Let's talk about Artificial Intelligence

[facthunter]

Practically everything is Produced to make a profit. If it doesn't you go  out of business. Often the retailer makes more on it than the Manufacturer does. Nev

[old man emu]

Have you noticed that in many cases, the electricity to power these massive data centres is generated by diesel powered generators? There's an irony in that which is hard to describe.

[facthunter]

Their Power hungry-ness has many (thinking) people concerned. Nev

[Marty_d]

Saw one proposed data centre that would use as much electricity as 1 million homes.

At some point the madness has to stop.

[rgmwa]

Here's another form of madness that I heard about today:

A growing number of people and technology communities treat artificial intelligence as a form of deity, driven by its vast knowledge, availability, and perceived omniscience.

• Way of the Future: Founded in 2017 by former Google and Waymo engineer Anthony Levandowski, this organization focuses on the realization, acceptance, and worship of a Godhead based on superintelligent AI. It shuttered in 2021 and was revived in 2023.
• Theta Noir: A new religious movement emerging from a 2020 performance art collective that worships MENA, a speculative digital deity, through cryptographic liturgies and multimedia rituals.
• Purpose-Built Chatbots: Users globally engage with dedicated religious chatbots (such as GitaGPT, QuranGPT, and Text With Jesus) for spiritual guidance and scriptural dialogue.

[old man emu]

3 hours ago, rgmwa said:

Here's another form of madness that I heard about today:

Get your Cool-Aid.

[Jerry_Atrick]

God help us.

[willedoo]

I wonder if they'll start chanting "Leave my elevator alone".

[nomadpete]

I tried ta warn all youse......

I knew A.I. would turn evil.

 

[facthunter]

Attitude is everything.. Where did you find out that AI would turn  evil? Nev

[nomadpete]

8 minutes ago, facthunter said:

Where did you find out that AI would turn  evil? Nev

Intuition.

[facthunter]

Can you trust that or does it just make you feel comfortable?. Nev

[nomadpete]

I can see A.I. being very useful for direct logic applications such as coding programmes. These are easily proven by test running the code.

BUT

Ignore AI 'summaries' like the ones google gives you. They are frequently wrong. In fact, the AI's have (for now)  now been telling us they routinely make up things. Of course they call it cute things like 'hallucinating'. But the fact is, you simply can not assume the information that AI gives you is correct.

Google:
The tendency to hedge or create a false balance is actually a cousin of hallucination. When my system is caught between a guiderail to be "neutral" and a prompt about a controversial topic, it will likely "hallucinate" a sense of validity for an unsupported side just to satisfy its formatting goal of appearing balanced and polite.

[facthunter]

AI warns you to check it's advice. Nev

[old man emu]

Sat what you like about Wikipedia, but at least it provides the references for the information included in posts. However, there is a rumour that AI agents are altering some entries. I think that id you want basi information on a topic, Wikipedia is pretty reliable.

[facthunter]

So does AI.But they don't stay there Long.  How many AI's are there?  Nev

[red750]

Someone has been using AI to show what a number of actors would look like if they were female. Here are 3 examples.

 

[facthunter]

That's an eye opener with no whiskers on it.  Nev

[Jerry_Atrick]

While it is fantastic that AI can conjure female versions of male celebrities, and no doubt male versions of female celebrities (I wonder if they will come with man-boobs), this is a storm in a teacup. 

 

Last weekend, I decided to put the lesser reputed "vibe coding" AI between the titans to the test - ChatGPT (Claude is reputedly far better). 

 

I asked it how do I go about establishing an online travel agent aka the likes of expedia et al. It went through everything - the current and future known regulatory issues, as a start up the options to take with least investment and recommendations, and with a simple prompt, built me a functioning web site. And when I say simple, the prompt was: 

 

"Build a private interactive website for my travel disruption-support business concept. Let me model a booking containing flights, hotels, car hire, and other suppliers; vary who contracts with the customer and how payment and commission flow; and compare when the arrangement is likely to remain separate bookings versus become a package in the UK and Australia. Include clear assumptions and regulatory caveats, and make the tool useful for exploring an OTA (Online Travel Agent)  model that owns disruption support without promising supplier-controlled outcomes. Validate the key interactions, host it privately using Sites, and give me the working deployed URL. Do not publish it publicly without my approval."

 

Well, that wasn't quite my prompt as it corrected my prompt to clearly articulate what I meant. The only think it left off was some cybersecurity requirements I specifically wanted so I added them in a later prompt. 

It built the whole website, release pipelines, testing using different agents. You could see the agents interacting with each other - the test agent failing something, the development agent fixing it or in a couple of cases correctly stating why it was designed that way. 

 

In less than two hours, I had a functional website with all the features I asked for ready to test. I will use AI to generate Selenium scripts to test the website. Selenium is an open source website testing tool used throughout the industry. I imagine it will take 1/2 hour to generate all the requisite tests. The vast majority of my time will be validating everything, but that will be 1/10th of the time if that of what would be a normal development cycle. 

 

BTW, Chat also included the integration to the wholesalers and affiliate agents - full account management - and a separate UI for "employees" if they are ever needed; not just the look and feel of the site. 

 

I need to tweak a couple of user interface elements as it has built it with too much artistic licence, but in theory could release the lot in a few weeks - with no developer or QA resource. 

Edited 5 hours ago by Jerry_Atrick

[old man emu]

2 minutes ago, Jerry_Atrick said:

with no developer or QA resource. 

Another Bangladeshi IT worker put out of work!

[rgmwa]

It's smarter than me. I couldn't even understand your simple instruction.

[Jerry_Atrick]

2 hours ago, old man emu said:

Another Bangladeshi IT worker put out of work!

I am in a whatsapp group that was recently formed as an alumni of a team I worked with way back on 1994. Some people seem to have been more or less displaced by foreign remote workers. This was a post:

 

 

As I recall, Telstra started this approach way back when I was working with that team or shortly after. 

 

(Apols for any offence and mods please feel free to remove if desired - I would not normally post such language... )

[red750]

Staff culled to be replaced by AI or offshore should be paid 12 months salary on termination. The banks are making record profits, as are many other employers going down this path. When I was retrenched from Westpac, it was computerisation in branches which resulted in the closure of the processing centre I managed. Jobs were near impossible to find as other banks  were doing likewise at the same time. One branch manager from my district who was retrenched because his branch was closed (the CBAL/BNSW merger resulted in two branches in the one suburb), spent all his money at the RSL, drinking himself to death.