[ExI] AI and Innovation

[Keith Henson]

I find AIs to be as bad as humans in insisting on known approaches to
a problem.  They are stuck on sorting and drying trash when the
trash-to-fuel project needs neither.  One of them was stuck on the
project as being an "energy vampire" when the reaction makes more than
twice as much syngas as the electrical energy input.

Keith

On Tue, Dec 30, 2025 at 11:19 AM Adrian Tymes via extropy-chat
<extropy-chat at lists.extropy.org> wrote:
>
> On Tue, Dec 30, 2025, 12:38 PM Ben Zaiboc via extropy-chat <extropy-chat at lists.extropy.org> wrote:
>>
>> There seem to be a number of people here who often make use of AI
>> systems, I was wondering if anyone has noticed any AI coming up with
>> novel ideas? Can they innovate at all, as opposed to sifting through
>> large amounts of already known information, creating summaries, doing
>> calculations, etc.?
>>
>> What I'm really asking is, if given a problem, can they come up with a
>> solution that nobody has thought of before?
>
>
> That depends on how broadly you stretch "nobody has thought of before".  There's an old saying, "nothing new under the sun", referring to how "new" ideas are - to greater or lesser extents - remixes of what's come before.
>
> For instance, most new power sources turn into novel ways to boil water and drive steam turbines.  The Internet's beginnings can be traced back through telegraph lines, which in turn can be seen as a wired variant of semaphore signaling.  Large language models...well, there were language models before, but much smaller.
>
> Thing is, there is just so much already known information out there that, many times, what matters is not whether it makes anything truly new, but something new to the user.  Someone in a lab you've never heard of may have the perfect solution to your problem, but you've never heard of it.
>
> In my own work, especially where I am doing things that (apparently) literally no one has done before, it is a useful research assistant, bringing to light aspects that have been explored before so I can focus on the novel parts: see what the known challenges are (quantifying the ones I spotted, and bringing to light ones I missed), to help me think of solutions to them.  Occasionally it has what looks like a novel suggestion, but when I've looked into the origin rather than just running with it (I usually just run with it), these seem like logical consequences of the inspiration I told it of (which I had not yet realized), or that I didn't think of that particular angle first and it found a source that had.  Then again, many of my own inspirations so far could be considered logical (if not previously significantly explored) consequences of what's come before.
>
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat