flembat

When purchasing any ARM based computer a key question for me, is how many of those can I purchase for the cost of a Mac mini, and how many Mac mini can I purchase for the cost of that, and does that have working drivers...

I have a couple of earlier RISC V systems that were advertised as nearly desktop performance: I always like unconventional systems, but cant find a reason to like these, they are much slower than similar priced arm systems, the software/hardware support is not as good, and the instruction set is also just not that interesting. Also once you run Linux, you are just running Linux, it is just like Linux only harder to install, and slower.

This is why I will never get in an AI plane.

That is quite a confession from AMD. It's not X86 at all, just every implementation. It is not like the ARM processors in Macs are simple any more, thats for sure.

I have had gemini running as a qa tester, and it faked very convincing test results by simulating what the results would have been. I only knew it was faked because that part of the code was not even implemented yet. I am sure we have all had similar experiences.

Look at computer systems that cost 2000 or less and they are useless at running LLM coding assistants for example locally. A minimal subscription to a cloud service unfortunately beats them, and even more expensive systems that can run larger models, run them too slowly to be productive. Yes you can chat with them and perform tasks slowly on low cost hardware but that is all. If you put local LLMs in your IDE they slow you down or just don't work.

The execs need to try actually using AI and verifying the correctness or lack of it in the results, THEY need to be more savvy about what AI is able to do. And if the results are garbage, and it wastes time rather than saves time they must be using it wrong, right?

You will find development much faster if you use a computer. Bad joke I know, when I was a kid I wrote out my programs on paper first and at the college you had to submit on paper so the county mainframe could run them and send back the results a week later... you can be much less careful about your thought process, when you get instant feedback.

There is a solution for young people they should look at this demographic disaster and realise they need to have children for the sake of their own future. Meanwhile the next generation or two to retire is going to have a very grim time surviving because we had two kids or none, compared to the five our parents had.

In the UK You don't get paid for lunch which is why real white collar hours worked are more like eight to six or a lot longer. And typically you get to work through your unpaid lunch, while your contract says 37.5 hours and 'any other time necessary to complete your work'. At least that's been my experience for the last forty years or so.

Let's be real the goal was just to write some rust.

I let a popular agentic ai refactor some code for me today, it looked really nice, but despite the fact that it was meant to be splitting a file full of functions that already compiled and worked it rewrote them all and broke them comprehensively, it then tried to research how to fix them, even though working code was right there in the file and just kept making the code worse and worse. It also had some limit which meant it left the code partly refactored, ran out of its quota left everything broken and then when restarted suggested refactoring some more classes.

I think it's the antivirus and security protection that makes windows slow at opening apps. As for people saying the M series macs are slow, those are the first macs that felt fast to me in many years.

When workers save time, through automation, they are expected to complete more work. They don't get paid more. It is still the same job. Of course the employer benefits. The reason productivity does not increase is all the incidental nonsense and time wasting activities that employers like to indulge in.

When the puppet pulls back on his strings.

Really great idea, I currently work for an AI, compiling and debugging its code, at least that's what it sometimes feels like. Who is the agent here exactly? The fact that the AI has no understanding at all of what we are doing, and does not apply the information it does know to solve problems is challenging. At least if it debugged the code, it would be able to see that it is clobbering the same registers that it is using, instead of me having to explain that to it. Fortunately I am talking about my hobby projects, I pity people who are doing this for a living now.

Must be all the cholesterol they warned us about. I am sure we were told not to eat eggs. Well cool hand Luke eat them and he had a long and healthy life.

FAA post authorised the alien drones to fly out of the sea and hover over people's rooftops. case closed and no lies told.

Orlando springs to mind, the pavement just siddenly ends and people who see a couple walking with a kid pull over and offer assistance, from personal experience.

People who renounce contamination from digital content will be paid by AIs to answer their questions on simple terminals.

I needed transformers to talk to a local copy of Qwen, and a GUI, using tkinter on macos intel ventura. It all worked eventually. I have not got the AMD GPU to work yet. I don't hold python in anyway responsible for the GPU problem. There was some issue with urllib, and it still complains that my numpy is the wrong version. Its all ok.

Well I was describing what I actually experienced last week. I did get my script to work after hours of faffing around, your article may have been one of the articles I read, in any event thanks for writing it. There is no way I would describe the package installation process I experienced as automatic, since pip had no idea what requirement was missing. Python was in an environment, a package used by an import statement required it to be downgraded so I had to downgrade it, by finding installing and setting up an older version. I then ran into other problems I described. Once python is set up and working it seems fine. Everything new I do in python seems way harder than it should be.

As for basic in ROM, I am not advocating BASIC, but many versions of BASIC (not in rom) have had built in support for GUIs for decades.

Write your python, run script, find missing import. Install missing import, find that you actually need a different version of python as the import does not understand the latest version yet. Find that the import depends on various other imports, which python and the package manager, does not tell you it needs. Downgrade your local python. (and lets not discuss managing multiple versions.) Import library again and find you need to install a C compiler. Install a c compiler. Later find that the library is not compatible with your systems version of openssl. And it turns out you now need a version of python that has some extra static code linked into it, because you need to display a simple GUI. But you still need the GUI library as well.

The irony is that an LLM that knows more about most topics than most teachers, and most parents, and can patiently discuss any topic one to one, is a superb way to learn new information. Its early days, right now, children need to be taught to verify the LLM is not making things up, by checking for references, look at additional sources, and use the LLM to LEARN about topics not COPY answers. Beyond that we need to consider what really needs to be held in our own minds now and what detail we can leave to 'AI' tools in future. Spending hundreds of hours learning things you can just ask an LLM to do is not that useful.

I would like to see this new knowlege used to generate tiled desktop wallpaper. Also does this tiling seem like a form of compression?

Free speech, should not just be for people we agree with. Although I agree conspiracy to commit a crime is something else.

My point was that AI faked content could certainly get you locked up in the UK.

An individual is not responsible for the culture or government in the country they live in.

In the UK a government was just elected with a historical absolute majority by only ten million people, and now first time offenders are being sent to prison for making stupid offensive statements online.

It was a long time ago, but I vaguely recall using a standard graphics extension on CP/M. It had a lot of gray scale fill patterns and could draw geometric shapes on a bitmapped display. I was using a z80 system on green monochrome display, and enjoyed watching the graphics slowly plotted onto the display.

Is there a font to use that contains these? I am really glad to see them. In the past when hoping to write a game in a terminal I found only one space invader symbol.

It helps a lot if you use a strongly typed language with a strict compiler and get the LLM to write plenty of tests. Then you need to understand that the tests are logically correct. The LLM is also good at documenting the functions, so you can review that matches your intentions and the code as well. Also the LLM will pick functions from your own source code library to compose new programs for you. So the reuse of your own well tested code should increase confidence.