Lon Binder

Serving all-you-can-eat software since 1824. CTO @MeetOphelia. @FoodMayhem's fella. Dad to two spicy kids. Pepper's dog trainer.

[Lon Binder @lonbinder]

@brilliantlabsAR What is the resolution of the display? And apparent resolution at some distance? I'm trying to think about visual angle capabilities.

[Lon Binder @lonbinder]

I may have been too greedy at the library!

[Lon Binder @lonbinder]

This feel way too important to not share broadly! Wearing hearing aids could help cut the risk of dementia, according to a large decade-long study. Note addtl research in comments showing that this effect holds even for mild hearing loss.
https://t.co/6Es9XDcmXV

[Lon Binder @lonbinder]

Great walk down "memory" lane https://t.co/fpsLahjDGt

[Lon Binder @lonbinder]

Oh, and one more visual... imagine that as the factors evolve, the bottom of the curve pushes to the right (greater along X-axis), it's a wave over-taking the jobs to fall completely to the left of it. YIKES. /6

[Lon Binder @lonbinder]

My working theory is return speed (y) on #ChatGPT usage is sinh(x) / clarity ^ efficacy; where x = time it would take you; clarity is how clearly you express the task to GPT; and efficacy is for the model (it's getting better each update). /1

[Lon Binder @lonbinder]

There's opportunity in clarity. You see evidence of this where people have produced wild output from #ChatGPT or other #LLMs (such as falling in love or simulating intelligence and making the news). Those are professional AI researchers with high clarity. /5

[Lon Binder @lonbinder]

So there's a sweet spot in the middle for usage, it's not great for minimal use cases and terrible at complex use cases.

But what to watch out for are the control "levers" of clarity & efficacy. Now, mostly, @OpenAI controls efficacy, although you can tweak it here or there. /4

[Lon Binder @lonbinder]

That means tasks that take us minimal time also take #ChatGPT minimal time, there's not much ROI.

Tasks that take us "some time" take ChatGPT minimal time, so there's good ROI there.

Tasks that take us a "long time" take ChatGPT extraordinary amount of time. /3

[Lon Binder @lonbinder]

In this calculation, the "time" it would take you refers to both the time you take thinking (working through the solution) and then executing the solution. For easier tasks, the proportion of your time is heavier weighted towards doing than thinking. And the inverse is true. /2

[Lon Binder @lonbinder]

RIP CPU god https://t.co/zjb4ZdVEcv

[Lon Binder @lonbinder]

Hoping to see meaningful changes to the DEA's proposed telemedicine rules! Great points from @ZarkGray https://t.co/oJ8JHVCigq

[Lon Binder @lonbinder]

I don't understand this #agile diagram.

[Lon Binder @lonbinder]

Amaze. https://t.co/ccmivK5BpE

[Lon Binder @lonbinder]

For now, producing this already puts one at top 1% of the class anyway. Better than just doing the work itself. That won't be true for long...
https://t.co/AedPAU7Evp

[Lon Binder @lonbinder]

This is truly gorgeous.
https://t.co/2pLZe35MVJ

[Lon Binder @lonbinder]

https://t.co/0cVWOoYgSl

When chatGpt starts sending atlas instructions... well, I, for one, welcome our new overlords.

[Lon Binder @lonbinder]

I defy you to convince me that Kryptonian crystals in Superman (1978) wasn't predicting neural network language models like https://t.co/IcNsESTZBJ

[Lon Binder @lonbinder]

Thanks to "smart garage" technology, I now need to reboot my garage. Yay?

[Lon Binder @lonbinder]

History of scrollbars! https://t.co/puL9a1ND0W