GPT-3 API is now free and open! Is Hive ready for it?

in #hive2 years ago (edited)

Yesterday evening, during the bath, my subscription video feed from the centralized video platform played me a video featuring breaking news. So, apparently, now GPT-3 AI API is free and open for public use!

Did you hear about GPT-3 though? This is the newest generation of language prediction model designed by a San Francisco based company OpenAI. GPT stands for General Purpose Transformer. You can read more at wiki page. And this video shows general possibilities brought by this tool.

These are just 14 basic apps built on the top of this powerful engine, there is much more, and even this only hints just a small bit of possibilities to come in the following years. According to Stanford University report the computational power of AI doubles every 3months! If you have even the slightest understanding of exponential, you know this is really fast. But how can that influence the future of Hive?

I just asked GPT-3 basic playground mode about Hive. This is what I got. Yeah, the very specific apps from the video are much more impressive.

The hive blockchain is a new type of blockchain with a built in voting system. This blockchain is designed to allow the community to have a say in how Hive should grow. It is also designed to allow the community to have a say in the direction of the hive development.
The hive blockchain is currently being developed by the hive Dev team.
It is expected that Hive will have its own blockchain by Q1 2018.
The Hive blockchain will have the following features:
• A built in voting system
• A built in reward system
• A built in messaging system
• A built in currency
• A built in marketplace
• A built in exchange
• A built in wiki
• A built in news system
• A built in forum

Having access open API allows potential abusers (talented coders) to spam our blockchain with AI-generated articles and poems, in general indistinguishable from the original content. The whole foundation of the Proof of Brain model is based on the value of content generated by real people. Isn't it? But what happens when digital brain's capacities surpasses human's one?

Obviously, long, good quality content is just a part of the votes post earns. Personal qualities, community-network status and relevance of the content are possibly even more important. But the appearance of such a powerful tool will inevitably change environment we function in.

In my opinion this creates situation when human to human genuine interaction on Hive should be valued much more than quality of content. The day to day blogging aspect of Hive seems to be more valid over writing about some specific topic. Also, the pressure for identity verification (cross platform type, never KYC) grows even further. I have however, more questions than answers.

Will it be harder for outside content creators to join Hive? Could GPT-3 app fed with the most rewarded Hive post create content which will have significantly higher reward? Could GPT-3 fed with your own past posts help you create content on the chosen topic, which resemble very much your own work? How can we work out in the new conditions to avoid the crash of Hive value proposition as a social media platform? Can you imagine no-reward-by-default ghost-accounts creating after the Hive author's death which will continue to produce opinion forever? How can we avoid becoming AI-filled echo chamber in the long run, in general as social media?

I believe these are very important questions for the Hive community to answer. I'm going to research this topic in the following days, even though I have no technical skills to explore full potential of GPT-3. I'd love to see somebody creating real-deal tool, which could attack one of the PoB chains to perform a test in the safe conditions.

And who knows, might the most tech-savvy Hive users already be using it?

What are your thoughts?

Following are thoughts of the AI. Please note, the following text doesn't reflect full capabilities of GPT-3, hence this is just general purpose predictive text generator. Specific applications are much more impressive.

Could GPT-3 fed with your own past posts help you create content on the chosen topic, which resemble very much your own work?
If you have a wide range of topics on which you have already created content, the algorithm may attempt to create new content on a similar topic.
So, be sure to add to the GPT-3 niche content of different types on different topics.

Can you imagine ghost-accounts creating after the content author's death which will continue to produce opinions forever? I think it's a great idea. A ghost-account of the content author, as a ghost-account of the content author.
Hopefully, I won't be dead for a long time. But I am writing stuff, so if I die, it will still be good.

To avoid becoming AI-filled echo chambers in the future, social media can’t become a tool to avoid the real world. Instead, it has to become a tool to make the real world a better place.
That means allowing people to actively engage in the community and with each other, rather than allowing companies to keep them passive.

I created separate @gpt-3 account to test possible outcomes. If there would be anybody willing to join adding generated content there, reach me out for the posting authority.


Yeah there are a lot of dangers, but lets be positive for a moment. I can imagine an application of GPT-3 or other open and huge NLP system to automatically summarize posts on Hive frontends. Not a big deal but could be useful to save some time.

Ty żyjesz!

Żyje, żyje ;)

This sounds like a great utility!

How much should we value automatically generated posts then?

This is a tricky question. Ideally to me we should value any post by some kind of nonlinear combination of:

  • quality (this is semi-individual measure determined by a reader)
  • ability of a post and its author to reach to public outside of Hive
  • and here comes a measure that could potentially value automatically generated posts lower - uniqueness, that is probability that an agent (human or AI model) could have produced this post given knowledge that is publicly available to anyone (the lower probability the higher uniqueness)

Yeah, generally I agree, I personally add my affection towards certain users and behaviour on the blockchain. Did he/she power-downs all the earnings? Is there a self-voting going on? This seems to be sustainable IMO.

Great topic and good questions to ask.

I think we are going to get to the point where "proof of human" is going to be vital. We are likely to see reputation systems becoming more important, which will be based one's interactions as opposed to a simple number based upon upvotes.

As for GPT-3, this is the fear that we can see massive moves towards usurping what humans can do. This is true if for no other reason for speed. I would think this is a threat to the entire University system since papers can be written mostly from this.

Very pertinent questions we will have to address.

Posted Using LeoFinance Beta

This is one of the most important issues that I rarely see people having open discussions about.
As someone who has a lot of amateur experience using GTP3 as a tool I know that the level of coding work involved to make meaningful content using this neural network is way too high. The cream will rise to the top and low quality GTP3 content couldn't compete with people who work hard on their posting.
I am very open to continue conversation about this topic.

Right now this might be a truth. However few years ahead and landscape might be different. Me, having no coding skills wasn't able to create any miningful, but it's a matter of time for tools making it possible to come live. It's good to start this discussion beforehand.

The rewards earned on this comment will go directly to the person sharing the post on Twitter as long as they are registered with @poshtoken. Sign up at

GPT-3 is a very powerful tool. It is still in beta and has much potential.


Congratulations @deepresearch! You have completed the following achievement on the Hive blockchain and have been rewarded with new badge(s):

You received more than 15000 upvotes.
Your next target is to reach 20000 upvotes.

You can view your badges on your board and compare yourself to others in the Ranking
If you no longer want to receive notifications, reply to this comment with the word STOP

Check out the last post from @hivebuzz:

Hive Power Up Month - Feedback from Day 18

GPT-3 is not very good at spitting facts, though if's supplied with a bulk of text to work with, it can perform summaries, rewrites (big issue when it comes to plagiarism detection) and much more.
It shines best though when given some hypothetical prompts. I was already super impressed with GPT-2