I'm playing around with the Grok AI and so far its an absolute beast - I really like it and compared to Gemini and ChatGPT the answers are much more coherent. A little bit frightening maybe, but absolutely helpful.
No experience but i object to them stealing Heinlein’s term for knowledge/understanding. Alreay ruined Tesla’s name —nat least short-term
Grok is a neologism coined by the American writer Robert A. Heinlein for his 1961 science fiction novel Stranger in a Strange Land. While the Oxford English Dictionary summarizes the meaning of grok as "to understand intuitively or by empathy, to establish rapport with" and "to empathize or communicate sympathetically; also, to experience enjoyment", Heinlein's concept is far more nuanced, with critic Istvan Csicsery-Ronay Jr. Wikipedia
https://en.wikipedia.org/wiki/Grok
dave
Grok is a neologism coined by the American writer Robert A. Heinlein for his 1961 science fiction novel Stranger in a Strange Land. While the Oxford English Dictionary summarizes the meaning of grok as "to understand intuitively or by empathy, to establish rapport with" and "to empathize or communicate sympathetically; also, to experience enjoyment", Heinlein's concept is far more nuanced, with critic Istvan Csicsery-Ronay Jr. Wikipedia
https://en.wikipedia.org/wiki/Grok
dave
Ah, the xAI effort - that just took over "X" for how many billion was it? Facebook may be best for discovering and communicating with everyone you've ever known, but I personally wouldnt touch either with a 10 ft pole. I'll gladly go along without, thank-you.
Is that just the Billion Dollar Baby shuffling money around, cause he's short of cash?
jeff
It reminds me of the old joke about calling your company 'X' and cashing in on the free publicity.
OK, Moderators, please delete this thread. Edit: You all should know better. I should have known better - I don't want to risk my membership here.
AI and the people creating it are certainly "controversial" right now. I find more distaste in the people, than the AI...
iTony, iDave, iNelson, iZenMod, ietc...
Then we can IPO ourselves and retire in a life of luxury, with enough money to buy a slew of PL XS300s and Maggie 30s and a magnificent tropical compound in Pago Pago... upwind of the fish cannery, of course.
In the old Temple of VTEC, we changed our user names to EX. I was TonyEX.... ( Honda trim levels ended at EX then ).... We all thought that "Type R" as in TonyTypeR, was too long. Unfortunately TonyEX did not ring the IPO bell like iTonyEE.
Then we can IPO ourselves and retire in a life of luxury, with enough money to buy a slew of PL XS300s and Maggie 30s and a magnificent tropical compound in Pago Pago... upwind of the fish cannery, of course.
In the old Temple of VTEC, we changed our user names to EX. I was TonyEX.... ( Honda trim levels ended at EX then ).... We all thought that "Type R" as in TonyTypeR, was too long. Unfortunately TonyEX did not ring the IPO bell like iTonyEE.
Last edited:
I have used Grok quite a bit for help in coding. Mostly Python, js, css, HTML. It’s not the best AI coder, but it’s way faster and better than I am - and it’s so darn enthusiastic. 🙂 And it’s good at fixing mistakes, its own or mine.
Claude and GPT get the code right faster, but not always. Google AI as used in Colab is also good at checking and debugging code.
Claude and GPT get the code right faster, but not always. Google AI as used in Colab is also good at checking and debugging code.
I doubt AI is good at writing device drivers and interrupt handlers... in a real programming language?
And, can it do multi-core, multi-processor, networked Python? Which, BTW, is a script.
For Python, believe it or not, Visual Studio is awesome... so is Komodo.
Trust me, somethings are simple... like writing another ring buffer, perhaps creating some framework for a known design pattern, but for "software" and "firmware" you can not come up with standard ways of doing it... which is how the AI induction tree "learns".
Writing "code" is not doing software. It's like saying that getting a driver's license allows you to race in F1.
Writing JAVA and html stuff for web interfaces is not real software engineering.
0xDEADBEEF
And, can it do multi-core, multi-processor, networked Python? Which, BTW, is a script.
For Python, believe it or not, Visual Studio is awesome... so is Komodo.
