General browser info
Rsadeika
Posts: 3,877
Not sure if anybody has or is using the 'brave' browser. The interesting part is the AI that is used when you request some info. For example:
when will this AI start getting some good information about the Propeller P2 languages.
This is the question that I asked.
The answer:
"
As of now, there is no specific timeline available for when AI systems will start providing high-quality information about Propeller P2 languages. The development of language-agnostic APIs for Propeller 2 hardware abstraction layers (HAL) is an ongoing discussion, with proposals suggesting the use of modified Spin syntax to create a common interface across different programming languages.
However, this effort is still in the early stages of community discussion and has not yet resulted in a standardized implementation or integration with AI systems.
The idea of a language-agnostic API for Propeller 2 HAL implementations is being explored to simplify cross-language development, but no concrete progress or release date has been announced.
While AI systems are rapidly advancing in understanding and generating human language, including complex technical topics, their ability to provide accurate, up-to-date information on niche hardware platforms like the Propeller P2 depends on the availability of comprehensive, well-structured data.
Currently, AI models remain largely opaque in their internal workings, though research into more interpretable models is progressing, which may eventually improve their reliability in technical domains.
However, this research is still experimental and not yet applied to hardware-specific language support.
Until such time as a standardized, well-documented API is established and widely adopted, AI systems will likely continue to provide limited or inconsistent information about Propeller P2 languages.
"
Not sure what that answer really means, but I do find using 'brave' for other stuff very useful, does not seem to have any sort of bias that is noticeable.
WOW, this technology is something else.
Ray
Comments
There's a good reason you didn't understand it. That's a pile of garbage spewed out. There is no HAL or generalised APIs being discussed. Any Spin based discussions ended when the Prop2 was launched without it hardwired. It's all vagueness with no links or quotes that it should be basing its statements from. That said, Flexspin has a bunch of mix'n'match capabilities internal to itself.
And LLMs ("AI models") are not advancing at all. Nothing's changed in the three years of public attention they've had. There's zero signs of progress on AGI. They're still like a search engine, they can't reason or design something that isn't just a regurgitation.
Just ho-hum, and what a waste of money.
The way that I am using the 'brave' AI is, specific programming questions. At the moment my latest project is with python and some rpi units.
I decided I wanted to learn how to use python sockets, when I asked the specific question, the response was an understandable explanation, and it provided functional python example code. The program code actually worked, which to me was very amazing, considering I have had some example code that did not really do the job.
The one thing that I did learn is, your questions have to be very specific, otherwise the response is very vague. Before this, when I asked a specific question it would just come up with a bunch of links that did not really answer my question.
The next query that I will be after is, can the P2 become a server using sockets.
Ray
There are millions of Python programmers, and probably more than a billion or more lines of Python code to scrape -- this makes a big difference. There are far fewer of us coding in Spin. My approach to AI is not to ask for Spin code, but to ask for generic code that can "run on anything" (which usually results in C-like code) that I adapt myself. I've been doing this with OLEDs that I want to experiment with. Converting from AI generic code to Spin/Spin2 takes little time in most cases, and I fix the AI mistakes/hallucinations along the way.
You know asking a search engine gets the same results. There is many Python examples in the open already. And that's also why the LLM can give a decent answer. It's just spitting out copies of what is already done by people.
Not amazing at all. Expensive and inefficient.
If the topic is "General browser info" then it is my duty to shill for Vivaldi.
I've been using Vivaldi since 2018 or so and it never let me down. No shady dealings, no attempts to push stuff I don't want, no unnecessary forced UI changes, can configure just about anything you want and the mobile version has ad block.
I have noticed a few people have a problem with AI, I consider AI to be a tool, if the results are not what you expect, then is it the tool or the user that is at fault.
Ray
I pointed out how flawed it is for your uses. I'll leave you to decide how much of that is the tool and how much is the user.
PS: Really, my biggest beef is the extraordinary waste of investor money, and related electricity and water costs, being spent by these irresponsible LLM companies building all those data centres. For what is mostly a gimmick.
PPS: Here's a graph of quarterly money already spent on just the servers that go in these data centres. A six fold increase in an already mature market is completely stupid! And possibly what's even more nuts is that "New Normal" is actually supply side limited. It would've been even higher if manufacturing was able to able to keep up.
Nice graph, seems to show, that we are nearby the "Peak of Inflated Expectations"
https://en.wikipedia.org/wiki/Gartner_hype_cycle#/media/File:Hype-Cycle-General.png
Hmm, it's merging with established replacement purchasing, it has to return back to replacement levels. Which, if LLMs in data centres becomes a write-off, would be right back at $20B per quarter. Or a little higher, say $25B, if LLMs end up being of some value. There's gonna be a lot of defunct data centres either way.
I absolutely agree that A.I. is a tool ... and like a tool, a hammer, screwdriver, etc. a tool can be used constructively or destructively but it is YOUR decision on how to use it.
A good friend of mine summed up A.I. in one sentence .... "A.I. is basically a house of mirrors that reflect averaged bits and pieces from a communal contribution." <- The HIVE mind.
Honestly I think it is time for a rip-up and redo that looks a t LLM's from a foundation approach rather than a vast endless lookup table. An example of what I mean... remember sentence diagramming from grade school? Most people hated it, but I sorta liked it ... to me it was like programming. The fundamental approach of A.I. should look at language from that perspective to build and make new connections.