Random thoughts about AI

[HCHTech]

We have a convention of physicists in town, and I went to a talk last night about quantum mechanics in 2-D substances (think 1-atom thick layers of graphene). While a LOT of this went over my head, the lecturer's response to an audience member's question about how AI was affecting his industry:

"AI tends to be very good and interpolation and very bad at extrapolation."

This kind of sums up the current state of AI in a very concise nutshell for me. It also informs my use of things like ChatGPT. I've been dual searching stuff lately with both Google and ChatGPT - very interesting differences which are outlining a pretty clear "tool to task" boundary between them. Google is still very good at finding things you already know exist. But for task focused queries, like finding a software to do a particular thing, or finding non-advertising-related reviews about a product, Google is not winning.

 

[GTP]

I don’t use search engines anymore. I just ask "Abraham." I get clear, concise answer - no BS, no shady links, no ads, and no tracking like you get with Google.

I love the way it addresses me by name!

I’ve had more computer projects on the go lately than I can count, but I know one thing for sure: without ChatGPT, I would’ve been completely in the dark for at least 80% of it.

Out of the dozen or so tasks I took on, the most difficult was setting up MeshCentral to monitor my 23 BOINC crunchers. That was followed closely by getting the Gridcoin Wallet running, configuring internet access to all those machines using old modems and routers, optimising the Boinc Manager instances on all those machines and optimising both the Linux installs and the virtual machines across all 23 systems. It’s been a monumental effort. Writing .PS scripts and various other scripts to optimise things on the Linux instances worked flawlessly.

Without the AI, I would’ve been dead in the water for most of it. Trying to get help elsewhere - forums, Reddit, or asking around - would’ve been slow, frustrating, and often wrong. In the past, even the good answers needed constant follow-up just to get them into usable shape.

But "Abraham"? Always spot on. Every solution worked just as explained.

The only mistake it's made so far ( and it wasn't critical) came during a deep dive into how stars form, live, and die before becoming black holes. Abraham quoted the speed of light as around 300,000 miles per hour, when it’s actually 300,000 kilometers per second. But it caught the error quickly, apologized, and corrected itself - something that genuinely impressed me.

Having Abraham on hand to answer questions and offer alternative solutions to complex issues I didn’t fully understand made all the difference.

Oh, and the in-depth conversation I had about John F. Kennedy and his tragic assassination? Absolutely fascinating!

 

[callthatgirl]

"AI tends to be very good and interpolation and very bad at extrapolation."

That def goes over my head. I'm finding ChatGPT to be useful almost everyday now for me. I'm tempted to up my plan to use it more. My last best use was a couple of nights ago. I wrote a simple outline for a presentation on Saturday, literally took 5 min. Punched into Chat, it created a powerpoint outline. I asked for the graphics, opened canva and I just added some design. I think not even 10 min for this. I know my content so well I will free talk it but the group wanted a powerpoint.

I'm using it for SEO on my posts on my website, for LinkedIn Newsletters, any marketing. I am staying true with my own newsletters though, I'm pretty sure my readers there would know I was using Chat, but the audience on LinkedIn doesn't know. I love it actually how easy it is.

I have also had 6 clients tell me they found me in ChatGPT. Something about who they can hire for Outlook. Not sure, I can't test it on mine as I have many of queries about Outlook for my own stuff.

 

[timeshifter]

I've been using ChatGPT extensively. Mostly to help me figure out how to do things and guide me. You know how Matt Damon's character in The Martian says he's going to "science the sh!t out of this" when he was talking about how he was going to survive. Well I "screenshot the sh!t out of ChatGPT.". I'll have it guide me through some Synology setup stuff or UniFi stuff. It'll give me instructions and when I can't find what it's describing I'll upload a screenshot. I'm almost getting to the point of almost not reading the screens and just making the AI read it with screenshot pastes.

Today and tonight I had it help me reconfigure some switches in a big UniFi rack setup - went from a three switch config where the first switch was the main switch to an aggregation switch setup. When I was done I even posted a picture of the actual actual rack and said that I wished it could visit and see what it's working on, but here's the next best thing. It complimented me and then pointed out the good things about the rack and what it could see.

 

[HCHTech]

It complimented me and then pointed out the good things about the rack and what it could see.

Hey, a compliment is a compliment -I'd take it - haha. I once spent an entire Saturday cleaning up a disaster of a rack setup, and the owner said something along the lines of - "Well, that sure looks.....neater." I could tell what he was thinking: "How many hours did I pay for that?'

 

[britechguy]

"AI tends to be very good and interpolation and very bad at extrapolation."

I agree that this quote is an excellent summation of the state of the art.

