• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Master AI (Artificial Intelligence) Discussion/News Thread

Anyone tried coding? It is both mind blowing - and frustrating - specially if trying to do it for free. Here is an example. I wrote the specification in the quotes below the image. I tried Gemini, Deepseek, and Chat gpt using the best models I could get free access to.

This is the result from Deepseek, after a few retries to fix bugs. So far I haven't got a usable result from either Chat GPT or Gemini - they both stop me using them after a few bug fixes. Gemini until tomorrow. GPT until later today.

While deepseek does limit access with "server is busy" messages. I don't have to wait so long until getting another go. It also creates the "prettyest" display.

Bug fixing (if just asking the ai to fix its own mistakes) can be a "one step forward, two steps back" process, with fixes often breaking the whole thing.

Deep seek generated 937 lines of code.


View attachment 448355

Did you use a tool like aider ? or just the text interface?
 
I use AI to create summaries of YouTube videos.
I use it to get detailed reports with tables to compare products or features of different models, searching online magazines, blogs and forums, telling it which countries (I make a list...) to search.
I make it do calculations, search for news and so on.
At the moment, for me it is a more advanced search engine.

It makes mistakes, is verbose and tends to repeat things, but it will surely improve over time.
 
Anyone tried coding?
I found Claude 3.7 was very good for python coding - it keeps a pretty good 'model' of the code so you can successfully ask for new features / enhancements up to a point and then if you can describe correctly how to split your application up into into separate 'modules' you can continue but eventually it does looses the overall concept and gets very frustrating to continue so you more/less have to take over.

But, I also use IntelliJ Idea ( mainly for Java but also Javascript ) and their new AI Agentic Assistant (called Junie) is actually pretty fantastic though very slow. IntelliJ keeps a internal "model" of the code with all the references / dependences and effectively this "grounds" the Agentic into the codebase very well and the agentic approach is to first split the overall problem you ask it into many logical steps ( this is on a huge codebase ) and so far it has done some excellent work ( I can even see if run the unit test - see it failing, undo the code it wrote and the try a different approach and continue. It is almost magically. I see this as the future of large scale coding with these LLM ( small stuff is fine with Claude, deepseek, Chatgpt).
 
I found Claude 3.7 was very good for python coding - it keeps a pretty good 'model' of the code so you can successfully ask for new features / enhancements up to a point and then if you can describe correctly how to split your application up into into separate 'modules' you can continue but eventually it does looses the overall concept and gets very frustrating to continue so you more/less have to take over.

But, I also use IntelliJ Idea ( mainly for Java but also Javascript ) and their new AI Agentic Assistant (called Junie) is actually pretty fantastic though very slow. IntelliJ keeps a internal "model" of the code with all the references / dependences and effectively this "grounds" the Agentic into the codebase very well and the agentic approach is to first split the overall problem you ask it into many logical steps ( this is on a huge codebase ) and so far it has done some excellent work ( I can even see if run the unit test - see it failing, undo the code it wrote and the try a different approach and continue. It is almost magically. I see this as the future of large scale coding with these LLM ( small stuff is fine with Claude, deepseek, Chatgpt).
Fascinating on the one hand, frightening on the other ...
 
Manus is great for research, is an agent on its own and is different in that to all others. Gemini 2.5 Pro dwarfs others when it comes to coding including Claude. Google might’ve held back at first, probably worried that pushing AI too hard could hurt their search business. But now they’ve jumped back in with a vengeance, and with the kind of massive resources they have, it’s hard to see anyone catching them up anytime soon.

Honestly, with how fast AI coding tools are improving, I wouldn’t be surprised if average coders are out of jobs within five years. AI might not replace every job, but people who know how to use it definitely will replace those who don’t.
 
Did you use a tool like aider ? or just the text interface?
Just dropped the text specification into the deepseek chat.

Then described things that were not working correctly for bug fixing.
 
