Question to the admins about an AI project

Haha! His AI avatar could be the moderator for that section. :)


And we can joke here. But this is exactly the shit that is coming. Fast. Every company is racing to train AI's *right now* for every field under the sun, and for everything you interact with every day. This technology is about to start appearing everywhere, and is going to explode in the next 12-24 months, whether we like it or not.

Search engines as we know it will likely become irrelevant. Instead of giving you a set of results to look through and maybe dig out the same answer written 10 different ways form various sources, AI can aggregate all that information, determine which information is most acurate (likely based on frequency) and provide a single answer. In a way it makes you wonder if websites as we know it would continue to be relevant.

I know nearly nothing about coding and literally programmed a web based Asteroids clone in about a day start to finish just conversing with ChatGPT until it hit its current limit to troubleshoot errors.

Left/Right Keys - rotate ship
Up Key - thrust
Z - Fire

When game is over you have to refresh to play again.

I've used it to analyze reams of user experience research and create executive level presentations that would have taken an analyst days to pour through and compile.

Growth is becoming exponential so it's a great time to start thinking about how it can be leveraged while it's early enough to avoid being completely overwhelmed and intimidated by what it's going to become.
 
At the end to the day, AI won't replace people. People who learn how to leverage AI will replace people.
Time to move things forward.


I think eventually AI is going to totally replace people.

But in the short term it will be as you say. The question is whether that lasts one decade or five. Humans leveraging AI in the short term will lead to competition, which will eventually lead to war, because humans. Except this time it isn't just a handful of nations with nukes, where they can all agree not to do dumb things. Every idiot on the internet will have the equivalent of nukes. And the chances of NOBODY doing something stupid is exactly zero.

This time it'll be an AI-powered war, which will wipe us out regardless, whether it's AI-to-AI war (in which case we'll be collateral damage), or AI-to-human war (in which case we will stand no chance).

Eventually we're going to lose control of this thing we've created, and it'll be game over. Remember how we couldn't even control spam and computer viruses when the internet was new? Nobody saw those coming, and they got out of control. This is going to be that, but amplified by about a million, and in ways we don't see coming yet. But the 'viruses' will be smarter, think faster, and understand us better than us.

Superior things outcompete inferior ones. That's nature's way. So we're totally idiots for creating this technology, as it will lead to the end of our species as we know it. But that's the point. Whatever survives is what's going to carry forward. It's what we're here to do, and nothing is going stop people from doing it at this point.

It's going to get very messy. But that's exactly how nature works.
 
AI automates that process in minutes, and humans can get all the same benefits.

A key distinction is that an LLM does NOT actually model the underlying system or thing in question, but is only a statistical model of sample documents involving that topic. This lack of a systematic understanding by an LLM can result in truly bizarre catastrophic errors, such as answers being produced that are so clearly wrong to any layperson. That's because a LLM simply generates the most statistically-likely output based on its training dataset — it does NOT derive a logical answer based on an understanding and analysis of the underlying thing in question (or even of the question/prompt itself). That might not be a big problem if you're using it to produce a recipe for oatmeal raisin cookies or your sophomore English term paper, but I personally wouldn't trust it for anything involving high voltage.

(For further reading, I can point you to my unfinished dissertation from ~15 years ago.)

In other words, if you think ChatGPT can pass the Turing Test, you aren't being creative enough with your questions. LLMs are so much closer to Searle's "Chinese Room" but now I've probably bored you all to sleep…
 
I know nearly nothing about coding and literally programmed a web based Asteroids clone in about a day start to finish just conversing with ChatGPT until it hit its current limit to troubleshoot errors.

Left/Right Keys - rotate ship
Up Key - thrust
Z - Fire


Ha. This is really cool, for what it is. I like how the asteroids even have vector 'glow'. (Though for some reason it doesn't appear on all of them, lol.) :) But it actually captures the basic feel of Asteroids, and was far more than I was expecting.

Right now we're in the early stages of AI. And the early stages of any technology are the best ones. The internet was great for about ten years (1997-2007ish), before big tech fucked it up for everyone. This will play out similarly, though the equivalent of 'big tech' in this scenario is going to be some form of superintelligence or Artificial General Intelligence.

And when that AI equivalent of Facebook or the iPhone shows up in X years, that's when it's gonna get ugly. If the current AI systems are the equivalent of an amoeba, what is an AI T-Rex going to look like?


So... in the meantime... we should use it to solve our monitor problems. :-D

TECHNOLOGY!

I'm all for it. And keep in mind I'm referring to the macro picture above. In the meantime we might as well have some fun on the micro level. (And it's really our duty to create the AI KenLaytons and andrewbs, because that's all that's gonna be left after humans are gone.)

