We are actually taking a first step toward this now with a Fake News Detection browser plugin for Brave browser, which is being done as part of an alpha version for NuNet.io … Kabir Veitas (kabir@singularitynet.io) is leading this effort … the software dev is being led by Dagim Sisay in our Ethiopia office… but you have a bunch of other good ideas here that could build on the initial app Kabir and Dagim etc. are working on…
2 Likes
Neuralink gets a lot of press attention but I am not currently convinced they are actually at the leading edge of BCI technology. Also they are not very open in their development. I think it might be more workable to make partnerships with some top academic BCI labs, some of which were ahead of Neuralink in many ways when I last checked… Would be great to hear w/ someone active in the BCI community about which academic labs would be ideal for collaboration. What we would really like would be a lab interested in sharing anonymized data in an open way which could then be analyzed using AI tools running on sNet … that might be the place to start… followed afterwards with real-time interaction btw AI agents and BCI devices… Def. a super important area.
I mean, if you think Big Tech controlled social media are bad, wait till you get ahold of Big Tech controlled BCI. This is your brain on AdSense ;-o … We had best get decentralized infotech fully working and widely deployed before large-scale commercial BCI rolls out or it’s going to be a really f**ked up Singularity…
2 Likes
Deep NNs for audio are definitely moving in this direction. Our top-secret simulated Ben Goertzel voice bot is getting better each month and will be inflicted on the world fairly soon 8D … At the moment the best speech models of this nature are big-ass NNs and expensive to run, but I’m sure they are distillable. We have also been looking at ways to make NN generated speech more emotionally expressive, for use in the Awakening Health eldercare robot project.
If any third party devs out there have some cool new tech in these direction, it would be awesome to see them put on the SNet platform and this would certainly be a prime target for DEEP Funding…
1 Like
Hi Ben,
I work with ERP implementations of Oracle Netsuite. We strive to digitalize how corporations work and enable data driven management. Major issues we see today are (i) business process are performed in a very manual way, leading to cost inefficiencies, time delays (on deadlines) and human errors (corrupting the data). (ii) Important business decisions are (due to this) made on bad or even absent data/fact basis. I think the next SAP, Oracle, Dynamics(Microsoft), Netsuite will be in crypto. I also think AI will play an important part in taking ERP (business systems) to the next level. I think developing a solution to integrate with ERP systems and provide AI based business intelligence would be a great way of adapting/introducing both AI and crypto technology to this market.
And once you are integrated with ERP systems I think the possibilities would be endless. I think all business data and process eventually will be automates and analyzed by AI, and I hope to see SingularityNet play a part here:)
This is a good direction for thinking – it would be cool if community members w/ some AI experience could take a careful look at all these AI offerings and think about which might suit SNet best in near term…
Create a SNet AI Which Can Advise People Who Want To Make Use Of AI About Currently Available Options
Since I am not getting many responses from my other posts, maybe this idea will help me work out a sensible way forward for my own various AI-related and interconnected projects - and it seems to me like it would be a good thing for SNet to have anyway? - sort of like an AI enhanced chatbot? (which is one of the things I need for my Avatar Phi Rho).
1 Like
@bengoertzel Is anyone at SNet currently working on the universal AI representations or on developing working memory that SNet AI can utilize?
2 Likes
Have you seen Replika? I’m playing with the beta version. I haven’t done a dive into the development yet. I’ve only interacted with it just to see how it responds. It seems as though they’re trying to teach it a wide variety of emotions.
Have you seen this? I can see this going somewhere if it gets enough attention.
Add this to the real world geographic digital cache effort. EOS has a neat game that has assigned titles to geographic locations to block chain.
Please don’t mind my squirrel brain but I keep seeing how SingulairAGI ADA could be the catalyst to bring great ideas together and benefit everyone.
Given the proof of testing requirements to be considered legitimate, I can see bad actors either changing their toxic behavior to remain a part of the global community or they’ll be left out.
Nice 
Now you mention it, I did listen to a podcast about this from Nature 2.0. I think the main objective was to demonstrate NuNet principles. I didn’t expect you guys to actually follow up on the fake news browser plugin.
That’s what I like so much about this space. Always thinking big, both horizontally and vertically!
Nice idea. There’s a good mix of technology in there, so I think we need to hook out which elements really advantage from SingularityNET, as opposed to blockchain generally. I know there’s a lot of research into using AI to identify plant disease, weather patterns, etc. I think there’s a case to be made for bringing together these diverse types of solutions into something more orchestrated.
EDIT: And scrolling further I see that is precisely what you propose. Nice.
Oh yea, the OpenAI Dota bot recently upgraded to playing as a full 5man team (vs 5human players) and they co-operate, heal each other, protect each other, even bait opponents so that their team mAIte can jump them from hidden positions…
I can see a lot of crossover between the way they implement that, and some of the things SingNET is wanting to prove its capable of doing. If you are curious I’ll PM you some more detailed thoughts on Telegram
How about funding the improvement of the worlds crops by harvesting data at source, using infrared sensors like the above link suggests, maybe SNet could offer AI services to the acadmeics pursuing this interesting research ?
