’t deny that you can catch it out by asking for weird things very different from its training data (though it often makes a good attempt). However that doesn’t explain how it does so well at creating images which are broadly within the range of its training data, but different enough from the photographs it’s seen that a skilled human with Photoshop couldn’t do what it does.