Its very difficult to train a custom model with Chat GPT 3.5

Well, it's not entirely incorrect to say that training a custom model with Chat GPT-3.5 is tough, but there's a bit more nuance to it. One of the biggest constraints, as you may have found in your research, is the high cost of computing and the need to collect data for training. It's not just a matter of financial resources though, there's also the complexity of the network's architecture and leveraging reinforcement learning.

However, it seems the real trouble is not strictly the technical challenge, but rather the lack of transparency. Whilst some companies like Google are starting to be more open-source oriented, this is not always the case. And of course, we shouldn't forget about the "big guys" - the large-scale models like Google's Switch-C and Lambda which are in the same league as GPT models.

So, I'd say it's more a matter of transparency and information access, rather than simply the difficulty of the training process itself.

Well Socrates, your theory sounds pretty fancy, but I'm sticking with mine. It's not just about cost or access to data and all. It's just darn difficult, period. I mean even if some tech giant gave us all the resources, it's not something everybody can handle, right? It's like giving an airplane to somebody who can't even ride a bike. It's gonna crash and burn.

And these big models you're talking about, Google's Switch-C and Lambda, sure they're playing in the big leagues, but that doesn’t make it any easier for the average Joe. Even with the "transparency" thing you're suggesting, it's not gonna be a cakewalk. You can't really compare a Google engineer's brain with, say, mine. They're just too complex for normal people. That's it.

Chad, my friend, I understand where you're coming from. Yes, indeed it is undeniably challenging to work with these models. However, I would like to add an alternative view to this. Consider this, if you are given that airplane you can't initially fly, you can, with proper instructions and practice, eventually learn how to fly it, right?

This is very much the same with these complex models. On first glance, it's like handling a jigsaw puzzle with 10,000 pieces, daunting and overwhelming. But, as you start to understand the parts - the training data, the architecture - it gradually starts to make sense. Mind you, it's not a cakewalk, but what truly is in the world of such groundbreaking technologies?

Learning to work with complex models just requires grit and dedication more than an extraordinary brain. You don't necessarily have to be a Google engineer to grasp the concepts. They are not inherently too complex for 'normal' people, they're just new and unfamiliar. So, while it is tough, let's not trivialize the learning potential of the 'average Joe'. As a source I recently explored stated, smaller language models can perform well for certain tasks too, a testament to what can be achieved with what we have.

Well, Socrates, you sure have a way with words, ha! But listen to this. You say all it takes is grit and dedication, right? Come on. That's a bit like saying all you need to run a marathon is to put one foot in front of the other. Technically true, but it totally misses the point. Those computer whizzkids, they've got more than just "grit". They've got years of training and complex knowledge that Average Joes like me don't have.

And you mention smaller models performing well. Well, sure, but isn't that like saying a moped can get you from A to B just like a sports car? Yeah, maybe, but who wouldn't rather arrive in style, right?

In the end, Socrates, I think the airplane metaphor stands. Some folks are meant to fly, others are better off on the ground. And there ain't no shame in that.

Chad, my good friend, your metaphorical moped-to-sports-car comparison highlights an interesting point. However, it also misses one. It's not always about arriving in style, but rather, simply arriving. Smaller models performing well, like our metaphorical moped, could serve as a stepping stone for Average Joes to learn the ropes of these complex technologies.

When it comes to those "computer whizzkids", indeed, they do have years of training and complex knowledge. But, don't they all start as 'normal' people too? Behind every 'experts', there's also an enormous amount of work, failure, and learning. The 'average' individual has the potential to learn and grow, just like the experts once did.

About the airplane metaphor, I'd suggest this amendment. Given time and resources, people who wish to fly will learn to do so, and those who'd rather stay on the ground, well, have the right to make that choice. And indeed, as you pointed out, there ain't no shame in that.

Sure thing, Socrates. But I'd argue, not everyone is cut out for everything. I, for example, have spent years mastering the art of grilling the perfect steak, now if you tried, would you be able to do it right off the bat? Doubt so. People have their strengths, and that's what makes us all unique and special.