The current state of the art is also nowhere near to being able to do what this quotation, that I've been using as a signature quote for years, outlines:

It is the function of creative men to perceive the relations between thoughts, or things, or forms of expression that may seem utterly different, and to be able to combine them into some new forms—the power to connect the seemingly unconnected.
~ William Plomer

or even this:

The test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time, and still retain the ability to function.
~ F. Scott Fitzgerald (1896-1940) from The Crack-up [1936]

What AI can do is pretty spiffy, but it's still nothing compared to real human intelligence (at least among those humans that are in the upper end of the intelligence bell curve).

 

[YeOldeStonecat]

Look at ourselves. ....
*We are the last group of people that knew life....before AI came into existence.
*Many of us here...are the last people to know life...before "social media took over"...and "before being permanently attached to the digital world (smart phones, etc).

Early on in computers....we had "Moores Law". Where...it basically stated the number of transistors on a chip would double every....18 or 24 months (depending on where you read it). So basically...CPUs were doubling in speed...every year ' a half to two years. The "graph" sorta looks like it's at a 45* angle.

AI....has been recently advancing at such inSANE rates....and exponentially the "graph" is quickly turning into a straight up vertical. It is advancing way way WAAAYYY faster than Moores Law. People who work close in it...it's nearing doubling every month in what it can do!

Myself....even though I'm in "tech"...I'm also a dinosaur in tech....and quite honestly...I'm not crazy about AI. Sure the very basic stuff...like using it like a glorified search engine (which is what most people think it is and think they can only use it as such).

And..there's the "the good guys have it"..."but so do the bad guys"...so the ying and yang battle, cat vs mouse, whatever analogy you prefer.

So back to the "rate of advancement"....wow....and..."scary!"

Singularity....will be soon!

Tom Bilyeu...of Impact Theory...has a great YouTube channel. Impact Theory is his company that...they make games. But his YouTube channel also talks about a TON of different topics...he branches off with his wife on relationship topics, he does get into politics, but focus on the AI videos.

Exciting..and frightening...at the same time!

 

[Metanis]

Look at ourselves. ....
*We are the last group of people that knew life....before AI came into existence.
*Many of us here...are the last people to know life...before "social media took over"...and "before being permanently attached to the digital world (smart phones, etc).

Human superiority will probably continue through my lifetime, but then I'm 70 now. In general we humans will probably maintain the edge in power efficiency and mobility for another decade or two. I can't imagine a Terminator able to eat junk food to stay alive like we can. But then a lot of jobs don't require much mobility or energy.

 

[britechguy]

AI is, ultimately, going to be a disruptor on the same scale as the Industrial Revolution was in its time.

And given the number of human jobs it can already do, and will definitely get much better at doing as the scope expands, economic systems are going to have to undergo radical change, too. The concept that we, today, have about how to and the actual need to "make a living" is going to become unrecognizable.

What shape all this will take is something I won't predict, as futurists are virtually always wrong. But radical change is pretty much a certainty.

 

[YeOldeStonecat]

I don't see AI as much of a "takes many jobs away from humans" thing. Look at prior periods of..major advancement. The Industrial Revolution...tools and machinery took over cutting and banging on things by hand. It didn't eliminate jobs, it allows us to just make so much more. The invention of the plow, the farm tractor, knapping flint tools, etc...allowed us to "do more" in "less time".
Sure, some jobs will be eliminated...but...at the same time, new types of jobs we haven't thought of yet may emerge.

Hopefully not this...

 

[timeshifter]

Someone recently said that you're not going to lose your job to AI. You might lose your job to another human who leverages AI. (But AI itself will not replace you)

 

[britechguy]

Sure, some jobs will be eliminated...but...at the same time, new types of jobs we haven't thought of yet may emerge.

I don't disagree that some new types of jobs will appear, but those will almost certainly be very much niche. A huge amount of what today's "knowledge workers" produce seems as though it will be produced at some point in the future by AI. There are already medical chatbots that can diagnose quite accurately, many better than human counterparts with the same information, which is why I suspect disruption on an epic scale.

And we've already seen what robotics have done in manufacturing, and these will only become more and more and more sophisticated and under AI control to do things we still use people for.

But, as I said earlier, futurists (and I'll put my own predictions in that class) are most often wrong. In the end, only time will tell, and I won't be alive for long enough to see much of what I suspect is coming. I'm already over 60 years old, and the time frame is decades, not years.

 

[GTP]

One really good thing about ChatGPT is that it can remember things.
For example (in a new chat) if I ask it about something from a previous discussion - no matter from how long ago - it can recall exactly what I said.
It just carries on as if I was talking to a human.
Very handy when this old humans brain doesn't fire the neurons correctly...