1 Like
xd___
In the OpenCog Hyperon initiative (led by Alexey Potapov and. myself) we are working on an improved version of OpenCog Atomspace which provides both universal AI representations and working memory
Current version of OpenCog is being used for working memory with the Grace eldercare robot (for Awakening Health) and soon for Sophia via the SophiaDAO / Sophia Collective project
The AI-DSL initiative, led by Kabir Veitas and Nil Geisweiller in conjunction w/ some folks at IOHK, is addressing universal representation of properties of AI systems (different from universal representation of knowledge for AI systems, but related). …
So if I am interpreting this correctly, those important areas are already being addressed and those interested in working towards that should probably get in contact with those teams. And so what is needed in addition right now is more useful almost strictly narrow AI projects which Hyperon will be able to make use of?
I guess I am asking: what do you think are the most important types of algorithms Hyperon will need access to in order to get to decentralized dominance that it does not yet have?
I looked at low code platforms a while ago an this just popped back in my head: Create a plugin for Low code platforms such as Mendix or Outsystems to make access to SNET AI services really easy.
I think Mendix is getting a bigger foothold in manufacturing, after being bought by Siemens. Connecting robotics (and logistics and any business process) via low code with SNET could be a major accelerator. Especially if we can add a function to easily create requests for new services right from the low code platform.
3 Likes
Walking through the AI product websites of different cloud providers would give 1) a list of AI services that have gone through a corporate selection process (whatever criteria they use) and 2) some ways of categorizing such services, probably in a two-level hierarchy.
Given that result, the question is then which selection process to apply in order to arrive at AI services worthy of pursuing in community projects. I think we should talk about the criteria flowing into such a filter mechanism. Here are some initial thoughts:
I) Is there an open-source implementation available for the task the commercial AI service is solving? If yes, how performant is it in comparison? I assume there are some cases where a) there is an OS implementation and b) it has similar performance. These should be rather trivial to provide on SingularityNET, it is only a question of getting it to run reliably, or perhaps retraining it on more or newer data. Then I assume there’s the more common case where there is an OS implementation but it lacks a bit behind the state of the art, e.g. open GPT2 versus closed GPT3. Here the question is whether the less performant OS implementation would still be a valuable service with regards to other criteria. And then I assume there’s the case where there’s no OS implementation (or only heavily outdated ones), where it would require serious R&D by a dedicated team of experts, which is probably not within the range of a short-term community-based project. But thinking of this, there’s constantly new code produced by experts in academia, some of which is provided in GitHub repos and/or in Docker containers when a new paper is published or soon afterwards. Could SNET position itself as a simple way for researchers to provide their code in form of a working service, so that other researchers can look at it without having to play around with the tech tooling to reproduce results? That could be especially useful when the research topic is at the boundary of computer science, where scientists from other fields need to have a look at it but don’t have time to tinker with compilation instructions, dependency hell and such.
II) Can the service work together with other services in order to achieve first steps towards cognitive synergy? Combining services opens up a combinatorial space that grows quickly, so it might help to reduce the complexitiy in the beginning by only thinking of cases where two or perhaps three services work together. A simple example could be where service 1 summarizes a text and service 2 does text-to-speech conversion, so that a user can provide some document as input and gets a short audible summary to listen to while in a car or public transport.
III) Can the service attract public attention? Since SingularityNET is in a growth phase, this might be the most important criterium. Are there some services or service combinations that are knowingly of high attraction to a diverse set of people? Since vision has the highest bandwidth of all of our senses, and 3d rendering has motivated the development of dedicated hardware, perhaps going into the direction of some high-quality 3d output could be one viable path. Blender is an incredibly powerful 3d modeling tool and, as far as I know, Unity 3D and Unreal Engine dominate indie game development and deliver code that can be deployed on many platforms (mobile, desktop, gaming consoles, …). So an AI service that constructs and/or controls something within virtual 3d space in these powerful and beautiful environments could draw quite some attention, even more so if passive or active user input is added to the equation, so that it becomes an individualized experience that can be shown to others.
I’m sure there are many more criteria, so a prioritization may also become a necessary topic of discussion. Let me know what you think!
1 Like
Hello all,
I dont know if it is really possible and the ways to do it but I was thinking about AI tools to help decentralisation around casino / poker.
For online poker there is (a growing) problem about cheating (with RTA or bots ) and old problems like collusion and I guess AI could help to detect cheaters.
Could also be AI to help players detect their leaks…
Some decentralised solutions are emerging and imo it is very important for the players to have the best tools for security and comfort. And maybe sophia could connect to it and become the world best robot poker player ? 
I saw there is a lot of competition between chess bots. And I am curious when hyperon will be fully launch on singularitynet could it be possible to create a “better” bot ?
For poker and chess, even if computer can beat humans, it is way too difficult to play 100% like it and could create some problems if you want to improve and follow the “computer advices” without thinking twice about it. So maybe it is possible to find a way to make the outputs less obscure for the end user (maybe with multiple AI ?) ?