poll IS DOMAINER GPT THE NEXT STEP?

[QUAD DOMAINS]

BY DEFINITION:

DOMAINER GPT - An Ai query tool that exists for the sole purpose of providing domain investment insights and guidance.

Is this the next step for domaining and GPT technology? Comment below.

🗣 Mel - QUAD DOMAINS

 

[trsaso]

Why not to use AI or ChatGpt You can automate most part of the hard work using AI.

 

[rokoroko]

AIDNTools are the right answer?

 

[Fayaz Ahmed]

I've posted this here on namePros a few of days back: Are domain names becoming less valuable in the near future?

So you can see that I'm aware of this current trend with Ai.

However, I don't think that GPT technology has what it takes to be a decent domain investor.

The main reason is, as the GPT definition suggests, these are Generative Pre-trained Transformers. What that means is, it's a system that was trained with existing content to generate similar sounding content.

So at best, a good GPT will sound like an average domainer, who takes in some good advices from great domainers, but at the same time, also takes in some poor understanding from beginning domainers.

Besides, at the very core of a decent domainer, there must exist the understanding and intuition of new and emerging trends. GPT tech. by definition has neither. It can only generate content based on existing content, it can't predict the future. There are other types of data analysis Ai that can possibly make some predictions, however, I'm not aware of any such innovation that can predict domain name trends.

So unless something significantly changes in the Ai tech to incorporate business naming prediction for the future, we'll not see someone becoming a decent domainer only by taking suggestions from GPT. Also, even if Ai catches up with newer innovations, it'll most certainly not be called GPT - because it'll not be Generative Pre-trained Transformer anymore.

 

[Acion]

1. it will be easy for AI program to search from .xyz then xyz .com so com will not be the only king in future.
2. AI will always be dependent on data created by human beings so humans will remain there lol.

Regarding #2, that is not the case anymore. Most of my vision models generate their own training data now. There is a discriminator (also a model) that judges the quality (FID, etc) outputs (thousands of them) and we feed it back in. This is incredibly useful in fine-tuning models for purpose where there is very little real data.

In fact, there is not enough human data to train models now. We've exhausted the entirety of the data available and it's still not enough to pass fit especially in large language models. The thinking used to be that larger models and bigger GPUs was the key and even reusing data on multiple epochs in training looked good. Now, it turns out, it's much more important to train for longer on more data which the whole of humanity does not have. So, now we make it. Ton's of it.

You wonder why the big companies aren't locking you in claiming copyright to your use of the outputs of these models? First, there's no point. Second, you are becoming a part of the process. You're rating the outputs of the model by your usage of that data on the internet in some way is a great way to grade it. Rock on!

 

[Grootian]

Regarding #2, that is not the case anymore. Most of my vision models generate their own training data now. There is a discriminator (also a model) that judges the quality (FID, etc) outputs (thousands of them) and we feed it back in. This is incredibly useful in fine-tuning models for purpose where there is very little real data.

In fact, there is not enough human data to train models now. We've exhausted the entirety of the data available and it's still not enough to pass fit especially in large language models. The thinking used to be that larger models and bigger GPUs was the key and even reusing data on multiple epochs in training looked good. Now, it turns out, it's much more important to train for longer on more data which the whole of humanity does not have. So, now we make it. Ton's of it.

You wonder why the big companies aren't locking you in claiming copyright to your use of the outputs of these models? First, there's no point. Second, you are becoming a part of the process. You're rating the outputs of the model by your usage of that data on the internet in some way is a great way to grade it. Rock on!

Thank you. you just yourself validated my point. these models ran out of data which means not enough "Human" generated data research there. you can do more micro refine these models but at the end there training will remain dependent on human data. those discriminator models also judging the quality based on available correct human research or incorrect human research. it is possible when AI would be able to do correct research itself to produce that data and we are far behind in it atm. Am Example, Doe needs an eye operation you can train AI to do that for you based on Already existing research. great work AI. what about a new eye disease ops AI can bring in new data and now those discriminators can not judge this new disease coz no data available. good thing is it brought back new data now a human doctor will work on that data to make it correct or incorrect for discriminators. so judgment will always be based on humans coz bits can not think. what about if i train my bits for 0 as ON AND 1 as OFF dangerous right.
Untill Human Progress is there AI cant be Perfected. so its not dangerous to humans or take over as some people believe. ya we would lose our jobs so its better to create or have some kind of digital asssets as these will be your future full time work. and i would recommend to buy a domain and build your own asset on it ya i am biased lol. i am sure future assets will include how many websites social accounts apps you are leaving for your childrens.

 

[papeter]

The problem with GPT is that it programmed to deliver popular (and not necessarily factual) responses.

In fact, let's test it out.

While the advice isn't necessarily bad, it's not particularly useful either. Some of it is even a bit questionable, and could easily be misconstrued by inexperienced domainers.

You could probably summarize the list of criteria for a useful domain on a sheet of paper the size of a business card. Beyond that it's just people determining what a particular domain is worth to them.

The new CHATGPT version 4 might be more useful.

 

[MKA]

The new CHATGPT version 4 might be more useful.

No, because it's likewise conditioned to produce popular responses, not accurate responses. Sometimes these overlap, but not always.

But there's another issue to consider as well, and that's the fact that Open AI are responsible for the financial advice given by Chat GPT. If Chat GPT was to tell me that Tesla is undervalued and that investments in Tesla would yield 10 - 30% in 5 years, and I invest and end up not making this money I have legal grounds for suing Open AI for bad investment advice. This would also apply to domains. If ChatGPT tells me it's a good idea to spend $10,000 on a domain because it will sell for at least $25,000 in 3 years, then unless that actually happens then the responsibility is put on Open AI. So Chat GPT will be coded so to not produce these responses.

To be fair. If you're new to domaining it might be useful because it might highlight things you didn't know, like making the domain memorable or avoiding hyphens. But that's the extent of the use I see for it. Keep in mind, it's not magic.