I do not think this is true. Careless unmanaged use of AI quickly creates "AI-only codebases" that can't meaningfully be worked by an unassisted human.
It would be economically non-viable to have humans carefully go through and "refactor" them; no one would pay for this
TL;DR: A single cortical pyramidal neuron is not a point; it's a powerful, noise-robust, general-purpose computational unit.
Dendrites are not just wires to help neurons connect; they are the substrate for complex nonlinear computation. (15/15)
preprint: https://t.co/m2a8ieuo8A
What can a neuron compute?
Real biological neurons are complex, but how capable are they?
Using a new method, we found that a single cortical neuron can classify cats vs dogs, recognize spoken words, and solve 10-bit parity, all tasks thought to require entire networks. (1/15)
An unusual title and an unmistakable voice helped make “Mmm Mmm Mmm Mmm” one of the most distinctive hits of the 1990s. Released in 1993, the song paired Brad Roberts’ deep baritone with stories of outsiders and misfits, turning the Canadian band into unlikely international stars.
@GaryMarcus https://t.co/U5ge4WVknC
It's always a risk and a pain to call it out, but there sociopathic systems that work almost the same - with loops of "idealization, devaluation, discarding" to get to the resources. The late @hintjens wrote a lot about it https://t.co/7OK7k0j4jd
The saddest thing is that the proponents will move on to the next bubble - from blockchain to “AI” and later from “AI” to quantum computing. One can also observe it in the small: the speed at which crowds move from one model/framework/harness/whatever to the next one - within a couple of days
@GaryMarcus No wonder though - via Conway’s Law - that the technology is based on massive plagiarism and appropriation, and some proponents have trouble properly citing and attributing as well
One could easily imagine the assisted coding to have been a marketing trick to boost the sales of @mfeathers's "Working Effectively with Legacy Code" book