One the other hand, there’s inescapable anxiety around the possibility that this tool, having digested vast quantities of medical information and text, is disgorging results with an inhuman and robotic absence of embodied experience and contextual knowledge, bringing an unknown degree of risk of grievous error.
I hope no one is substituting AI/Search for serious medical care. My application is that it is 2:00am on Sunday and you have chills. Is something seriously wrong with you or is it side effect of the new medication/vaccine you took? Before AI, I would do a search and top results would be whoever games the SEO system to show up on top. Now with Google AI, you get a summary with links to where it came from. It is much faster, easier to digest than clicking link after link when you are not feeling well and access to health care is very limited.
 
Access to health care will be reduced very fast in near future, due to several reasons ... If AI will cure it??
 
I hope no one is substituting AI/Search for serious medical care. My application is that it is 2:00am on Sunday and you have chills. Is something seriously wrong with you or is it side effect of the new medication/vaccine you took? Before AI, I would do a search and top results would be whoever games the SEO system to show up on top. Now with Google AI, you get a summary with links to where it came from. It is much faster, easier to digest than clicking link after link when you are not feeling well and access to health care is very limited.
I think many of us have the same experience. Based on what has been found by many in areas where we have some experience or expertise is that AI is good most of the time but often wrong about certain details. AI suffers from the age old garbage in equals garbage out. My health provider has a help line 24/7 with a medical advise professional.
 
Having made a comment previously on AI on a different thread on this forum and felt it was perhaps misplaced there, I welcome the pragmatic creation of this catch-all AI thread.
My 2 cents: I'm not against AI per se.
I acknowledge its potential benefits in e.g. healthcare research. But I would caution against the elision of 'AI' with recent LLM technology and the consequent normalisation and tacit acceptance and assimilation of LLMs into everyday life. While ostensibly useful they are fraught with errors, by design made to benefit billionaires, at their current trajectory made to demand planet-killing levels of unrenewably sourced electricity, are credibly accused of being taught by widespread theft from unacknowledged, unpaid, actual human creatives, and 'Grok'? the digital child of a nazi-saluting narcissist?
So no, I don't want to hear unqualified, misplaced inputs about the benefits of 'AI' in this, its current unreconstructed, apparently unquestioned by many, form and so will, as suggested by the OP, "just ignore it".
Good luck.
 
Just hope the benefits/advances in healthcare etc outweigh the obvious risks and pitfalls (gross understatement?) of the inevitable ongoing weaponisation myself.
(The genie is out of the bottle... Fingers crossed.)
 
I played chess between chatgpt and deepseek. (chat gpt start with white figures...i copy paste that to deepseek and viceversa...). Deepseek wins everytime :).
 
I played chess between chatgpt and deepseek. (chat gpt start with white figures...i copy paste that to deepseek and viceversa...). Deepseek wins everytime :).
Interesting, Deepseek is 5th and ChatGPT 3rd in these rankings at the Chatbot Arena
 
I played with free versions. Probably they have some chess dedicated engines for competition...?
 
ASR readers may enjoy

How AI can help supercharge creativity

Forget one-click creativity. These artists and musicians are finding new ways to make art using AI, by injecting friction, challenge, and serendipity into the process.

 
ASR readers may enjoy

How AI can help supercharge creativity

Forget one-click creativity. These artists and musicians are finding new ways to make art using AI, by injecting friction, challenge, and serendipity into the process.

It can certainly help humor :D , repost of a video I just posted in the thread ''a call for humor''
 
Slightly creepy though (and symptomatic of the majority of AI videos) They all feature young, conventionally attractive, frequently scantily clad women. One did feature a chimp - but alongside one of the women.

Just a little bit of diversity wouldn't go amiss.

Oh yes. THEY STILL CAN'T DO FINGERS (check out the fingers on the fretboard of the woman with the chimp)
Yeah...I suppose AI itself is wondering why it can't put a finger on this problem :p
 
Back
Top Bottom