This one's for Alex

Is the face swap tool part of the ChatGPT suite? Do you want to create a short video on how it's done? I've got a similar project where I want to "Release the Cracken", and it's my son with his constant chatter that is the monster.

No, it's a different website called Pixlr.com that I use for the face swap. It's got the best face swap technology that I've come across and I've tried a whole bunch of them.

I would say though as far as the regular AI sites go, ChatGPT is pretty good at making pictures, but it doesn't keep the face the same. So, I can tell it, "Hey, make me a picture that looks like this with the Access logo in the background and a red tie," and it will make a picture with a dude that kind of looks like me, but it's not good at keeping my face the same. So, then I just take the picture that it generates and I move it over to Pixlr and then I do a face swap. And it works really well.

I will say Gemini is really good at maintaining the face. Like, if I upload a picture of myself and say, "Change the tie to blue," it can do that and keep my face the same. But it's not great at everything else like putting a logo on the screen in the background. It had a hard time with that. So, every AI has got its own quirks. I've found that ChatGPT and pixelar is the best for doing face swap stuff.

Yeah, maybe I'll make a little video on it. Pixlr is not free, though. But it's not expensive. It's really cheap. You get like a thousand tokens for like seven dollars or something stupid like that. And each face swap you do is like two tokens, so it's not expensive.

I'll throw this here... The AI bubble may be about to pop. Nobel Prize-winning economist Robert Shiller just predicted that the AI bubble will bust. Also, Michael Burry (the guy who did "the big short" on the sub-prime lenders) just revealed he has shorted Nvidia and Palantir for over $1B (yes that's a B)... The Asian markets have tech stocks falling tonight. We'll have a good idea soon... I sincerely hope everyone here has protected their assets.

Michael If it were to burst, it probably wouldn't be as bad as the dot com bubble burst in the early 2000s.  This AI hype is not entirely on hype alone.  It has already had tangible and positive results, as evidenced by the fact that we all are probably using it daily every time we do a Google search.  I've been using it to learn Visual Studio (which also has AI features built-in) in the past 2 months.  I've been asking Google non-trivial tech questions such as "How to change column header background color in datagridview in VB.Net."  And it has given me pertinent, accurate, concise, and user-friendly answers about 70-80% of the time.  That may seem like a bad percentage, and it's actually a good percentage since this is equivalent to asking a random person on the street, except that a random person doesn't give you pertinent answers 70-80% of the time.  In the early days of ChatGPT (~3 years ago), AI would give me nonsensical answers on coding.  I had to warn the people on this forum not to take AI's answers seriously.  But now I sing a different tune.

Michael Also, AI is not exactly an overnight phenomenon.  It has been in use to a certain extent for decades, mostly in tech fields.  It gained notoriety mainly because it started to affect arts fields.  Writers, painters, musicians, artists, etc., who for decades had thought they were immune to automation, collectively pooped their pants when AI suddenly could write term papers, legal documents, draw arts, write music, make videos, etc.

Kevin Google search has been FUBAR since it incorporated AI. I would argue that presenting the wrong information with authority is far worse than a search from five years ago. I haven’t seen any formal studies, but most DIY folks I know won’t trust AI searches.

The legal briefs created by AI frequently quote nonexistent cases. Several firms have already suffered the consequences there.

I’m not against tech. Like many of you, I grew up watching the promise of tech through Star Trek. Please listen to Better Offline’s recent episode on vibe coding to hear experts discuss how AI is not helping coders like the MSM would have you believe.

IMHO, you should use a real picture. Your videos are better than everyone else's because you speak in a real honest manner. It seems like a real honest photo should go with that. I am personally tired of all the fake or filtered images on everyone's FB profiles.

Michael  Regular Google searches have the same problem with accuracy too: wikis, forums, news reports, etc.  There is hardly anything online you can accept at 100% face value.  There are always going to be errors -- everywhere, online and offline -- unless you look at original verified sources.  Whether you look at AI or non-AI info, you still have to check and verify, like you would do with every info.  The info from AI is like any other info: it adds value to your work when it's useful, and you can discard it when it's not.  As I said, it has had the tangible effect of saving me time when I write program code, even with only 70-80% accuracy.

To Kevin's point, the adoption of AI by Gen Z is huge.  They have embraced it like we embraced google 20 years ago.

Matt  Age may affect adoption, but the bottom line is ultimately what decides its truth worth.  Even with "only" 70-80% accuracy that I've observed when it comes to coding, it still creates a positive net effect to my bottom line, which is the time it takes me to find the right answer.  Am I going to turn down a right answer because it comes from the "wrong" place (or just someone else's idea of a wrong place)?  No I'm not.

