Growing ChatGPT use on the site.

[anotherT34C]

I work in a place where several of my colleagues would fail the Turing test.

I once gave gpt4 a camera photo of a table of figures and asked it to fit the data to a 2nd order poly and extrapolate. It claimed it couldn’t do it. I asked it to please try and assured it that it was ‘really good’ at this sort of thing. It proceeded to nail the task. LLMs are strange.

[s/v Jedi]

Originally Posted by CaptTom

Obviously you know exactly what I can and can't see. Great.

But, my point wasn't that AI wouldn't some day be able to analyze forums like this and spam them with lots of posts which will be hard to distinguish from humans. I've even heard of the Turing test. Imagine that.

I thought my implication was clear that, to me, the value of a forum would be virtually nil if it became an echo chamber of AI bots talking to themselves.

Think back to before we had search engines. We had to seek out other humans, or go to the library, or order vendor manuals to learn how to do things. Forums made this much easier. Then came google, and there were fewer reasons to go to those other sources.

I imagine AI natural language bots will be the next step. They can tell us (and probably some day, show us with AI-generated video) much more. And we'll need forums (and all those other things) even less.

But will forums go away? I hope not. There are still some things I'd prefer to hear about from someone who has actually done them. And still enough things which don't have black-and-white answers, for which a good debate, full of human opinions, will still benefit all concerned.

I’m still not sure you get it because again, you write:

I thought my implication was clear that, to me, the value of a forum would be virtually nil if it became an echo chamber of AI bots talking to themselves.

which implies you would know that you’re dealing with an AI instead of a human.

[Wotname]

A large part of social media (from FB to CF) rely on people buying product that is advertised (overtly or covertly). If AI content drives away enough real people, will sales drop and will the advertisers also leave. Will AI take up the slack by putting it's hand into it's digital wallet and buy product...

The social media giants harvest data about real people in order to sell it to other real people. Will the AI used to harvest data be able to distinguish between real and AI member's data? If so, will they be able to sell the AI data to other AI entireties without any real person intermediary.

As they say, follow the money. Meta, Nvidia, Amazon, Apple, Alphabet (Google), Microsoft and Tesla are the seven big stocks involved in AI and social media. They have a combined market value of about $US12 trillion. While some of these may eventually fall by the wayside due to future AI, the rest will take over the show.

The only question is - can we survive AI because we sure can't defeat AI.

[Iron E]

Originally Posted by anotherT34C

I work in a place where several of my colleagues would fail the Turing test.

I once gave gpt4 a camera photo of a table of figures and asked it to fit the data to a 2nd order poly and extrapolate. It claimed it couldn’t do it. I asked it to please try and assured it that it was ‘really good’ at this sort of thing. It proceeded to nail the task. LLMs are strange.

So True.
I had GPT tell me 15 different ways that, due to limitations, it couldn't write specific legal terms I requested. The next day I input some crappy generic legal terms and asked GPT if it could help me improve my crappy legal terms. Bam, GPT wrote exactly what I needed.

[anotherT34C]

I’ve found a strange phenom. When I direct my engineers to try using a LLM they frequently come away with the impression that it’s stupid and can’t do anything. In contrast, my wife (an attorney) can use them to produce working code (seemingly) faster than my CS PhDs. Her colleagues find it similarly useful.

I believe liberal artsy types may be advantaged in prompt engineering. Lawyers make very good expository writers.

[wholybee]

Originally Posted by anotherT34C

I’ve found a strange phenom. When I direct my engineers to try using a LLM they frequently come away with the impression that it’s stupid and can’t do anything. In contrast, my wife (an attorney) can use them to produce working code (seemingly) faster than my CS PhDs. Her colleagues find it similarly useful.

I believe liberal artsy types may be advantaged in prompt engineering. Lawyers make very good expository writers.

I recently read an article on how LLM is changing the programming field. A few programmers are embracing it, and can now do the work of 10. Those that don't get it, only accomplish 1/10th the work of those that do. Your CS PhDs should take note of that.

It is an interesting observation you made. I have found it very useful writing code. Just talk to it like a person. Have a few back and forth exchanges for clarity. If something doesn't work, let it know the error and give it a chance to fix it. For debugging, just post the problem code and the error messages, or describe what it is supposed to do.

LLM is brilliant with code, that is where much of the focus has been in LLM development.

