₦airaland Forum

The recent releases of state-of-the art Large Language Models (LLMs), the GPT architecture behind chatbots like OpenAI's ChatGPT have developed reasoning capabilities. OpenAI's ChatGPT o1 model, o3 and ChatGPT 4o were tagged reasoning models at launch. Soon, Anthropic, Google, Facebook, Alibaba and Deepseek followed up with their own reasoning models.

Auto-regressive models learn reasoning from the input data when training, algorithmic search and also from other methods like Reinforcement Leaning added to them post-training. With massive amounts of data and billions of parameters (weights or randomized coefficients adjusted during the training process), these models make sense of the data and map relationships they can later query, in high dimensional graphs.

To describe this in a more understandable form, the auto-regressive models correlate their past knowledge or data with present/future solutions they output. If they are then trained or fine-tuned with data that includes the several steps before a solution is arrived at (let's say for Math problems), the models tend to learn these steps and develop what can be described as chain-of-thought process to achieving the results. This is meant to match and internalize the true data generating process during training so the models can be able to understand the steps involved in the overall underlying pattern. Post training, the right way for the models to do the chain-of-thought reasoning, is then reinforced. The reinforcement learning is not only rewarded at output, but also for the right process steps the model takes and right search algorithm choices.

The above method is improved more upon by achieving a sort of artificial self-reflective behavior. This behavior is arrived at through explicit reasoning. The model outlining its own steps as it solves a problem. These steps go back into the model as input and a chain of reasoning is created, which also relates to our initial explanation of the map in a high dimensional space the model has developed to create relationships.
Inference time compute and mixture of experts (MoE) are terms you may come across while studying how these models work. These two methods can optimize a model's accuracy and performance in using the right resource as a strategy.

The mystery or manhunt of the healthcare CEO killer has been all over International news for a few days. New York Police Department has some of the most advanced surveillance and investigation tools and also a large budget and manpower.

New York public spaces, buildings and shops mostly have surveillance cameras which presents the investigators with a massive amount of data to comb through and investigate.

With the help of the type of AI image analysis which narrates activity from different video footages as described in the article below, could the work of the NYPD be done much quicker?

https://medium.com/@boltzmind/could-this-type-of-ai-help-arrest-the-healthcare-ceo-killer-f8041e4ca9c6

AI is the future. Align yourself with the future and help in training AI at https://theiapp.com

Do not miss out on Theia AI project. It is in its early days and you can become an innovator if you register now. AI is the future.
https://theiapp.com

Theia AI project is unique and practical. You don't have to tap anything, you are uploading images and adding labels/annotations that will be used in training computer vision AI. start here https://theiapp.com

You can begin to contribute images to the AI training program https://theiapp.com

No tapping! Upload images and add labels to earn your share of Theia AI https://theiapp.com

Contribute images and labels to earn your shares https://theiapp.com

Theia is a long term AI project but you need to also start contributing to earn your share for when AI becomes very big.
https://theiapp.com

Hello Sunnyrado.
You can register on the app and you upload images with labels indicating the objects in the picture. You can go to your profile and see your Theia points grow with time as you contribute these images and labels.
The images and annotations will be used as data to train AI.

Sunnyrado:
Good evening, am interested in the theiapp but please you will have to put me through on how to go about it...as I am new to it...08163341407

Personally I am focused on training AI and earning a stake of the project through my contribution on images and labels at https://theiapp.com

Gamification brings engagement. People play games for very long stretches of time without any monetary reward. So the monetization on makes doing something we love even better.
Projects like https://theiapp.com are also working on gamifying their image and labels user contributions to train Artificial Intelligence.

Give Theia AI project a try. It is slightly different and gives stake to users for uploading images and adding labels to train AI.
https://theiapp.com

Is it a memecoin and is it endorsed by Elon Musk?

No wonder I see everyone sharing DOGs or CATS or Major on their status. I also launched an AI training group on telegram for https://theipp.com

The UBI idea has a lot of pushback from conservative economists and politicians, there has been a lot of debate around its sustainability and at the very least the idea as proposed seems incomplete. Sam Altman of Open AI recently contributed a UBI experiment in which participants received $1000 monthly payment.

The same problems that arose from the Marxist ideas of an economic Utopia may arise from many of the proposed present day UBI concepts. There are fundamental human needs and innate conditions that cannot be ignored when designing a new system for society. It is not just about economics and the need to earn a living, there are so many other drives and needs required for survival and the stability of a community.

See the possible future outcomes of an AI jobs disruption and the possibility for a Universal Basic Income https://medium.com/@boltzmind/ai-blockchain-and-the-idea-of-universal-basic-income-ubi-de03a796c384

I honestly don't get what it is with these game coins, how do people see the utility? I need to be educated here. Personally I understand AI ,ore and computer vision and know what contribution it is bringing and will soon bring to the world.

That is why I only invest my time and effort in making computer vision AI get smarter and know many different objects in different contexts. I upload and label image training data to https://theiapp.com for a small share of the project.

Internet of things is when devices around us connect to the Internet just like our phones and computers. When we have our watches, cars, bicycles, traffic lights, traffic cameras, street lights, homes, turbines, electricity meters, factory machines and more all connected communicating through the Internet while optimized by AI, that will be great. Imagine the traffic light updating all the nearby cars to stop because there's a cat crossing the road. Life will be awesome.

I love a lot of AI projects that also reward their communities. One that really stands out to me is Theia AI training community https://theiapp.com
They are trying to go big, building novel computer vision models that will be ultimately owned by the community.

I gave it a try. found it even much more convenient than i Thought.

Let's share AI memes. Not AI generated memes.