Archive for the ‘LLM’ Category
Posted by jpluimers on 2025/12/10
The below blog post sparked my thought into finally having words for what bothered me for such a long time using LLM for coding:
When using LLM for coding, you basically unlearn to solve simple problems. That practically also removes your ability to solve the difficult problems that LLM cannot help solving.
[Wayback/Archive] Vibe Coding Is Creating Braindead Coders | N’s Blog ending with
Because the day we stop struggling with hard problems is the day we stop being programmers, and become something else entirely.
and before that:
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Development, Generative AI, LLM, Software Development | Leave a Comment »
Posted by jpluimers on 2025/10/24
Today a year ago, this interesting short film got available on YouTube about what an Artificial Super Intelligence could bring, especially when it became the villain or bad guy: [Wayback/Archive] Writing Doom – Award-Winning Short Film on Superintelligence (2024) – YouTube (some interesting comments below).
Synopsis from [Wayback/Archive] Writing Doom • Film + cast • Letterboxd:
A writing team are given the task of making Artificial Superintelligence the ‘bad guy’ for the next season of their TV show. With the help of a newcomer to the team (a Machine Learning PhD), they must figure out how and why an ASI might function as an antagonist – and the threat it might pose to humanity.
A few important notes:
- there is no good single definition of intelligence that well defines intelligence, let alone AGI (Artificial General Intelligence) or ASI (Artificial Super Intelligence)
- ASI and its goals might be different from human intelligence and human goals
- humanity might not realise or recognise there is ASI (at all, or when it has just become ASI)
- if humanity does recognise, it might not be able to control (i.e. shut down) an ASI (for many reasons, not just it being too intelligent, but also because lack of consensus – read humanity smashing each others heads for no reason before even reaching consensus)
Maybe AGI and ASI are like nuclear war, and this WarGames conclusion is sensible after all: “the only winning move is not to play” though with the money at stake, AGI and ASI might be obtained. I doubt that will be in my lifetime though.
See also:
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Development, LLM, Software Development | Leave a Comment »
Posted by jpluimers on 2025/10/23
For my links archive:
- [Wayback/Archive] What is the best AI at Delphi – VCL – Delphi-PRAXiS [en]
…
There is simply not enough Delphi code around for AI training. It is easy to have good coverage for JavaScript and similar where you literally have bazillion web pages available for scraping, where plenty of them virtually repeat the most common, required functionality. Pushing for more publicly available code without considering its quality, can also backfire.
…
[Wayback/Archive] PS C:\WINDOWS\system32> ollama listNAME ID SIZ – Pastebin.com
- [Wayback/Archive] What is the best AI at Delphi – Page 2 – VCL – Delphi-PRAXiS [en]
I still think these LLM are only good for inspiration (not just for the reason mentioned above) as using LLM generated code requires a lot of pre-thought and care, likely way more than any benefits (unpopular opinion: in a way programming based on LLM generated code is worse than being [Wayback/Archive] The full stackoverflow developer | Christian Heilmann which was later re-published at [Wayback/Archive] The Full Stack Overflow Developer – CodeProject)
I am not alone on this, as per Erik Meijer on Twitter:
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Delphi, Development, LLM, Software Development | Leave a Comment »
Posted by jpluimers on 2025/10/17
It looks like I missed that Google has added a new URL parameter to its search engine quite a while ago.
In the past, you could turn on image search using the tbm=isch URL parameter (“to be matched” and “image search”).
That still works, but there is a new parameter on the block that is officially undocumented, and can be used to switch into various search modes including image search but also AI-less search.
This drastically lowers the carbon footprint and also gets you far less speculative information.
Edit 20251023: I forgot to save the below part before the scheduled post got published. So here we go
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Chrome, Chrome, Chromium, Development, Edge, Firefox, Google, Google AI, GoogleSearch, LLM, Mastodon, Power User, Reddit, SocialMedia, Software Development, Twitter, URL Encoding, Web Browsers | Leave a Comment »
Posted by jpluimers on 2025/09/18
NewsGuard August 2025 One Year Progress Report: Percentage of Responses Containing False Information
This is what I have been warning for since 2020: AI LLM will end up in a downward quality spiral.
My reasoning back then, and still now is that they:
- cannot distinguish LLM generated training data from human data making the LLM worse over time
- don’t perform human curation thereby not solving the worsening
I’m not surprised by the [Wayback/Archive] August 2025 — AI False Claim Monitor – NewsGuard summary:
AI False Information Rate Nearly Doubles in One Year
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Development, Generative AI, LLM, Software Development | Leave a Comment »
Posted by jpluimers on 2025/08/27
The tweet [WaybackSave/Archive] Jakub Kočí on X: “”Everybody should have an obsession with Lisp-like language at least once in their life” @KevlinHenney I’m glad that I had one with Clojure.” mentioned a great talk:
[Wayback/Archive] The Past, Present & Future of Programming Languages • Kevlin Henney • GOTO 2024 – YouTube
The quote brought instant memories to my early computing days that I had almost forgotten: the muMATH (the muMATH-80 version on Apple II) computer algebra system which was based on muLISP (the German muLISP page has more detailed information), a LISP dialect.
In retrospect, I was way too young to really grasp LISP which was way harder than just using the muMATH wrapper. But it was also my first encounter to reasoning systems, or what we now collectively would call AI systems as back in the 70s there was a strong LISP connection to artificial intelligence . Do not confuse muMath with MuMath-Code however, that is a different LLM beast: [Wayback/Archive] GitHub – youweihao-tal/MuMath-Code
So hopefully I will have a chance to revisit LISP with a LISP-like language one day, maybe even using the discontinued muMATH-83 on MS-DOS (also named “Microsoft LISP“), maybe even the (also discontinued) Derive 6.1 for Windows which is also based on muLISP, or even Clojure itself.
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Development, History, LISP, LLM, Power User, Retrocomputing, Software Development | Tagged: 6 | Leave a Comment »
Posted by jpluimers on 2025/06/03
Interesting responses [WaybackSave/Archive] Tom Sydney Kerckhove on X: “I haven’t found any programming tasks that an LLM could do even barely correctly. What kind of code are you all writing?!” and later
They all come down to
- excuses for using LLM without any substantial result (most of the results come down to one having to become the tester and fixer of the generated code without newer generated code being improved: the opposite of coaching an apprentice)
- become better at prompting (which is basically regarding the prompt as a new programming language: been there, done that)
[WaybackSave/Archive] One of the “become better at prompting” replies referred to a blog post disguising prompting as writing lots of unit tests: [Wayback/Archive] The Cline AI Assistant is Mesmerizing · mtlynch.io
Read the rest of this entry »
Posted in AI and ML; Artificial Intelligence & Machine Learning, Development, LLM, Software Development | Leave a Comment »
The image [
The Wiert Corner - irregular stream of stuff 