Leo: The Power of Recommend Engines
This is something that I have been harping upon. Quite frankly, I am surprised that more aren't talking about it on LEO.
When it comes to Web 2.0, recommend engines are commonplace. We know the algorithms are working overtime on sites such as YouTube and X to provide people with content it things they want. The goal is to get people to engage more, hence remaining on the site.
We would be remiss, of course, to exclude Meta from this conversation. Zuckerberg might be the king of this, bordering on a psychop that it uses on people.
Either way, this is something we really do not see in Web 3.0. This is especially true on Hive where data enters and is buried.
For example, when was the last time you were able to easily pull up a thread from two years ago?
It is not easy. This is why the recommend engines are crucial.
Showing The Power of the Recommend Feature
Things are improving. Hopefully, we can see a massive step forward in some areas. That said, I do have some reason for optimism.
Here is an article I was replying to earlier.
Notice the date on it. This was from two months ago and was located at the bottom of an article. While the recommendations can be wonky, this one was actually somewhat related to the topic.
When I clicked through it, this was something that I had voted up but did not comment. Hence I was able to drop a comment on an older thread.
Is this a big deal?
On its own, not in the slightest. However, when we step back, it can have a major impact in the aggregate.
As more content gets in front of people, it can stimulate activity. If nothing else, this generated another page view.
The key is to think about this across the entire platform. What if everything that is entered through Leo has the potential to show up. That means we will be accessing what is in the database instead of just having to depend upon the most recent additions.
Of course, a lot of stuff submitted is time sensitive, such as the price of Bitcoin on a given day. Quite frankly, we do not care about the price of a coin 3 months ago.
There is, however, a lot of content that falls outside of this. We can see how the stickiness of music, movies, motivation, and success can apply. The same is true for entertainment.
Is a joke less funny if it is two years old? If it is time sensitive, then it is. Outside of that, it still applies.
Shorts: TikTok Style
Over the last few days, the stability of shorts have improved. It appears we are about to move onto the next phase of this development.
Here is a short that was put out by @khaleelkazi:
https://inleo.io/threads/view/khaleelkazi/re-zsxnixldhp
As we can see the shorts feed page should go live. There were some bugs, especially on mobile that still require attention. It seems the team is on it.
Here is what the page will look like.
https://inleo.io/threads/view/khaleelkazi/re-leothreads-2mrhhmvlv
This is obviously an video that was posted horizontally, when shorts prefer vertical. Hence we can see how it was adjusted to fit the layout. Just like TikTok or YouTube, the page is scrolled down, with a new short appearing.
Here is what it looks like in action.
https://inleo.io/threads/view/khaleelkazi/re-puoykvfqxu
Having a page like this is crucial. When it comes to getting content in front of others, it is a must. As stated, this is a natural part of Web 2.0 yet is absent for much of Web 3.0.
Leo, it appears, is changing this.
The next question is how the algorithm is trained to feed into this. For the moment, since there aren't a ton of videos, it probably doesn't matter. But over time, if this is a heavily utilized function, we are going to need the engine to sort through the content.
Here is where things can change completely.
When shorts were mentioned as being in testing a few weeks back, I mentioned that it was a 3 step process:
- shorts
- dedicated page
- recommend engine
According the one of the videos, we will have the second one implement soon (hopefully tomorrow). After that, we will have to find out about what the machine learning capabilities are.
Posted Using InLeo Alpha
Very timely opinion and suggestion.
It's not just about adding data, how do we manage it and most importantly how do we retrive it spontanepusly, that is what the economy and USD of a social engagement is.
It is all a part of it. The challenge is what is in the database to begin with. Here is where the community dictates things. An algo isnt going to do anything is people do not participate in training it.
As for the retrieval, that is a problem for Hive.
It is up to the people and the community to decide how diverse they can build the data, add the data. Wider the data, better gambling of it by AI.
Without a doubt. I am not sure most realize how all this works. The more each person inputs, the more the output is improved.
Both on the individual and collective level.
I think we need to enhance this feature. I get recommended some odd things. But then do I really want Khal knowing my porn kinks via cookies? Lol
There are improvements required. Of course, it is also up to the individual to actually participate in training it.
