What is your Opinions on new AI Technology

It seems to me there is no intelligence, maybe like a hand held calculator having calculating abilities seems like intelligence? Artificial something.

 

How we learn about Ai has an effect on how we think of it.
When listening to any of the Pioneers, there is a strangely similar dialog.

 

The term artificial "intelligence" strikes me as being premature. We have come up with several systems that can make mostly correct decisions in a limited domain. They can do one thing when commanded, but they do it blindly, with all the intelligence and awareness of a bowling ball.

Part of the problem is that we are trying to copy over-complicated high level programs when we don't particularly understand what they are or how they work. We should be trying to map out and copy something like the brain of an earthworm - something with only a few hundred neurons, not millions. When we get to the point where we can do that well, we start looking for something with perhaps a thousand neurons, work our way up.

The tools we currently have are promising but not worthy of the word "intelligence".

 

Tools like ChatGPT can write, or so it seems for now, text that pretty much succeeds at hitting a certain style, but can be spotted easily enough by a human who reads it carefully and looks for semantic sense and soundness of argument.

So it doesn't hold up to what a well-educated adult human can write.

On the other hand, look how long and how much work it takes for a new human to reach that condition.

 

"Humanity has been using spark plugs for over 100 years now."

But that might be the first time a human ever typed that sentence. (And that's far from the worst on that page, yikes! How did you find this?)

It's just one more tweak of a reading skill, I guess. We sometimes noticed odd phrases in books or magazines that were originally written in human languages other than the one printed in our own copies.

Then we saw and came to recognize the patterns of automatically-translated webpages, and occasionally stores like this one:



Now it's a bit more subtle with the machine writing and singing and all. But I'd bet my sixth finger that we'll be able to catch these things out for a while yet.

 

All of us are on our own journeys into the future on our own paths and in our own environments, while sharing many of the fundamental similarities that make us humans.
Still,
there are so many things within,
within our bodies and minds that we don't have the slightest clue about.
Same for so many things happening without or outside each of us.
They call that stuff, life or the human condition(s).

Or,
there's stuff happening around us, within us and through us 24/7/365 we can't detect or understand, even today, and possibly never will be capable of understanding. Oh well ! Ever wonder what some of those things might be?

Here are a few more acronyms for those who are on their own paths of learning what AI is and it's history.
large language models LLM artificial general intelligence AGI

I don't have a clue how to produce a LLM or even what an AGI might look like, both a good one and a bad one.

I have access to the tools to build one, but mine would be slow and probably a bad one, besides I have my hands full digging through 250,000 lines of html from a 450 mile Hybrid Assistant Reporter Trip I recorded a month ago. Which console and GUI tools to use to break it up into shareable chunks.
( Damn the scatter graph has got to have somewhere in the realm of 50,000 lines by itself, it's the biggest ).

I don't want to post a youtube link here to any specific video that I've consumed regarding AI, but I will share this much.
Mustafa Seleyman !yt by way of duckduckgo.com / ( ddg.gg ) if your browser supports it.

You can find other youtubes and other research personnel that might fit your personality better while looking at the results from a single search pattern.

 

As hilariously bad as so much of that "article" is, there are also some pretty impressive things about it.

Out of all the terabytes of internet data some LLM had hoovered up, it included "induction coil" as a related topic, and included the figure of 20,000 to 25,000 volts, which is about right.

It mangled the description of the four-stroke cycle, but it clearly did pick out sources of that description and deem them relevant, and synthesized a (goofy) description out of what it found.

While it didn't do a great job, it might be interesting to talk to some middle-school teachers about assignments where they ask to review a bunch of information sources, pick out what's relevant, and synthesize something from it, and how many of their students do as well.

 

freaky

 

If you want to plot a course for the 22nd Century, then you need to look back - at least as far as the 19th - lest you find yourself veering off course and going in circles.

The Luddites were members of a 19th-century movement of English textile workers who opposed the use of certain types of cost-saving / wage stealing machinery, and often destroyed the machines in clandestine raids....
Luddite - Wikipedia
I myself would almost certainly be replaced by AI were it not for the fact that they haven't developed a cost effective robot that can clean fibres, remove and replace circuit packs, and drive a car.
I also have the advantage of being nearly old enough glide in for a landing on Runway Retirement, right.

Either way I find it comforting that AI will allow people who are too old or too mentally enfeebled to 'retire in place' longer and with more mobility and more dignity.
Well...unless you're a certain politician.

Medical research will be more efficient once they control for GIGO and groupthink.
So will research in all of the hard sciences.
Durable goods will be cheaper and last longer.
EFFICIENCY not only means more economical but also more ethical and more environmentally sustainable.

We're BETTER OFF now than we were back in the 'good old days.'
We're GOING TO BE better off in the future, probably with a few bumps along the way.

Period.
Full stop.

 

What gives me pause is that it will be, for the near future anyway, humans deciding whether the AI capabilities get used more for those beneficial ends, or more for other ends such as more effectively deceiving and scamming other humans. Perhaps you've rolled that possibility into "a few bumps along the way".

In one scenario, the capabilities for deception and persuasion could even contribute to the election of governments that outright devalue those kinds of beneficial ends, in favor of prolonging the enrichment of cronies. A bump like that could last a pretty long time, if human generations are used for the scale.

 

The AI so far is a bunch of hot garbage for most the applications it has been misused for. It ruined the internet. Google is broken, just AI generated drivel everywhere. Forums like this are the last bastions of fact, assuming we don't end up with swarms of posters are just AI registering for forums and posting drivel, which will probably be the next step in this "progress".

AI cannot give factual answers on a reliable basis. It hallucinates and presents false things as truth because someone said it somewhere.

It's more or less useful to learn to do something like code, compared to stackoverflow, and it is useful to generate simple images quickly that don't need to be super perfect, like my avatar image. It's cool for generating mildly entertaining toys. g-code for 3d printed or woven doddles.

However, it is utterly rubbish for writing articles or for use as a search engine, at this point in time.

Facial recognition advancements on the other hand are immensely useful, and it shows process for mind reading technologies, despite 1984 vibes. The ability of the language learning modules to rapidly translate ancient texts on rolled scrolls is absolutely phenomenal.

AI is a tool, and the key is to find the applications it is good for. Reading ancient scrolls, wonderful. Replacing articles written by auto mechanics on how to fix your car? Rubbish. The thing is, at least at its current state, the AI doesn't actually understand what a car is or how to fix it. It just knows what people wrote and patterns. It can repeat stuff people said but not understand how the parts go together nor tell fact from fiction.

 

Ai in Full Self Driving Supervised (FSD) is brilliant. FSD V12.x has driven about 3,000 miles including Atlanta traffic and done brilliantly.

As for the text AI, they have a real problem with plagiarism. Something I've suffered once on the job. But it can be solved by teaching it how to footnote which even Wikipedia knows how to do. In this narrow case, we agree.

Bob Wilson

 

Self driving cars have a way to go.

Might be a great asset for a blind person or a drunk person in good weather, but these days they still stop for signs printed on t shirts and I wouldn't trust one on a snow covered road.

 

Mine has already gone over 3,000 miles. But if you'll wear the "t shirt," we'll have FSD will do the test. Of course we'll also have to compare it against you wearing a blank "t shirt."

The only false "STOP" sign events are when high winds have twisted the sign to appear full on to the car. That is exactly what it should do for a correctly placed sgn. BTW, it is easily overridden. It does not do a 'California stop.'

Bob Wilson