Re: Phix/Euphoria Rosetta Code Rubik's Cube
- Posted by petelomax 3 weeks ago
- 268 views
Nice one. Why not post that on rosettacode, after all it's a free world.
I've moved on a bit from the initial shock of thinking AI is going to replace us humans and take over everything. If you are thinking you can no longer do what thousands of other people have already done, you'd be right, but the day AI advances to the point where it could have replaced the likes of Einstein, Edison, Tesla, Hawkins, Copernicus, Gauss, Newton, Darwin, Freud, Ghandi, Aristotle, da Vinci, Archimedes, Euclid, Euler, Pythagoras, and Hammurabi is a very long way off, quite probably many decades if not millennia, and almost certainly requiring precisely the kind of innovation the current crop of AI has absolutely zero chance of ever doing [to] itself.
Done right, which these things rarely are, AI should free us from petty detail and allow us to operate on a higher plane. Granted, that will be a moving target: once AI gets trained on thousands of people operating on that higher plane, it'll usurp that too, but that should just push us up another level. Now if you're waiting patiently for governments and big business to "do it right", you'll have a very long wait and be left stranded at the bottom of the pile, and instead you should realise you already have all the (free) tools you need to take that next step right in front of you, right now.
While not strictly my thing, education stands out as one of the more promising areas to explore, since (as it stands) all AI can ever do is teach stuff taught before it existed, and that will essentially remain true no matter how many higher planes it usurps. Those already in education may have a significant edge, but few devise brand new lessons. Then again, AI might spot patterns after relatively few iterations that are "good enough".
Right now, I'm pushing ahead with renewed vigour on Phix 2.0, in the quite probably mistaken belief it might yet make a difference, not so much any more in helping people write code, but helping them read all the stuff AI has just spat out at them. That is taking the form of getting AI to help me write bits I can later glue together, in a way I've currently no hope of ever adequately explaining to it, if you get my drift. But hey, it's my hobby, not yours, leave me alone!
I've even paid for Google AI and started training it so that it can write GUI programs for me.
You don't actually have to pay for it (but it's your money I guess). Admittedly I have to hop between Claude, CoPilot[ick], ollama, and ChatGPT, with wildly varying results, and spend long lonely hours actually coding by myself, waiting for those pesky timeouts to elapse. Now the shine has worn off it a bit, I realise how naff/dumb as a bag of rocks it really is, albeit often useful I'm not actually sure it's worth much real money at all. I'm not the expert here by a long chalk, but it sounds like 100% free in-home AI/Claude Code is now doable under two grand, the big costs being 64GB of ram and a decent GPU card. I actually set up a poor man's one here, just 16GB of ram and no GPU, but it's pretty pants, at least 10x too slow.
There doesn't seem to be much demand for desktop programs anymore.
True, but if you can somehow get the same stuff to run in a browser as well, developing on the desktop is 100x nicer.
With my p2js, that's done for over 1,200 rosettacode entries already, though I'm still working on the nice GUI front-end and some sort of clever PHP-based backend bridge (I imagine but am struggling with), to finish the job.