Moving from anecdotal to science-based findings:

https://devclass.com/2025/02/20/ai-is-eroding-code-quality-states-new-in-depth-report/

BTW, anyone see how NVIDIA stock is doing today? Or Nadella's comment on lack of power for AI?

Kevin Google is *intentionally* inaccurate with its searches. It has moved almost entirely to search results through which it can profit-- sponsored, popular links with ads. Five years ago, you could search my name and get back a page of pretty much me (due to my exceedingly rare last name). Now, you might find me; mostly you get Mike Holmgren (Google thinking it knows better than you as to what you're looking for AND driving you to sports sites that pay them) and other unrelated material.

You can search the unique *HTML title* from one of my web sites and it literally will not appear in the results. Try it yourself: KAKT: Camelot (prologue) - Video Games for Old Guys. I think I'm pretty safe in saying it's because Google doesn't make any money there. Hopefully this clarifies what I meant by Google searches being FUBAR.

Michael   Does that "science-based" report take into account the qualifications of the coders?  AI is only a tool.  At the hands of poorly qualified individuals, no tool would help or make worse.  Instead of looking at "code quality," we should look at the bottom line: the quality of the end product and the time it takes to make it.  

I have my own "science-based" findings too: my tangible time reduction in writing and debugging code in Visual Studio.  If I had this in my old job, nobody in my company would care what "science-based" reports said.  Our own concern for the company's bottom line would trump everything.

Sometimes, both sides can be right.  No one is absolutely right or wrong.  There is "gradation" between the opposite sides, and there is "context."   If you can use AI to your advantage, what does it matter?

Michael  Google has made tons of bad products, as we all know.  For this forum and and discussion, I'm only looking at and evaluating its AI answers on coding.  If we bring up all its questionable practices and services, there is no end to it.  No company or product is perfect, and it's up to us to take the good (like YouTube) and reject the bad (like Stadia).

Kevin I'm guessing you didn't read the article. They looked at *211 million lines of code* and derived multiple findings. This is science. The personal experience of one person is what in medicine is called a case report. It's an interesting finding that only has value in suggesting something that might benefit from further study. It is not a finding that should drive behavior. If my nephew was vaccinated and was later found to be autistic, that is not a finding that should drive behavior. Scale matters.

If AI helps you code, great. A careful coder, willing to heavily edit the pieces AI generates, might gain some benefit. However, when looked at scale, this study and others find that AI repeats functions and doesn't refactor well, if at all. In many instances it is creating spaghetti code because a large number of coders do not take the time to edit, refactor, etc.

Today's market watch: Nvidia only dropped about 3% before the boosters swooped in to prop it up...

Michael If your argument is "I guess you didn't read...," then you don't have an argument.  What makes you think I didn't read the article?  Because I disagreed?  Software creation is not just about coding (heck, in the near future we might not even need to code anymore).   And again, the end result is what counts.  This study DOES NOT look at end results and a whole lot of other things.  If a software is bad, there is no easy way to tell if "code quality" is actually the cause, or how much of it is actually attributable to AI, or if the underlying design of the software was so poor to begin with that not even the world's best coders could make anything good out of it.  It's ridiculous that looking at "spaghetti code" or "duplicate code" would give you the impression that you could make any kind of grand conclusion about anything.

Michael All the software made in the pre-AI era, in the 80s, 90s, etc., were they all successful?  If you want to claim that AI makes things worse, it needs to be backed by historical data as well.  Maybe bad software today is caused by poorly qualified people, just like in the past.

Michael I really hate the term "vibe coding." I've got a whole Captain's Log coming up on this. Interesting report. Copy/paste = the sign of lazy coding. Type it in your damn self! And I wonder about Google web results too. My site should show up super high on relevancy for Access-related terms. I don't. My YouTube videos do, fortunately.

Joe yeah... the picture I use now is a real photo of me. Granted it's about 15 years old and I can't say that it hasn't been filtered at all... but...

Kevin don't forget Google+ LOL

As far as AI creating "bad software" all I can say is that I get results that are usually about what Kevin Yip says... 80% of the time it's correct. Granted, I never ask for full programs. Usually it's something that I already know how to do, but I want to save time. Something like "write me a quick sub that will import a text file where every other line starts with Q: and the other with A: and I want to put that into a table with fields Question and Answer." I just did that last night, and it worked fine. More elaborate coding however I have to usually go over it 3 or 4 times to fix bugs the AI introduces. So... it's not there yet, but it's getting better.

Richard Love all the pictures.  As for AI (aka LLM), its still in its infancy state. I just hope we don't give AI any real responsibility until it gets to the "adult" stage.  Image having an AI in the "terrible twos" stage that has access to launch codes.  Yes, that is a reference to "War Games"!  LOL!!!

Sam War Games - such a great movie :)