OpenAI o1 might be the final nail in coding's coffin — "If OpenAI's o1 can pass OpenAI's research engineer hiring interview for coding at a 90-100...

fjtorres5591

Active member
May 16, 2023
315
76
28
Visit site
Too much hype and pearl clutching, not enough understanding of the limits of neural nets.
Maybe this will help:
In a nutshell, neural networks models, no matter how large, how complex, how well-trained, converge on a similar error rate that is far from non-zero. No clear answer yet but it appears to be due to the nature of language or the neural network approach or both.

They all level off at an errot rate well above the reliability required of mission critical software. So while AI processed code might theoretically be marketable for gaming there are entire industries were only human vetted code will be acceptable. That is where human coding will survive: some body will have to tell the model what to do and then verify it actually does what it is supposed to do.

As always, the truth about AI software remains that it won't be software that puts people out of work but other people making better use of software. AI is a tool learn to use it to your benefit or face the consequences. Causevthe djinn will not return to the bottle.
 

CadErik

New member
Feb 6, 2015
12
1
3
Visit site
Kevin's articles lack some depth and go for catchy titles. Not only the errors, but the AI will just spit out code it learned from existing open-source codes which is not always accurate or even good. I'm sure today's AI could replace some jobs, but this would not be any better than an average junior developer. I do not have a single doubt that the AI will pass interviews, or the turing test but these do not require to solve actual problems. Not saying this won't change in the upcoming years but wording in Kevin's articles is not accurate.
 

Members online

No members online now.

Forum statistics

Threads
325,738
Messages
2,247,288
Members
428,401
Latest member
comprarpermiso1