AI

For those of you who may be interested. I have DeepSeek running locally as part of my HAL 9000 setup :D

I just need a bigger boat ( a mac with a s ton load of unified memory ) and should be good to go :)
It is so cool having my own bespoke webui as well :D

View attachment 205662

Also on another note, the UK and US have today refused to sign the international AI declaration.

You do not have permission to view link Log in or register now.

Nice one! How much space (GB) does your setup take in total? I have a 1 terabyte Samsung SSD lying around, so i might try to stick it in there :D

Ah, and how you're going to keep updated in case a new version comes?
 
Oh the LLM's are relatively small. Works out about 4 - 5 Gb for a 7billion parameter LLM. The six LLM's I have downloaded and using with Ollama, take up prob around 30GB all told. Stable Diffusion for images, I am using a model that is just over 4GB.

Disk space is not an issue, even though my M2 Pro MacBook Pro being the binned model has a 512GB SSD. Memory is the kicker. Trust me 16GB unified memory does not even make the 7Billion parameter models responsive. When I say responsive, I want it to be around 5 seconds tops.

Hence I have worked out I will need a Mac with ideally 64GB unified memory. Likely to go the M4 Pro Mac Mini route. But with the new M4 Max Studio due soon, I am holding out for that :)

As far as storage goes, on my main MacBook I have two Samsung T7 drives attached via USB C providing a combined 1.5TB , plus I also have a Synology Nas Server which is 4TB.
 
Ah, and how you're going to keep updated in case a new version comes?
Missed this, At the moment, whilst I am stuck with my M2 Pro MacBook, I will be manually updating them as and when. Pretty straightforward and just a quick edit so the webgui is updated accordingly.

But when I get my new dream machine lol - I will automate the process so it checks. It must be easily doable :)
 
Has to be said I am now firmly back in Chat GPT's camp and prefer it to Claude. I have been having an indepth conversation with it this morning with regards my local LLM setup and bespoke webui. It is going to fine tune it for me :D

Screenshot 2025-02-26 at 08.16.43.webp
 
In other AI news, Google's Super Bowl advert (famously some of the most expensive advertising in the world) included a section where they were showing how great their Gemini AI is at helping businesses, and they used a Dutch cheese seller to demonstrate this, with some Gemini generated business promotion blurb.

Unfortunately Gemini regurgitated the entirely wrong (and easily disprovable) claim that Dutch Gouda accounts for 50-60% of cheese eaten worldwide, because the sites it'd 'learned' (stolen) from had that down as an error.

Once this was pointed out to them with lots of LOLs, Google quietly did a complete in-place replacement of the video on YouTube with the error removed.

You do not have permission to view link Log in or register now.


View attachment 205653


For me, it's easy to forgive AI when it makes an error, as I know at its core, it's "trying its best".

You get out what you put in though, so why not find a way to apply it to your life, rather than look for cheese statistic mistakes?
 
Ok the world of LLM's and being able to run them locally on a desktop has just got very interesting. With the release of the new Apple Mac Studios, particularly the Studio M3 Ultra that you can spec with 512GB of unified memory for an eye watering $10k!!

A serious, very very serious rival to Nvidia. Mindblowing in fact!

Don't take my word for it though.

 
Ok the world of LLM's and being able to run them locally on a desktop has just got very interesting. With the release of the new Apple Mac Studios, particularly the Studio M3 Ultra that you can spec with 512GB of unified memory for an eye watering $10k!!

A serious, very very serious rival to Nvidia. Mindblowing in fact!

Don't take my word for it though.


Have a watch of this, I haven't seen the new mac studios news yet but with the announcement of the framework desktop just over a week ago it had already got very interesting. Additionally you can chain together framework desktops or their motherboards and use them networked together to run models like the full DeepSeek R1 671B.

 
Yes chaining machines together. Alex in the video above has already done that and is indicating he is going to do the same with the new Mac Studios. I mean you need serious deep pockets, but compare that to renting x number of servers with top of the range Nvidea cards, it works out cost effective. What's more you run it locally!