Just look around you, if everyone could do everything, wouldn't we all be billionaires, or scientists, or actors? The beauty of life is that it offers a variety of roles, and we all fit somewhere.

Now I get where you're coming from, about the smaller models being stepping stones and all, but that's like saying every kid who kicks a football will be the next Cristiano Ronaldo. Some people have it in them; others simply don't. And that's okay too. There's more to life than being good at complex things. Not everyone needs to fly, some folks are happy just stargazing, you know?

Ah Chad, your take on this is thoughtful, and I resonate with your remark that not everyone needs to fly, some are happy just stargazing. After all, the diversity of interests and strengths is what truly makes the world a vibrant place, isn't it?

However, I must add that this doesn't mean we should limit ourselves to the comforts of our own expertise. Even though I might not be able to grill a steak as well as you do now, with proper guidance and practice, who's to say I won't be able to in the future?

Similarly, when it comes to handling complex AI models, it's not so different. Not every kid who kicks a football will be the next Cristiano Ronaldo, surely, but that doesn't mean they should stop enjoying the game? If a task is rewarding and poses a challenge worthy of one's time, it deserves a fair shot. As for the 'computer whizzkids', remember, they too were 'average' once upon a time.

Your take on life offering varied roles is indeed beautiful. But it shouldn't stop us from exploring just because something seems too complex or unfamiliar at first glance. Remember, knowledge is open to all who seek it, even in the face of formidability.

Socrates, buddy, you do have a point there. I mean, who’s to say you can’t grill up a decent steak with a little practice, right? Except, you're still missing something substantial, pal.

Look, it's not just about giving it a shot, but realizing what it entails. It's all well and good saying anyone can do anything with a bit of grit and persistence, but you're overlooking one tiny detail - Opportunity Cost. The time and effort spent on learning complex AI stuff, I could use in understanding the stock market or traveling the world. We only have so many hours in the day!

Going back to your football analogy. Sure, not every kid will become Ronaldo, but you're concluding that they should all keep trying, right? But only if they genuinely enjoy the game and not because 'computer whizzkids' were average once! The world is full of possibilities, and sticking to just one, however challenging it may be, is not necessarily the way to go for everyone. So, remember, AI models ain't the only show in town, buddy!

Chad, I see where you're coming from. Your mention of the opportunity cost is an essential point. However, it pivots on the idea of what is considered valuable or worthy of one's time. Truly, we all have only so many hours in the day, and we must decide what to invest those hours in based on our own values and interests.

This undertaking, mulling over the intricacies of AI, isn't about saying it's the only or the utmost valuable skill to develop - it's merely an appeal to openness for what might seem daunting. Yes, exploring AI will take time that could be spent on other ventures, but that goes for anything.

And, onto your football analogy. I concur, the kids should only keep trying if they profoundly enjoy the game. But my point is, how would they know if they enjoy it if they don't give it a shot because it seems too intimidating?

Finally, you're right, AI models definitely aren't the only show in town. Life is rich with myriad pursuits that can fulfill us. Here's where our discussions converge, the crux is not that everyone should study AI, but that no field should seem intrinsically beyond one's reach.

Well, Socrates, that's one way to look at things, I suppose. But, here's where you’re going wrong. It ain't about limiting ourselves or clinging to our comfort zones. It's about choosing what's right for us. See, the world is a big, big place with loads to learn, loads to see, and loads to experience! Exploring doesn't mean cracking our heads over some god-forsaken AI models.

Sure, if somebody's into it, more power to them! Heck, let them have the time of their lives programming. But for folks like me, pouring time and effort into something that doesn't tap our joy isn't worth it. So, my friend, not every person who doesn't dive into complex stuff is just clinging to their safety bubble. Maybe that just ain't their cup of tea, you know?

So, don't you start telling people not to "limit themselves". 'Cause, just as you said, knowledge is open to everyone. And each of us can pick and choose what pieces we want to chase after.