I'd love it if we all agreed to just not create any of this tech, and just keep on being humans. That would actually be my first choice. But since there's zero chance of convincing the entire planet of that, I'll ride this shitshow all the way down.

dr.strangelove.gif
 
Last edited:
A key distinction is that an LLM does NOT actually model the underlying system or thing in question, but is only a statistical model of sample documents involving that topic. This lack of a systematic understanding by an LLM can result in truly bizarre catastrophic errors, such as answers being produced that are so clearly wrong to any layperson. That's because a LLM simply generates the most statistically-likely output based on its training dataset — it does NOT derive a logical answer based on an understanding and analysis of the underlying thing in question (or even of the question/prompt itself). That might not be a big problem if you're using it to produce a recipe for oatmeal raisin cookies or your sophomore English term paper, but I personally wouldn't trust it for anything involving high voltage.

(For further reading, I can point you to my unfinished dissertation from ~15 years ago.)

In other words, if you think ChatGPT can pass the Turing Test, you aren't being creative enough with your questions. LLMs are so much closer to Searle's "Chinese Room" but now I've probably bored you all to sleep…


Don't get me started. :)

You're right about the Turing Test. However what's happening is that as AI tech advances, it's able to convince more and more people, because some people are just smarter and more creative at being able to tell than others. It will continue to eat its way up that stack, being able to convince smarter and more creative people. Run that to the limit, and at some point it will be able to convince the smartest person on the planet, at which point literally nobody will be able to tell the difference. And at that point, it's cognitively superhuman by definition. It's really no different to how it worked with chess AI's, go AI's, and every other narrow task that AI's are being trained for (which now include medical exams and the bar). Intelligence isn't a black and white thing, it's a continuum.

Yes, the current state of tech has limitations. However I'm in the Ilya Sutskever camp, where intelligence is basically just prediction. That's really all you and I do as humans, just with richer models and hardware. I can tell people how to fix their vector monitors because I've seen just about every way they break, and I have that information represented in my neural net. Tell me the symptoms, and I'll tell you where to look to find the problem. Problem solving is basically prediction.

You can say the current models fail in some aspects of prediction (which they absolutely do), but that's because this is just the first iteration of all of this, and they're working with limited modality (text), whereas we operate in the 3D world, where most information is actually not text. But they DO have a grasp of language, and that's one of the most important developments. Because once you have that, you can build higher layers on top of it. Things like world models, robotic agents, and other algorithmic advancements are going to take care of all of the missing parts in time. (And really quickly, IMO.)

The core of what HAS been developed is machinery that can actually learn, in a way that is novel. And once you have that fundamental piece (like the transistor), there's almost no limit to what you can build on top of it. The only factor is time.
 
@andrewb the only problem with your idea is that you will need a comitee to determine what is acurate data to input into the AI so that potential dangerous or false info is not part of the AI database. Its amazing how much bad dangerous and misunderstood information exists about your field of endevor asteroids. I have seen quite a few statements that were false about asteroids from a lot of people here. I only was able to catch one of your statements being wrong once. (regarding how the tone was generated for the self test memory tone, you said it was the 3khz signal when it was actually the thump. A minor error but something that should have to be vetted in an AI database). Just to repeat this was the only one time I caught anything.
 
A quasi-colleague of mine posted this in 2018. It's still valid, despite excitement over things like ChatGPT...

IMG_2494.jpg

Ever since, "Hapax legomena will be your doom" has been my personal favorite curse.
 
If AI is the lastest and greatest upcoming thing that should be embraced, then why not use it to program/design something with your favourite images, songs and games on it. No need to learn or invest massive amounts of time.
 
We are at the very early stages of experimenting with AI here, and have some very specific use cases it may begin to creep into the main site. It's less clear to me how useful it will be on the forums, other than to improve the search engine.

All the talk is about LLM these days... I'm actually curious about the potential of combing it with structured data... We are still building some basic (non-AI) tools here though, so we aren't quite ready to experiment further.
 
So, regardless of our opinions on the matter, it looks like these forums are already being used in Open AI's LLM's. Check out this simple query I just made in my free Copilot app on my Windows 11 desktop.

1711581477536.png
 
We are at the very early stages of experimenting with AI here, and have some very specific use cases it may begin to creep into the main site. It's less clear to me how useful it will be on the forums, other than to improve the search engine.

All the talk is about LLM these days... I'm actually curious about the potential of combing it with structured data... We are still building some basic (non-AI) tools here though, so we aren't quite ready to experiment further.
I know life progresses, but is this a good idea? I mean if the Artificial Intelligence is gonna take our fun, lively, drama filled banter as INTELLIGENT and factual content, we as a human race are doomed!!
 