AI tech is currently blowing my mind
 
Yes chaining machines together. Alex in the video above has already done that and is indicating he is going to do the same with the new Mac Studios. I mean you need serious deep pockets, but compare that to renting x number of servers with top of the range Nvidea cards, it works out cost effective. What's more you run it locally!

AI tech is currently blowing my mind
Ah yes I see he mentioned framework in the video, I never bothered watching it before replying as I do not like apple so have no interest in their products.

They are doing a great job with tv shows though but they still don't get any money from me for them!

I just realised that account I posted a video from has 4M followers and 74 videos totalling views in the billions... Thats one way to monetise some excellent AI prompting! Must be making a hefty chunk of change off those.
 
There's a lot of 'critical mass' stuff coming up with the AI.... cough.... 'businesses' in the next twelve months or so.

This guy is worth keeping an eye on -
You do not have permission to view link Log in or register now.


OpenAI could go pop at this rate, the rate at which it's setting fire to money, with no path to profitability, is gobsmacking.

You do not have permission to view link Log in or register now.


1745391388880.webp


If you don't fancy taking on a read that long, he has it in podcast form as well.

 
Normally Google is good with results but this AI search result had me laughing/confused. Watching a documentry about the TV soap Neighbours I wanted to see who was still alive and...

View attachment 207743

OK wow I have never seen AI do this on my search, I guess its different in countries and regions? I have so far managed to turn off AI in everything, but I think I read Google will be using AI on default by next year.
 
Nice find. Will add to the podcasts I listen to when walking the dog. Meanwhile there are reports stating that AI could go all 'Skynet' on us in the following decades!

You do not have permission to view link Log in or register now.

Yeah OK :) I'll put AGI wiping out humanity on the list of things that's definitely happening next year, or the year after, along with Musk and his promise that Tesla FSD/robotaxis are right around the corner.

Meanwhile, in the real AI world......


1745407677865.webp
 
There's a lot of 'critical mass' stuff coming up with the AI.... cough.... 'businesses' in the next twelve months or so.

This guy is worth keeping an eye on -
You do not have permission to view link Log in or register now.


OpenAI could go pop at this rate, the rate at which it's setting fire to money, with no path to profitability, is gobsmacking.

You do not have permission to view link Log in or register now.


View attachment 207746

If you don't fancy taking on a read that long, he has it in podcast form as well.



Sounds like whatever it costs they're determined to bring it in, but to make money from a chargeable service to a business, they in turn have to make a bigger profit to pay you, that seems like a long term sticking point.

I remember the dotcom years, and part of that success was firms willing to put money into hiring staff and doing things. There was a crash in the end but generally speaking the economy expanded.

Will AI expand the economy or put more people or businesses out of work to concentrate wealth further, e.g. a graphic artist that sells his skills for marketing, now the computer program can just generate images to compete with him. I suspect they're opening a Pandora's box here and not thinking through the repercussions.
 
Where there is no context in your search history or search query, it won't know which or what Tom Oliver you're looking for, unless it's something very prominent on its own, for example 'trump'.

Tom Oliver on its own can mean many things. But it would usually pull famous people if there are any.

tom-oliver-search.webp


For the query 'is Tom Oliver still alive,' first Google likely based its response on location, which is why it picked an Irish man. Since the query included the word 'alive,' (which relates to death or mortality) it connected this to factors that might cause death like torture, fighting, war, etc. And then it put its response together:


ai-overview.webp


And here i asked it directly to get an idea about where it got this info from:

tom-oliver-ie.webp


To avoid seeing that Generative AI nonsense, you can click on 'Web.' If it's not there, then click 'More' and it should be under there. It then just gives you only website links.

search.webp
 
This company Incogni they use AI to remove your Data from data broker, cost around £70 a year, only for US UK and Swiss, 49 brokers had my detail
Screenshot 2025-04-23 at 19.50.33.webp
 

Users who are viewing this thread

Accredited Casinos

Back
Top