Any freshman trying to use this programming technique would be flunked in Intro to Programming 101. Trying to use AI to write real code will only result is very poor results. LLMs don't have any semantic capability to know "why" thinks are. Using simple phrases as directions to AI does not communicate enough information for real requirements. Your AI generated code is just recollections of what the LLM has been trained on, which you have not seen. This is such a bad idea. It's just one more stupid way for desperate AI companies to try to make a buck. But it's a really bad business model.
So AI increases productivity among unproductive software developers by 25%? The difference between low productivity developers and high productivity developers is a factor of 20X So Increasing the bottom rung doesn't really achieve much. I supposed there are always stupid managers who will try to leverage this. Good developers will go elsewhere and should.
There is no serious business case that can sustain investment of $500 Billion. Microsoft, Google, Meta, Tesla, Apple, OpenAI, etc are all going to get hammered when stock holders realize those companies have overspent and have no means of getting back that investment.
This is definitely a bubble that will impact the stock market.
Using AI to code is a shallow victory at best. Once you deal with logic that it is less likely to have seen, the efficacy of the code drops quickly. Plus every single peice of code from AI has to be carefully inspected for bugs. Only a good coder can do that. So the likelihood of introducing bugs and subtle bugs goes up dramatically. Is that really a gain in productivity if you are now introducing bugs at a higher rate? Noooooooo.
Everyone can expect more BS from Sam as he's tries to remain relevant and grease the wheel for the OpenAI IPO, where he will cash out big time. Loads of completely unsupported BS until then. Thank you Media, for being an uncritical, unfiltered conduit for his sh*t.
AndrewZX writes: An excellent deep dive into the history and capabilities of the 1980's DEC Pro 380 PDP-11 workstation and the classic Venix UNIX that it ran.
All this AI power demand does not change the inexpedient fact that each and every nuclear power plant costs BILLIONS to build and takes years. Unless these companies want to own and build these plants, no one else is going to foot the bill. In other words, only money talks. It ain't happening.
Let's not kid ourselves. These companies will sell any kind of sh*t to make a buck, without regard to societal impact. They will undercut vast work forces without thinking twice. Our biggest threat is not self aware AI, it is un self aware corporations.
Humans have lived in the close presence of fire for over a hundred thousand years? I wonder if we may not have evolved some immunity to soot and carbon bits.
$75 BILLION dollars. What is the business model? How the F does Google expect to make money on AI that is crap? Where do they spell out the products they will sell? Now may be a good time to dump Google stock and think about shorting it.