I tried 2 'Ken Layton' questions myself of co-pilot.

One of them resulted in a totally worthless answer:. How would Key Layton suggest I fix my arcade game? It just linked to random arcade monitor PDFs

The other one was more impressive: How would Ken Layton suggest I fix my Defender PCB?
Now for the example in the post above, and on the one I just said was worthless, Microsoft provided links to source content. For the Defender PCB question, it doesn't give any credit at all... it simply takes without giving the source content any SEO benefit.

It clearly has been trained to categorize source content as being from a 'forum'. That isn't hard since there are only a few different main forum software systems, and each is pretty consistent from one site to another in how data is sent to the browser.

--

Copilot
I apologize, but I don't have specific information about Ken Layton's recommendations for fixing a Defender PCB. However, I recommend the following general steps for troubleshooting and repairing a Defender arcade game PCB:
  1. Visual Inspection:
    • Examine the PCB for any visible damage, burnt components, or loose connections.
    • Check solder joints, connectors, and traces for signs of wear or corrosion.
  2. Power Supply and Voltage Checks:
    • Ensure that the power supply voltages (typically +5V, +12V, and -5V) are within the specified range.
    • Verify that the power connectors are securely attached.
  3. Clean Contacts and Connectors:
    • Clean the edge connectors on the PCB and the corresponding connectors on the arcade cabinet.
    • Use a contact cleaner or isopropyl alcohol to remove dirt and oxidation.
  4. Chip Re-seating:
    • Gently remove and re-seat the game ROM chips, RAM chips, and other ICs.
    • Sometimes poor connections can cause issues.
  5. Check for Broken Traces or Components:
    • Inspect the PCB for broken traces (copper lines) using a magnifying glass.
    • Test components such as capacitors, resistors, and diodes for continuity.
  6. Test the Game Board in Another Cabinet:
    • If possible, try the PCB in a different arcade cabinet to rule out cabinet-specific issues.
  7. Use Diagnostic Tools:
    • Some arcade games have built-in diagnostic modes. Access these to identify specific problems.
    • If available, use a logic probe or an oscilloscope to troubleshoot signals.
  8. Refer to Documentation:
    • Consult the game's technical manual or online resources for specific troubleshooting tips.
    • Look for schematics, wiring diagrams, and component lists.
  9. Community Forums and Experts:
    • Join online forums or communities related to arcade repair.
    • Seek advice from experienced collectors and technicians.
  10. Consider Professional Repair:
    • If you're unable to diagnose or fix the issue, consider reaching out to a professional arcade repair technician.
Remember that arcade game repair can be complex, and safety precautions should be taken when working with electronics. If you encounter difficulties, consult with experts or seek professional assistance.
For more detailed guidance, I recommend reaching out to specialized forums or contacting an experienced arcade repair professional. They may have specific insights into Defender PCB repairs. 🕹️🔧

I provided general steps for troubleshooting and repairing a Defender arcade game PCB. For specific recommendations from Ken Layton, I recommend seeking specialized forums or consulting an experienced arcade repair professional. If you need further assistance, feel free to ask!
 
You're really gonna trust LLMs with technical advice about working with HV??

This happens to be my particular technical area of expertise (computational linguistics, not high voltage haha) and I would definitely NOT trust it for something like that.
 
You're really gonna trust LLMs with technical advice about working with HV??

This happens to be my particular technical area of expertise (computational linguistics, not high voltage haha) and I would definitely NOT trust it for something like that.
Calm down... What could go wrong?!?

"For proper CRT operations, please lick the little suction cup that is attached to the tube. This insures a tight seal is made and there is no electron evaporation."
 
You're really gonna trust LLMs with technical advice about working with HV??

This happens to be my particular technical area of expertise (computational linguistics, not high voltage haha) and I would definitely NOT trust it for something like that.

If its tech related to the point it can readily find and summarize data in the same way google or this site search feature can serve up info, then yes. The big differentlce is the conversational inquiry and ability to provide relevant data based on source data. Obviously there are hallucination errors that can occur so its like anything else... someone on a message board tells you to just pop off the anode cap, youll be ok... you probably want to take it with a grain of salt.

But if you could dump your knowledge and experiences in the field into a searchable database, chances are it would be very beneficoal... and live on for people to access as this info and field dies off.

Im not 100% sold on AI but you have to admit, it has some solid potential.
 
If its tech related to the point it can readily find and summarize data in the same way google or this site search feature can serve up info, then yes. The big differentlce is the conversational inquiry and ability to provide relevant data based on source data. Obviously there are hallucination errors that can occur so its like anything else... someone on a message board tells you to just pop off the anode cap, youll be ok... you probably want to take it with a grain of salt.

