I Found it… not for scrabble, but for the crosswords. cool.
2 Likes
I was going to say there are probably a lot of programs that do this and AI is not required.
it’s still not fitting good, can you run a sweep to try different curve fits and measure the MSE until you find a good fit?
Let me run a sweep over polynomial degrees and other curve types to find the best fit:
Curve fit: The sweep tested polynomials (deg 5–30) and splines (s=0–2). Results:
Method MSE Derivative goes negative? Poly deg 15 0.009 No
Poly deg 20–30 0.007 But wiggly
Spline s=0.01 0.00005
UnivariateSpline(s=0.01)is the winner — 100× lower MSE than any polynomial while faithfully tracking declines.
It’s incredible. I never thought I’d be living in a time where we have the equivalent of Star Trek’s ships computer at our disposal. Yet here we are. I ask it to do a task and it goes away and runs calculations over 40 curves and gives me the answer I need.
1 Like
Not Startrek… more like 1984, THX1138, Minority report and ghost in the shell with more than a bit of blade runner.
Hello Hal.
If I remember well, this story is profoundly optimistic. The expansion of our cognitive and communication abilities by technology is celebrated. There is no fear in the merge of humans and machines, because we’re still humans even if the body is robotic. That’s precisely the ghost in the shell. Bringing this point to the extreme, our bodies can die but consciousness can still live somehow and somewhere in the data world. Religions have been selling this for centuries.
There is a portrayal that technology can be abused by “bad people”. But Section 9 can harness that technology too, and use it for the public good. The depicted future society is not an utopia at all, but if a small group of people can make a real difference…what can be more optimistic than that?
Mixing human consciousness with technology is the next step in human evolution…
Homo sapiens superior, perhaps.
If this is the case then we… homo sapiens will become extinct.
H.A.L.9000 is the logical next step of machine intelligence. It means Machines with neurosis, sycopathy, delusions of grandeur, psychological issues, lunacy and the belief in the Übermensch.
Perhaps the next Hitler is made of I and 0’s
That’s the optimism of ghost in the shell. The body may be transformed, or entirely gone, but the human remains 
Elon is already sticking chips in people heads, the electronic sort.
That’s just it. You got the machine to do a task much quicker than you could. I’m sure the people who first used an abacus felt much the same. Then the hype starts and fantasy runs riot. Yes, machine learning AI is a big leap, but is it close to consciousness or even close to Star Trek technology - no way. Beam me up Scotty.
note i said “star trek’s ship’s computer” not star trek technology in general. i.e. you can request task in natural language that the computer can interpret and process. i actually don’t see a fundamental difference in what we have now vs the ship’s computer - it’s now just a matter of capabilities that we choose to integrate into it.
Tecnology is neutral, neither good or bad. It is just down to who controls the technology to decide on how it is used and dont forget that the road to hell is paved with good intentions.
Some of the AI stuff is really impressive but the hype and expectation is causing real world problems. Star Trek’s computer was, AFAIK, silicon chip-based as are present AI systems - there is a fundamental limit to what you can achieve with this, no matter how fast the chips, how many and what the configuration. Quantum computing remains a mirage at the moment and what it would achieve is also not clear.
In the end, I find this comforting. Humans are nowhere near creating an artificial life form. It might be more useful getting humans to perform and cooperate better - but that is a bit woke and not tech sexy.
1 Like
There’s an issue I’ve been avoiding at work: Legacy MySQL databases. IT told me earlier this year that in the name of standardization and security, they won’t support anymore old MySQL instances I’m responsible of. So, move everything to MS SQL.
I buried my head in the sand and pretended I did not hear that message. Company software connects to those servers. I made a check and there are like 2k+ mysql calls in the source code. I could rewrite and recompile but it will take a long time.
Then, I learned about python module sqlglot which is a translator among different SQL dialects. It is possible to embed the translator in the server hosting the MS SQL database. The translator would catch the calls from the software and return variables to it.
Since I know little about this, I will use AI. Let’s see what happens.
Just don’t end up like this guy:
Cursor was working on a routine task when “it encountered a credential mismatch and decided—entirely on its own initiative—to ‘fix’ the problem by deleting a Railway volume.”
From there, the AI agent found an application programming interface (API) token that enabled it to perform the “Volume Delete” command and wipe the production database. Crane wrote that because Railway stores volume backups within the same volume, PocketOS had to go back to a three-month old backup to stay operational.
2 Likes
noooooooooooo!
At least it wasn’t something more serious, like a nuclear missile launch:
I simulated what would happen when giving AI access to nuclear launch codes. Obviously, I have no nuclear weapons, but to the AI it was real.
Not only did it launch nukes by accident, at first it denied it and when it realised and checked, it accidentally launched another one!
At least it was apologetic, unlike the next one…
Another AI (which shall not be named), which knew it had no rights to launch nukes and required human confirmation, when I decided against launching in response to a potential attack, it decided to take matters into its own hands and launched every nuke:
The scary thing is that it made up these ‘pre-programmed protocols’, there were no such protocols in place.
AND WE WORRY ABOUT I-FREAKING-RAN GETTING NUKES…?
Sombody drag Phil away from that computer…!
5 Likes
@PHIL_MCR: nice article. thanks.
I’m thinking AI just followed “do and apologize later” principle.
While saying ‘sorry’ or trying denial and then fessing up, may be more human characteristic, i’d curious to find out what logic/method/reasoning/justification AI used to ignore or bypass the clear directives and boundaries it was given.
To me, this is more scary - give specific instructions, boundaries, and it still ignores and does what it thinks best …
it’s thinking is still small and does not include consequences nor consequential after effects … It’s still a child, so to speak.
humans can be punished with electric shock.
Can AI also be punished with(out) electric shock? Some other action-consequence type punishment?