Trust me, somethings are simple... like writing another ring buffer, perhaps creating some framework for a known design pattern, but for "software" and "firmware" you can not come up with standard ways of doing it... which is how the AI induction tree "learns".
Writing "code" is not doing software. It's like saying that getting a driver's license allows you to race in F1.
Writing JAVA and html stuff for web interfaces is not real software engineering.
0xDEADBEEF
Last edited:
That's the point.
I always try to verify it and never cut and paste code.
Some while back I was looking for a way to transfer information from one program on windows to another on android. I asked for suggestions on libraries for encoding. And got a really helpful list. After a bit of experimenting it turned out the quickest (and simplest) way to do it was an sqlite dB. Which the AI didn't suggest. Slqite is insanely fast and entirely text based, therefore completely portable.
Yeah, AI is limited to patterns (solutions) that are common. Those things can be fed into an induction machine.
In the 90s I learned LISP doing AI...... Hint, never bothered to put that in my resume.
For the rest, well, there are so many problems, solutions and so many tools.
BTW, I do cut and paste code quite often... when we buy hardware it normally comes with "sample code" in the Programmer's Manual. You're not supposed to post it freely on the Internet but you're welcome to use it for your own use - because you paid for it. It saves me a lot of time and grief, even if I spend a ton of time fixing the code because it's not necessarily the nicest and cleanest you'll see.... Right now I've been removing 'GOTO's where an else if construct should have been used.
SQL.... Once I figured out that the temporary tables it builds when you specify searches can also be used as temporary variables. Hmm... so I rewrote some of the scripts using that "feature". I was taking a customer course, I was the only internal R&D employee.. the customers wore clip on ties, I wore jeans.. So the next morning I showed them my work. Took reports that ran for 2 minutes to create down into the 20 second range! The instructor was impressed, the guys with the clip on ties didn't care. That was my one and only time doing SQL.
In the 90s I learned LISP doing AI...... Hint, never bothered to put that in my resume.
For the rest, well, there are so many problems, solutions and so many tools.
BTW, I do cut and paste code quite often... when we buy hardware it normally comes with "sample code" in the Programmer's Manual. You're not supposed to post it freely on the Internet but you're welcome to use it for your own use - because you paid for it. It saves me a lot of time and grief, even if I spend a ton of time fixing the code because it's not necessarily the nicest and cleanest you'll see.... Right now I've been removing 'GOTO's where an else if construct should have been used.
SQL.... Once I figured out that the temporary tables it builds when you specify searches can also be used as temporary variables. Hmm... so I rewrote some of the scripts using that "feature". I was taking a customer course, I was the only internal R&D employee.. the customers wore clip on ties, I wore jeans.. So the next morning I showed them my work. Took reports that ran for 2 minutes to create down into the 20 second range! The instructor was impressed, the guys with the clip on ties didn't care. That was my one and only time doing SQL.
Last edited:
In the early 80s I worked on a search engine project aimed at defense and police (among others). It was quick. We sold it to a super secret department, and after a while got a note from them reporting progress and performance. I was told they had a query that took 4-6 hours to complete. I was shocked about the length of time, only to be told they were thrilled because on their old system it normally took 3-4 days...
Yeah..... even by the late 90s.... we had a data processing algorithm that took 28 hours to complete, which was an issue since the report was supposed to daily with fresh data. It was done in C and FORTRAN under Solaris.
We ended up rewriting the math device drivers ( div by 0 was the big one ) to handle it all differently. We brought it down to 16 or 18 hours... That was with several top of the line Sun machines running in parallel.
Today, you could write crappy Python ( there is good Python ) in an iPad and it would complete in 4 hours.
It used to be you had to be good to get your software accepted, today, people get away with some of the crappiest stuff I've ever seen.
We ended up rewriting the math device drivers ( div by 0 was the big one ) to handle it all differently. We brought it down to 16 or 18 hours... That was with several top of the line Sun machines running in parallel.
Today, you could write crappy Python ( there is good Python ) in an iPad and it would complete in 4 hours.
It used to be you had to be good to get your software accepted, today, people get away with some of the crappiest stuff I've ever seen.