[ExI] Multiplying engineers

[Gadersd]

So much of software engineering is dealing with tedious boilerplate code. GPT4 is nearly perfect at automating the tedium. It still isn’t sophisticated enough to handle the more novel aspects of programming, but hopefully it will soon improve to the point where a developer can write the specification for a program and have it generated. The exponential growth of software development will be astounding.

I don’t think there is much inherent limitation in the model architecture as the models have demonstrated that they are capable of human level reasoning but often do not meet this potential unless the prompts are carefully crafted. Perhaps with better training methods the current state of the art models will reach effortless human level design and reasoning soon.

> On Jun 2, 2023, at 7:18 AM, Keith Henson via extropy-chat <extropy-chat at lists.extropy.org> wrote:
> 
> The sample I am going on is small, but one engineer I know who also
> knows a fair number of other engineers tells me that GPT-4 increases
> the productivity of software engineers today by a factor of 3 and
> expects it to double again and spread to the far reaches of
> engineering on a time scale of a few months.
> 
> I have wondered in posts on Power Satellite Economics (or maybe Howard
> Bloom's mail list) if China with its 3x number of engineers would
> outstrip the US, particularly in space.
> 
> If the CCP comes down as hard as they say on AI, and engineers in the
> US use AI as a multiplier this may be reversed and the US will surge
> ahead.
> 
> This should be apparent in a year or two if it is true.
> 
> Keith
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat