@OpenAI’s GPT-3 beta is literally mind-blowing. See the Davinci-Codex model in action below, writing Python code that runs flawlessly on an RP2040 microcontroller board from @Adafruit. Is this the future of learning to code? A thread (1/n)
I do wish I was a better coder (started late), but still find it very rewarding being able to write even the most basic scripts and programs. Learning how stuff works, building things, writing code, and taking things apart is the #1 way to develop #techne h/t @kevin2kelly (2/n)
In the Codex demo, OpenAI’s CTO @gdb said that coding is essentially two things: 1) the high-level cognitive aspects of understanding a problem and architecting a solution and 2) low-level implementation that maps a small piece of functionality to code. (3/n)
The first part is where humans shine (creativity), and the second part (tedious & time-consuming) is where Codex really shines. GPT-3, and it’s progeny, will make the days of combing through @StackOverflow pages for the choicest chunks of code a thing of the past. (4/n)
Again. This. is. mind. blowing. Both practically and philosophically. As a shoddy coder/maker and an even shoddier philosopher, the implications of this tech are profound. I need to dive deeper into @Plinz ’s work and back episodes of the excellent @lexfridman podcast. (5/n)
Practically, this could speed up the iterative cycle of coding for existing programmers, and unlock programming for a huge group of people for whom the syntax barrier to entry has been too high until now. #OODA@johnrobb (6/n)
What has been your experience with GPTx and Codex? How do you see this technology impacting your work as a coder/dev/human being? Is AI pair-programming the most likely model for this or will humans eventually be booted from the loop? Ro-bot? Co-bot? No-bot? (7/n, end)