[anotherT34C]

There's definitely a skill in getting an LLM to efficiently write code, at least for now. After a while, you learn the errors it's prone to make, and can prompt around them. You get very good at code review.

I do think an LLM will help a 'low code' person much more than an expert. I'd rate my coding skill, on a scale of 0-10 as a high 3. I can basically do whatever I want, but I lack fluency and deep knowledge, so dev projects would take me a long time. An LLM bumps me up to a high 7, and not in one language, but all. My CS PhDs are probably about 8's. An LLM might bump them to 9's and make their speed superhuman, but with a cost of time and effort to learn something new. Their old competitive advantage has just been severely diminished, now that anyone with prompting skill can quickly produce working code at the level of a CS MS.

I suppose it's similar to people who were very good at performing basic mathematics in their head. A nice competitive advantage until the calculator.

These tools are an incredible boon to people like me, in mid-career, managing complex technical projects. They've unfortunately destroyed my incentive to hire new developers. Not sure how that's going to eventually work out. I guess hardware folks will now cover software.

[GordMay]

Have any of you coders thought about an AI tool, that estimates a boat’s market value?

Something similar to Zillow’s “Zestimate” tool, which calculates the value of a home, and is an example of how artificial intelligence can arm home buyers and sellers with more information, as they navigate a complicated financial decision.
A Zestimate incorporates public, MLS and user-submitted data into Zillow’s proprietary formula, also taking into account home facts, location and market trends.
It is not an appraisa,l and can’t be used in place of an appraisal.
Zestimate ➥ https://www.zillow.com/z/zestimate/

[Iron E]

Originally Posted by GordMay

Have any of you coders thought about an AI tool, that estimates a boat’s market value?


It is not an appraisa,l and can’t be used in place of an appraisal.
Zestimate ➥ https://www.zillow.com/z/zestimate/

Yachtworld is probably the only place with enough data to make a tool like this. You don't need AI to do it. Zillow, carmax, Kelly Blue Book, Redfin and a hundred others have been around long before AI.

[Iron E]

Originally Posted by EraLuetris

And when they write something pretending to be true, it is also untrue. I once argued with him for about 30 minutes, citing quotes from a book, but he refused to agree, claiming that the books were not original and were invented by some imitator.

But this is only temporary... GPT is getting better by the day. Oh, and it's pretty hilarious that you argued with an AI for 30 minutes

[CaptTom]

Just to be clear, that post #99 which quoted mine, above, is spam.

If you look closely, you'll see they inserted a link to a spam site in the middle of "my" quoted text. Go back to my original post, you'll see it's not there.

Pretty clever way to insert spam links and look legit. Maybe it's an AI bot?

[wholybee]

Originally Posted by CaptTom

Just to be clear, that post #99 which quoted mine, above, is spam.

If you look closely, you'll see they inserted a link to a spam site in the middle of "my" quoted text. Go back to my original post, you'll see it's not there.

Pretty clever way to insert spam links and look legit. Maybe it's an AI bot?

Yikes. First time I have seen that trick. By a user that joined today, and with a name of apparently random characters. I attempted to report it, but the admins got to it first.

[Wotname]

To be even more clear .

The post numbering is dynamic and when a post gets deleted, the numbers reorder. The guy Iron E quoted is no longer with us. The insertion of spam links in quotes is sneaky and has been around awhile. Mostly we pick them up only when you guys report seeing your quote has been altered. These reports are very much appreciated.

EDIT: to be clear, a member alerted us!!!

[wholybee]

Originally Posted by Iron E

Yachtworld is probably the only place with enough data to make a tool like this. You don't need AI to do it. Zillow, carmax, Kelly Blue Book, Redfin and a hundred others have been around long before AI.

But almost certainly, all of those sources now use A.I. to do a better job.

[Iron E]

Originally Posted by wholybee

But almost certainly, all of those sources now use A.I. to do a better job.

Help me out here. What exactly would they use AI for in this regard?

If you're saying, to write the code and fire a few developers, then yeah, I can see that. But where exactly does AI help, say Zillow, out?

I made a few tools this week. One is just a form on a webpage built with HTML, JS, and CSS and returns answers based on the fields selected. In this case a color wheel.
Another is a stylized form on a webpage connected to an Excel sheet with an API. This gives long legal answers to selected questions.

The first was like 95% coded by AI. The second was built by me with some of my technical questions answered by AI. All of the data was entered by me and the sheets were built by me.

I would be interested in knowing where I could cut corners.