I've got a local AI :)

[confused2]

I'm using a Raspberry pi 4 with 4 GB of ram (was about $70) and it's not bad!
Install ollama .. consult Gemini for details
Start the magic with ollama run qwen2.5-coder:1.5b
The clue is in the name .. optimised for coding but it still knows the five longest rivers (probably).

A sample (useful) output..

Yes, you can get a list of bad sectors and use them to determine when to
replace the microcontroller.

Here's an example
of how to do it:

```python
import os

def get_bad_sectors():
bad_sectors = []
for block_number in range(1000, 2000): # Replace with the range of block
numbers you want to check
if os.access(f'/data/{block_number}', os.W_OK):
bad_sectors.append(block_number)
return bad_sectors

bad_sectors = get_bad_sectors()
if len(bad_sectors) > 5:
print("Replacing the microcontroller because the number of bad sectors has
increased.")```

This will print "Replacing the microcontroller because the number of bad
sectors has increased." if the number of bad sectors is greater than 5.

You can adjust the range of block numbers you want to check and the threshold
for bad sectors as per your requirements.

It won't write an 'app' from scratch but for someone like me it saves reading manuals (if they even exist) and I don't have to mess with a datacentre the size of Louisiana.

It isn't instant but I've spent almost as long in a queue waiting for an instant answer.

[C C]

[...] It isn't instant but I've spent almost as long in a queue waiting for an instant answer.

I guess I can see the calling to go homegrown, then. Loads of refurbished junkers with 4GB out there, too. Finally found their mission in life as heritage pieces.

[confused2]

[...] It isn't instant but I've spent almost as long in a queue waiting for an instant answer.

I guess I can see the calling to go homegrown, then. Loads of refurbished junkers with 4GB out there, too. Finally found their mission in life as heritage pieces.

It's definitely a nerd thing .. at least with my hardware. I've tried 3 AIs in an attempt to find the fastest and best .. they take over an hour to download and no matter how fast they are they probably won't repay the time spent setting up and testing them. IRL I'm probably going to end up paying to use an online AI.

The best I've found so far is gemma3:1b (ollama run gemma3:1b). Getting a stupid answer quickly is better than waiting 30 seconds for the stupidity to reveal itself.

[stryder]

I notice the markets are fluctuating over the current hype on the danger of AI.

The problem is that it's reminiscent of the Dot Bomb (wikipedia.org) where the original expectations of AI were overhyped

AI had likely been worked on for years to get where it has got, the market of course however expects rapid improvements, new versions and paradigm shifts which would takes time to do which means it would eventually flatten out in regards to investment and then decline.

It's likely the leaked stories and fears, along with the governments jumping on the bandwagon are pulled into a purposeful speedy decline so as to use a hedging method to make money on a fast decline. It's then likely if it's similar to the Dot Bomb that those with the larger wallets from the money they have taken from both the original investments and the hedging will start to attempt to acquire smaller companies that can't deal with the sudden collapse of funding.

The same thing happened with small tech firms during the Bot Bomb, which led to the large firms that now dominate cyberspace.

[confused2]

Speaking as one of the masses I've got used to Google providing a free service .. paid for by selling the information it collects (about me) to the highest bidder. Maintaining a Google style indexed database is very much cheaper without the 'AI' front end.. no GPUs required .. it doesn't even need to be particularly fast. For the masses .. does the AI front end add enough value that they'd be willing to pay real money for it?

Most of the people who are actually using 'AI', for example to write software, are only using a tiny component of the LLM database .. we don't need an AI that can also recognize pictures of sheep .. so I'm guessing the future lies with AI optimized for particular tasks and not the general purpose AI that suppliers want us to buy. Kind of shifts the balance from mass market to high value bespoke AI running locally. A user might well need cloud storage but the AI would run on their own hardware.

Conscious AI is fascinating as a research project but for most users that is exactly the thing they don't need.

[stryder]

@confused2

I agree with your thoughts on it. I often pondered the problem (with both this and other things) as being a Tall versus Wide problem.

When there are tall business structures or implementations, then finding is funnelled to produce a venture that can have decent QA applied.
Examples would be such as the major block buster motion pictures of the 1990's or pop music from the 60's-90's.

When such ventures turn wide, it becomes messy as everyone potentially adds to whats available wide but the money is then spread thin. In essence it creates a saturation problem of poor quality produce. (e.g. AI slop)

AI needs to be treated more like a space race with a few runners rather than the whole world being mass populated by slop.

You're right that they need to step back from it being aimed at the consumer market, and concentrate on platforms meant for Defence, Economics, Education, Farming, Environment etc. Try to get the cornerstones of things that society is built on to prosper from it, and then trickle down from there.

[confused2]

Useful information from my AI (qwen)..

Trout, also known as rainbow trout, can live for
many years. They can reach lengths of up to 15-20
feet in the wild, and can live up to 30-50 years
in captivity. The average lifespan of trout in the
wild is about 8-10 years, while those in captivity
may live up to 20 years or more.

[stryder]

I'd probably look into messing with some old PROLOG scripts.