Not remotely impressed with AI

I am hoping AI can fix Amazons search function.  I search Amazon for head lights for a 2019 Ford Escape.  I see a bunch of results, click on one of them, and it for a 2020 Ram 1500.  Select another one and it is for a XYZ.  The top says it will not fit your XYZ click here to see items that will fit.  

I have a Samsung Bespoke fridge.  It is having issues sometimes.  The fridge and freezer stop cooling.  I unplug it then plug it back in and it starts to work again, no error codes.  I search the model number for this issue, at the top of the search page it gives an AI generated response saying you need to keep it 2/3 res full for it to work it's best.  The next top selections all say about the same thing.   The rest are for another model, not even close.

AI is only as smart as the programer in my opinion, or for people not bright enough to come in out of the rain until someone tells them to.

Mike, I agree with hoping that many searches can be fixed to return what I want to see, not what they want me to see. On the other hand, I've found ChatGPT to be very handy when I'm looking for info that's published somewhere, but I can't find it, or I am too lazy to take the time. The other day I wanted to know the pros and cons of MRI's verses Cat Scans for personal reasons. Chat GPT provided a very concise explanation that was very helpful.

I'm trying an app, perplexity, that gives you linked footnotes to the sources it used for its answer. So far, it seems to work: the answers are complete and appropriate for the searches I've entered, including subjects I'm knowledgable about.

We ditched cable a few years ago in favor of fiber, so my TV-watching has mostly settled on watching YouTube videos.  I am coming across more and more AI-heavy videos that are just a conglomeration of B-roll footage and/or CGI, with a narrative that is much like what the OP describes—circular, repetitive, shallow, and often there are glaring mispronunciations that no human would make.

Jason

Peter Martin wrote:

@John in NM,

I see where you're coming from—AI-generated content often lacks depth, especially when it's optimized for engagement rather than substance. But AI isn't inherently bad; it's how it's used. For example, AI can assist with structuring complex information, summarizing dense research, or even debugging code. The real issue seems to be the commercialization of AI-generated content, where quantity is prioritized over quality. Would you find AI more useful if it were applied differently, perhaps as a tool to enhance human expertise rather than replace it?

Yeah, that's fair enough.  Easy to overlook the strengths after sorting through some of the sales copy bs that populates 90% of search results.

Thinking about it a little, any mind numbing and tedious comparative task can very well be handled by AI - QA on manufactured parts, photo evaluation for things like near earth object detection, etc.  Sorting through huge volumes of data looking for anomalies, AI doesn't get bored, doesn't get distracted, mind doesn't wander.

What it can't seem to do yet - I hope anyway - is creative invention.  I'm not sure we know where that comes from in human intelligence either, pretty certain we never will.  If AI can ever crack an original joke, we may be in trouble!

Yeah, a perfect case in point Jim!

After more thought, what I'm remarking on is really human nature more than a particular AI failure.  AI is very good at large volume tasks, repetitive comparison tasks, etc.  What AI cannot do is creative tasks - that is why it's writing tends to be so bad, and why there are often laughable errors in the results.  The human element is to take any one thing where you get excellent results and attempt to apply it as broadly as possible without regard to concern over the quality of results - AI is good at these handful of things, let's use it to do something totally different!

Humans have been making that mistake for eons.  We will likely soon be treated to an AI bubble, just as we had dot com bubbles, mortgage backed securities bubbles, telcom bubbles, etc.

I think anybody who thinks AI is not going to get better and take over lots more tasks is in the bubble, not AI.

Bill Howatt wrote:

I think anybody who thinks AI is not going to get better and take over lots more tasks is in the bubble, not AI.

I'm sorry, I was not clear in my writing.  By using the term "bubble" and examples of similar bubbles from recent history, I meant a financial market bubble, not the sort of cultural bubble that you may have taken to be my meaning.  Looking at what I wrote, my meaning was implied rather than spelled out - in the interests of brevity needed for casual discussion like this.

Sure, AI is going to get better.  It will certainly get better at what it already does well, and it may get better at what it currently does very poorly and most human beings do well - creativity.  Focusing on what AI does well will be an ongoing boon to humans.  If AI gets better at creativity, that may be a whole new boon to humans, or it may be the end of us.  We won't know until we get there.

I can only say, in retrospect, that I mis-titled this thread.  The thing with which I am not remotely impressed is really the popular cultural excitement with AI rather than AI itself.  We are short sightedly suckering ourselves and each other with an irrational excitement over technology and developments which we do not understand well.  The most obvious and dangerous result of this excitement is going to be a market bubble in which many people will lose their investments, and the economy (and thus also those of us who never made ill considered investments) will suffer some setbacks.  One need only remember the many previous market bubbles since the 1980's to see this coming.

So in a nutshell, I'm not complaining about AI.  I'm complaining about the foolish, trendy hype to which homo Sapiens have long been fascinated.  We do not learn from history and we repeat it, over and over again.  The only thing dumber than following the latest foolish trend is to try to ban the latest foolish trend.  We never seem to learn that one either.