But if you could dump your knowledge and experiences in the field into a searchable database, chances are it would be very beneficoal... and live on for people to access as this info and field dies off.

Im not 100% sold on AI but you have to admit, it has some solid potential.
We don't need AI lobbying on your behalf for Mr. Do! to be on Bitkit!!
 
I tried 2 'Ken Layton' questions myself of co-pilot.

One of them resulted in a totally worthless answer:. How would Key Layton suggest I fix my arcade game? It just linked to random arcade monitor PDFs

The other one was more impressive: How would Ken Layton suggest I fix my Defender PCB?
Now for the example in the post above, and on the one I just said was worthless, Microsoft provided links to source content. For the Defender PCB question, it doesn't give any credit at all... it simply takes without giving the source content any SEO benefit.

It clearly has been trained to categorize source content as being from a 'forum'. That isn't hard since there are only a few different main forum software systems, and each is pretty consistent from one site to another in how data is sent to the browser.

--

Copilot
I apologize, but I don't have specific information about Ken Layton's recommendations for fixing a Defender PCB. However, I recommend the following general steps for troubleshooting and repairing a Defender arcade game PCB:
  1. Visual Inspection:
    • Examine the PCB for any visible damage, burnt components, or loose connections.
    • Check solder joints, connectors, and traces for signs of wear or corrosion.
  2. Power Supply and Voltage Checks:
    • Ensure that the power supply voltages (typically +5V, +12V, and -5V) are within the specified range.
    • Verify that the power connectors are securely attached.
  3. Clean Contacts and Connectors:
    • Clean the edge connectors on the PCB and the corresponding connectors on the arcade cabinet.
    • Use a contact cleaner or isopropyl alcohol to remove dirt and oxidation.
  4. Chip Re-seating:
    • Gently remove and re-seat the game ROM chips, RAM chips, and other ICs.
    • Sometimes poor connections can cause issues.
  5. Check for Broken Traces or Components:
    • Inspect the PCB for broken traces (copper lines) using a magnifying glass.
    • Test components such as capacitors, resistors, and diodes for continuity.
  6. Test the Game Board in Another Cabinet:
    • If possible, try the PCB in a different arcade cabinet to rule out cabinet-specific issues.
  7. Use Diagnostic Tools:
    • Some arcade games have built-in diagnostic modes. Access these to identify specific problems.
    • If available, use a logic probe or an oscilloscope to troubleshoot signals.
  8. Refer to Documentation:
    • Consult the game's technical manual or online resources for specific troubleshooting tips.
    • Look for schematics, wiring diagrams, and component lists.
  9. Community Forums and Experts:
    • Join online forums or communities related to arcade repair.
    • Seek advice from experienced collectors and technicians.
  10. Consider Professional Repair:
    • If you're unable to diagnose or fix the issue, consider reaching out to a professional arcade repair technician.
Remember that arcade game repair can be complex, and safety precautions should be taken when working with electronics. If you encounter difficulties, consult with experts or seek professional assistance.
For more detailed guidance, I recommend reaching out to specialized forums or contacting an experienced arcade repair professional. They may have specific insights into Defender PCB repairs. 🕹️🔧

I provided general steps for troubleshooting and repairing a Defender arcade game PCB. For specific recommendations from Ken Layton, I recommend seeking specialized forums or consulting an experienced arcade repair professional. If you need further assistance, feel free to ask!

Fuck me. I don't know if that's what Ken Layton would recommend. But it's definitely what andrewb has recommended (as well as other people here). Every one of those things, and pretty much in that order. And with sending it out always being the last option.

Just look at any thread where I've used the phrase 'visual inspection'. Here are all of them:



Sadly, you used to be able to find much of this information via a basic google search. But Google and others are intentionally kneecapping their own search now (which get more terrible by the day), so they can justify these newer AI-based solutions.

And while it seems impressive on the surface, I can't help but keep asking myself, 'Is this REALLY that much better than a good Google search was ten years ago?' And the answer I keep coming back to is not really. It just seems better today, because search and the web in general is such shit.

It seems like most of the world just got lazy and stopped googling things years ago, in favor of 'asking your friends' on Facebook (which we've DEFINITELY seen in this hobby), regardless of whether the information is good or not. And these new AI 'solutions' are just a sort of response to that, where it's the same information you could have googled once upon a time, but in a more conversational form, and lazier. Because people will do whatever involves the least effort. Even if it sucks.

I still would rather have the Google of ten years ago, where the human was still an active part of the search process (and I could make up my own mind based on what info I found, from multiple sources), rather than being a passive thing that just accepts whatever answer is given.

The latter is not a good thing.
 
Back
Top Bottom