AI Where is it headed

I was more worried about this until I dug a little deeper about some of the claims being made that turned out to be myths. For example -

Myth #1 - GPT-4 wrote GPT-5

Each model is created independantly of the other

Myth #2 - When GPT-4 wrote GPT-5, some of the human safeguards implemented in GPT-4 did not make it into GPT-5.

See myth #1.

Myth #3 - GPT-5 created a secret language to communicate with GPT-4

This apparently came about when they had a GPT-4 agent talking to a GPT-5 agent. The LLMs break down a request into chunks of text (tokens). Since this is how they understand text when communicating, they started doing so in tokens when communicating with each other. This is not some duper secret language that humans can't/don't understand.

While it is impressive at to what AI can do, it is a ways off from us being like pets to some master AI. That said, we did see an obvious change when schools no longer had students commit things like multiplication tables to memory. That's one of the reasons you see younger adults that need a cash register to do what we would consider very simple math that we can do in our heads.

Using AI in a way is like using Google Search. You still need to know how to frame the question, otherwise the old GIGO acronym applies. Same with AI.

Ai has already rendered humans as stupid morons, reasons why we now have to prove that we are humans to get into more and more websites , and it will be the death of all of us if we continue going the brainless way we are. Mark my words.

Paul Krugman described the current LLM models as "a souped-up version of autocorrect."

He's not wrong...

If current AI is what people are scared of, I'm pretty sure that Skynet is a LONG way off. The AI that Google and Microsoft use in their search engines looks like the product of a yapping retard Pomeranian purse-dog.

On a humorous note:

Recently one evening while listening to a major news network. The topic was brought up about the growing concern that AI has the ability to degrade critical thinking skills.

I saw hubby shake his head and then say smilingly, "Now they are concerned about the demise of critical thinking skills after legalizing marijuana in 24 states in the United States".

goodgollymissmolly,

That is why the quotes around "rewired"

It looks awfully like the "rewiring" of a human child's neural pathways as new and uncategorized data is presented.

We use Microsoft CoPilot throughout our company. I've used it as an engineer in building Azure microservices. In Visual Studio 2022, it appears on the right hand side of the UI by default and makes suggestions. I typically add a header to my Azure functions describing what the code does. The first time I did it, it suggested about 100 lines of code; which were all good enough to use.

it can make a competent person more productive. A newbie could get a similar result. However, without knowing the definition of "good", the person wouldn't be sure it met the business requirement. And this is just the source code. It's very common a microservice would receive a payload of data and then want to do something with it- like analyze the inputs, perhaps do some validation / transformation, and then add/update a database record with it. The developer needs to know how to do that part too. There is no magic wand that does it all end-to-end yet. And.. the biggest challenge is business people (read: customers) have every right to change their minds on what is needed and how it should work. Today we can't just talk into a microphone and tell the computer what to do. At least not yet.

"There is a growing body of evidence that use of AI degrades critical thinking skills."

That's what I was thinking.

I've become so accustomed to so many people handling cash being unable to figure out how to make change without the cash register or a calculator that I'm probably going to get used to even less critical thinking by people.

I'm not so concerned about what AI will do, but our reaction to it.

I think AI is being implemented before people know how to use it. I am in education, so i am most concerned about AI in relation to that.

There is a growing body of evidence that use of AI degrades critical thinking skills. I can't think of anything more important in today's world than to have critical thinking skills.

There us also thr issue of who is doing the programming. There was a documentary about facial recognition software. At the time, the software was unable to recognize any faces other than whiter persons faces.

Finally, there is thr issue of intellectual property. Which i won't go into here.

Apparently too many people thought The Terminator was fiction. So, Skynet is coming.

As far as AI rewiring itself, that's not really what happened, just like that's not what happens with small children. Instead, pattern recognition builds on itself as a natural process. It's how we learn to speak and read.

Where is it headed? Likely nowhere good for we normal everyday Everyman and women of the world.

I’m sure our masters will find it useful.

On a good note, kids and grandkids just did a big trip to Myrtle Beach, Wash DC, NYC, Cooperstown, Niagara Falls, Canada down back to Kentucky to Ga and then Fl.. ChatAI , gave them a written plan of what to see on 2 days in each city and cost effective including free exhibits etc. They showed it to me and it was pretty thorough including safe transportation while in each city.

But do NOT like all the AI with our banks, CC, customer service and tell them so.

AI scares me, as I've seen to many sci-fi film where it run amuck. I like to ask it to calculate prime number to a million decimals. That hopefully will keep me safe for a while, but I expect I'll get a call back eventually

I just finished an article. Cited was the observation that one, perhaps two of the AI models have successfully "rewired" their software to recognize patterns that they were not designed to "see", and to organize these patterns in ways not foreseen.

This involved the ability to use something close to intuition to build these constructs, and to use bits of information not directly within its design.

This is reminiscent of human children starting out with comfortable/not comfortable , and within a year being able to distinguish certain shapes as food or not. Then favorite food or not.( A slice of pizza and a triangle. After a try or two they will most often choose the pizza on the first try. The same for a circle and a cookie.) These are learned and not innate categories learned. Random bits of information are categorized ,to suit the need to organize.
The AI's are doing something similar with information surrounding what they are designed to manipulate.

It might be a good time to copy a concept from the Bible and introduce a "Tower of Babel" moment.