Added later 14 min:

Peter Martin wrote:

I don't know if it's going to get away from us and try to kill us or turn us all into paperclips, but it is going to be very disruptive, weaponized as all technology is, and replace lots of jobs. Interestingly, the jobs it replaces won't be low-skilled jobs, but ones that traditionally required lots of education and experience. 

I have been using it to program and am gobsmacked at how it can do what it does. I think within 12 months it will be doing all coding that is now done by new and mid-level programmers. The once lucrative and secure job of being an excellent coder is about to just go away.

Now that got me thinking.... to what degree are we fooling ourselves that AI isn't replacing low skilled jobs?  We don't think of coding as low skill, but that could be due to it being such a recent career.  The old people in the field were in college when I was born.  That is a very new career.

Think about it though.  What is the creative part of software development - the algorithm.  What is the tedious drudgery?  Getting your code to be logically consistent so that it produces no errors. 

Now I haven't done any coding since the days just after C was a new big deal language, and scripting languages were so new that I didn't know about them yet - so I'm no expert in this field.  That said, it seems to me that AI is likely taking over and excelling at the part of coding that we humans hate to do and try to avoid doing, the tedious drudgery.

I also may have a very skewed perspective on this, because much of the software I am exposed to at my day job was developed in house.  There has long been a temptation with such development to focus on getting it running well enough to move on - say 90% level - than to focus on perfecting it.  Thus we have stuff where you have to restart it every 3 days because of memory leaks.  One of the kids developing software actually chuckled when I suggested they make fixing crap like that a higher priority.  AI could fix dumb stuff like a memory leak pretty darn quick!

I'm an old guy in the 8th decade. When I started work one of my jobs was a bit of scientific programming in FORTRAN. When I told people that I programmed computers they would declare me a genius and fall at my feet. By the time I retired, they were saying they have an 8 year-old kid that does that and knows all about computers!

Bill Howatt wrote:

I'm an old guy in the 8th decade. When I started work one of my jobs was a bit of scientific programming in FORTRAN. When I told people that I programmed computers they would declare me a genius and fall at my feet. By the time I retired, they were saying they have an 8 year-old kid that does that and knows all about computers!

What you're describing with the 8 year old kid is kind of what I'm complaining about - it's the popular culture thinking, repeating what sounds good and everyone else is saying, without thinking about it.

Long story short, I prefer hiring an older programmer - someone who learned their business when system resources were so limited that part of the job was making code efficient.  The kids have gotten lazy in my experience, they never learned to put their toys away when they were done with them.  It shows in the performance of their programs.

As AI gets smarter, we will become dumber.  At some point AI results will not be questioned.  It will become the Oracle of Delphi and we blindly worship it.

I read an interesting column just now, about AI (or more accurately, Large Language Models) and their use in faking writing assignments, whether professional ones or in education.  The author mostly focused on education because that is the easier outcome to predict.  First off, it is important to understand that AI is not smart - it can not produce anything creative.  It can assemble syntax more effectively than homo sapiens because it has an infallible memory and does not get frustrated with tedium.  Perfect for that 10th grade term paper that the teacher will just skim through anyway an give a B+.

What this author didn't quite predict in so many words, but I think will happen, is that as students use AI to fake their way through school, there will be a renewed premium put on genuine creative ability by employers.  The smarter students will recognize this and will take advantage of it.  If we can just avoid giving all the dunces a UBI when they are out of school, then this could actually be beneficial to incentivizing creative thought.  I doubt we will be that tough though.  Not the culture that gives C- work a B+!

Easy, north pole.  That may well be Musk's favorite interview question, but it's older than he is - I think I first heard that in junior high school.

The one I liked from him had to do with not wasting everyone's time in over populated meetings.  I see a lot of "meetings = productivity" fallacy in the day job.  Far worse since the pandemic and zoom becoming a thing.

I also interviewed with SpaceX a few years ago and was underwhelmed.  I liked their structure of several short interviews with individual people, but I didn't think the position they were looking to fill was a wise idea.  They wanted an engineer to design a methodology that I have seen fail on a consistent basis, all based on the misapplication of lubrication theory developed for low speed/high load bearings in modern huge windmills.  Maybe they can make it work, but I tend to think they were pursuing a fools errand - spending tons of dollars to avoid spending a few of them.  Unimpressive to us old people.  I actually laughed at the interviewing engineer, that probably sealed the fate in which I had already lost interest.  Maybe the kid will remember that when he hits 40 and realizes the folly of letting accountants make engineering decisions (and vice versa too).

You're right about AI text being far too wordy while saying either precisely nothing or the same thing over and over again.  To me it sticks out like a sore thumb.  First sentence might fool me, but 2nd or 3rd it's clear that a human brain didn't come up with it.

The annoying one to me is the YouTube videos that use AI-generated narration.  It’s about like the sentence repetition already mentioned.  It’s interesting at first, but then I’ll pick up on the repetitive statements, and the pacing of the narration is a little unnatural as well, along with a butchering of the abbreviations for units of measure.

Jason