System that can interact with AI.
Posted by: Freya
Date: November 27, 2019 03:04AM

Hi guys,

What do you think of a system that can interact with AIs in man different ways, depending on the type of inference method used and specificity of application.

In a paper - technology overview, we read:

"We can name three main modes:

1. Training data acquisition and processing

In this mode Bitlattice performs an initial gathering and processing of data that is later fed to an AI solution. Primarily in the ones based on neural networks several steps must be accomplished before the actual training can begin. Simplifying, those steps can contain normalization (making data homogenous), feature extraction (gathering prominent patterns) and classification (assigning meanings to patterns). Bitlattice can perform them in one pass while processing incoming data. The added values of using Bitlattice, apart from preparation, are potentially large data gathering network and possibility to decouple actual data from symbolic description passed to AI. Globally accessible training dataset is one of interesting potential applications.

2. Output storage and processing.

This is a rather obvious consequence of Bitlattice being a distributed ledger. Actually distributed ledgers at the moment do a similar job, but using mostly natural (human) intelligence as input. Bitlattice in this mode could accept input from multiple AIs and offer an aggregated and/or processed output. At that point the latter can further deterministically classify, the data provided by AIs and produce a conclusion. Given a large scale and a large number of AIs that could minimize risk of residual poisoning of training data and other forms of manipulation. That mode could also serve to combine different observations using strict criteria.

3. Active component

The compartments that Bitlattice stands on, expose a set of metrics that facilitate operations. Those metrics can also be used as an input for AIs. As they are uniform, there is no need to normalize them. They carry a general picture of the data stored in a compartment and therefore can be used by AIs to infer about the data structure and dynamics.

4. Artificial intelligence itself.

This is probably the most interesting mode. While Bitlattice's middle layer works like a decision tree, other strategies can be implemented as well. Simplifying again, distributed ledgers are networks of peers, but to become neural networks they would need several additional properties, like layered organization, persistent neurons, defined excitation levels, etc.

Bitlattice has or can have implemented instrumentation needed to act as a neural network. That idea is wild, but ultimately possible and potentially beneficial. While the global wide network in this mode won't be fast (due to physical limitations of signals speed and delays of network) the fact that the middle layer contains far less nodes than actual number of participating devices makes that idea at least possible to implement. The practical aspect here could be, for instance, making a "feeling planet" like project.

The above modes can also be mixed and extended, as the domain of AI is very wide and provides plenty of different tools to experiment with."

What is your opinion?

Re: System that can interact with AI.
Posted by: Maria Jhosph
Date: March 03, 2020 09:55PM

This allows them to excel in those areas that conventional computing often finds difficult. Artificial intelligence is the intelligence exhibited by machines or software. Conventional AI research focuses on attempts to mimic human intelligence through symbol manipulation and symbolically structured knowledge bases.A distinction is often made between conventional artificial intelligence, which is based upon machine learning, and computational intelligence, which applies iteration and computation to generate intelligent agents. Conventional AI relies on symbolic logic and formalism and is considered a top-down approach.If you are searching online AI certification course then visit: https://www.cetpainfotech.com/en-ng/artificial-intelligence-training-in-nigeria

The Ways Which We Can Engineer AI
Posted by: Vibhor Tyagi
Date: March 06, 2020 02:54AM

A really swell description. While there are many possible applications of Artificial Intelligence in the broadest terms, this article does a pretty good job of explaining the core ways we can engineer AI. While I do detest the nomenclature (I prefer ML or DL more aptly), this particular field of study is too interesting to ignore.

Engineer.AI

Re: The Ways Which We Can Engineer AI
Posted by: Engineer.AI
Date: March 10, 2020 09:00PM

The way I see it, all you need to engineer AI, are - a deep learning system, language for interaction (C, C++, Python, etc.), and an end-goal. One can engineer AI to do possibly anything, as long as they have a program that responds to DL, with a clear set of commands, and a strong final motive.
Sorry for being a bit philosophical, but apparatus is the least of the worries - it usually boils down to skill, perseverance, and rationale.
~Engineer.AI

Re: The Ways Which We Can Engineer AI
Date: March 10, 2020 09:02PM

You can also do artificial intelligence and use something else then deep learning. Deep learning is a very effective tool but there are many others in AI. :-)

Re: The Ways Which We Can Engineer AI
Posted by: Engineer.AI
Date: March 12, 2020 03:20AM

There are most definitely more modules to engineer AI, there are even a lot of them that are open source and free. Artificial Intelligence (broadly Machine Learning) is only going to expand from here on out. I would encourage anybody who is interested in engineering AI, to do so ASAP. The field is ripe for growth.
~Engineer.AI

Re: The Ways Which We Can Engineer AI
Posted by: Paula
Date: March 13, 2020 05:45AM

How about an AI that can create an AI? Do you guys think it is possible?

Re: The Ways Which We Can Engineer AI
Posted by: Engineer.AI
Date: March 16, 2020 01:50AM

In theory, this technology actually exists right now.

Google Engineer AI that creates other AI

The thing is that although this Artificial Intelligence by Google is quite accurate, it is just the basic version, only more powerful. Google was able to Engineer AI that tracks any and every identifiable object on an image. However, it was limited to just that.
For an AI to engineer AI a more sentient AI, the base machine will have to hold processing powers of ridiculous levels - just as that of a human.
~Vibhor Tyagi (Techie at Engineer.AI)

Re: The Ways Which We Can Engineer AI
Posted by: Paula
Date: March 18, 2020 06:13AM

But how can we compare with the processing power of humans? The brain is not a machine confused smiley How much power do we need? I have so many questionsthat need answers about AI.

This forum is powered by Phorum and provided by P. Fournier-Viger (© 2012).
Terms of use.