People seem to think algos are just going to recommend without any input.
this is true, I often engage in most the new things. Been hard to keep up though
that’s the key difference between web2 and web3
You’re giving data via cookies to web2
On web3, we just use all the data that’s already public on the blockchain (who you’re replying to, upvoting, etc.)
That's a great solution to the web 2 problem. I guess my comments and engagement switches alot so I probably confuse the heck out of it!
This topic, in my opinion, is one of the missing aspects of Hive and a reason why it doesn't attract more people from the outside or encourage newcomers to stay longer.
People prefer platforms that automatically feed them content tailored to their interests, rather than having to search for content they like. On Hive, the latter is typically the case.
Hive is essentially a vast database of transactions, links, blog posts, and more. However, unlike social networks like Meta and X, it is not designed with graph network algorithms that continuously work to connect users to other users and content. Hive operates more as a meritocratic system that expects users to compete for rewards "organically," with its recommendation system relying heavily on manual curation. Additionally, monetization is limited to the first seven days after content is posted.
Solving this problem is quite a challenge and is currently being addressed by frontend dApps like Leo.
I agree, but I would also suggest that "as more curated content based on users' activity, likes, and engagements is presented, more users may stay and contribute to increased activity."
However, this approach could also introduce new challenges, such as concerns over privacy.
Beyond that, if there could be a single or decentralized graph network algorithm that continuously evolves to connect everyone and everything on Hive, I think it would have a significant impact.
How could there be concerns over privacy? Everything people are doing is being fed into a public database.
Well, you're not wrong about that. However, concerns over privacy can vary depending on how a recommendation algorithm uses our personal data. It’s a double-edged sword.
When it comes to content recommendations that enhance users' experience on the platform, I think we can both agree that it’s generally acceptable.
But what if that data starts being used for other purposes, like selling it to brands, politicians, or others? Wouldn't that raise concerns about privacy and the use of personal data?
On the other hand, why do most people on this platform hide their identities? For example, I wear a mask in my profile picture, and you use a cat. Can we agree that it's because we don't entirely feel safe sharing our data here?
So, in my opinion, it depends on how the data is being used—or whether one can simply opt in or out of the algorithm.
Yes people can remain as anonymous as they want on here. That is a personal choice.
Again, when you mention the data, what personal data are you referring to? You mention politicians buying the data? Why would they do that? It is on chain. Anyone can scrape it.
If it is on chain, it is public. How personal people get with what they reveal, well that is up to the individual.
I myself is interested to work in solving this problem. If Id have a time, maybe, I'll try to dig into the database and implement some opensourced graph algorithm out there. Let's see 😄
You can push a privacy layer over the top certainly. That is something some have talked about. It could enhance the ability for people to mask what they are doing since right now it is wide open. Everything on chain is open to being scraped and parsed.
the privacy question is actually solved in a massively important way
On web2, they use your data and are sneaky about it.
On web3, everything we use is from the blockchain directly. We don’t collect device data which is what web2 is doing in the background
That is a key difference
That is great news. I do wonder how the recommend algorithm will work. Will it focus on showing related posts/comments/videos related to the post/comment/video, or if it will be customized to what the account wants to see.
To be effective, I imagine, it will have to take into account what the account does. How they feed all that in isnt my forte. They might be using LeoAI for that ultimately.
The ultimate version will ask LeoAI to sort content based on the individual user’s interests
That's great. I wonder if a three row recommendation would work. The first row is a recommendation of the user's interests [SPL, Travel, Educational, etc.], the second would be similar content to the one just viewed [they viewed a travel post, so other travel posts are shown], and the third row would be other content from the creator they just viewed.
This is really a suggestion that comes in a very wise timely and which I believe if picked inside will help the Leo
Convenient, quick search for the necessary information is what makes a person return to where he can quickly and efficiently obtain it. !BEER
definitely going to spend some time improving search in the near future
I believe that the Internet has become what we see thanks to search engines and libraries, this is something that should pay off handsomely, I hope :) !BEER